mpMRI-based habitat analysis for predicting prognoses in patients with high-grade serous ovarian cancer: a multicenter study

mpMRI-based habitat analysis for predicting prognoses in patients with high-grade serous ovarian cancer: a multicenter study | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article mpMRI-based habitat analysis for predicting prognoses in patients with high-grade serous ovarian cancer: a multicenter study Qiu Bi, Kun Miao, Yang Liu, Jing Yang, Ao Zhou, Wenwei Shi, Ying Lei, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6245251/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 29 May, 2025 Read the published version in Abdominal Radiology → Version 1 posted 9 You are reading this latest preprint version Abstract Purpose To evaluate the value of multiparametric MRI (mpMRI)-based habitat analysis for predicting prognoses in patients with high-grade serous ovarian cancer (HGSOC), and to develop combined models by integrating habitat analysis with clinical predictors. Methods This retrospective study included 503 HGSOC patients from four centers. A K-means algorithm was used to identify voxel clusters and generate habitats on mpMRI. Radiomics features were extracted from each habitat sub-region. After feature selection, habitat models were developed to predict overall survival (OS) and progression-free survival (PFS). Cox regression analyses were performed to identify clinical predictors and construct clinical models. Combined models were developed by integrating habitat signatures with clinical predictors. Model performance was evaluated using C-index and time-dependent receiver operating characteristic area under the curves (AUCs). Results Compared with the clinical models (OS: 0.713 and 0.695; PFS: 0.727 and 0.700) and habitat models (OS: 0.707 and 0.672; PFS: 0.627 and 0.641), the combined models integrating habitat features and clinical independent predictors such as neoadjuvant chemotherapy (OS: 0.752 and 0.745; PFS: 0.784 and 0.754) achieved the highest C-indexes for predicting OS and PFS in the internal validation cohort and external test cohort. The combined models also achieved the highest AUCs in all cohorts. Conclusion The habitat models based on mpMRI demonstrated potential value in predicting the prognoses of HGSOC patients, but no significant advantages over the clinical models. The combined models were expected to improve the prognoses from the level of individual clinical characteristics and habitat features reflecting intratumoral heterogeneity. ovarian cancer prognosis MRI habitat radiomics Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 INTRODUCTION Ovarian cancer (OC) is acknowledged as one of the most lethal forms of gynecological cancer, ranking sixth in terms of mortality rates among women [ 1 ]. High-grade serous ovarian cancer (HGSOC) is the predominant subtype, accounting for 70–80% of all OC-related deaths [ 2 , 3 ]. Previous studies have demonstrated that clinicogenomic factors, like BRCA mutation status and the International Federation of Gynecology and Obstetrics (FIGO) stage, are valuable in predicting the prognosis of OC patients [ 4 , 5 ]. However, these factors are neither consistently reliable as independent prognostic indicators nor adequately address the heterogeneity of clinical outcomes. Therefore, there is an urgent need to improve prognostic precision by integrating additional complementary indicators. MRI is a commonly used non-invasive imaging method that offers several advantages over CT and ultrasound, including the absence of ionizing radiation, high soft tissue resolution, and multiparametric imaging capabilities, As a result, it is increasingly utilized in the assessment of OC [ 6 ]. In particular, diffusion-weighted imaging (DWI), which leverages the apparent diffusion coefficient (ADC) metric, has shown significant predictive value for the prognosis of OC patients [ 7 ]. With advancements in artificial intelligence, reseachers have recently developed MRI-based radiomics and deep learning models to predict the prognosis of OC patients with promising results [ 8 , 9 ]. Nevertheless, while these methodologies are capable of extracting general features of the lesion in a high-throughput manner, they fail to adequately capture the spatial heterogeneity that reflects the molecular biological characteristics of tumors. Habitat analysis is an unsupervised machine learning algorithm that can automatically identify sub-regions within different tissues by analyzing voxel similarity, thereby enabling the assessment of intratumoral heterogeneity [ 10 ]. Dextraze et al. [ 11 ] found that spatial habitats based on multiparametric MRI (mpMRI) were associated with the prognosis of glioblastoma patients, and each prognostic related habitat had unique signaling pathway changes, suggesting that the association between image-derived phenotypic measurements and molecular characteristics. Since the tumor microenvironment was closely related to treatment resistance and survival prognosis, Bi et al. [ 12 ] developed a mpMRI-based habitat radiomics model to predict platinum resistance in HGSOC patients, and the predictive performance of the habitat radiomics model was superior to that of conventional radiomics model and deep learning model due to the advantages of combining artificial intelligence and intratumoral heterogeneity, which preliminarily confirmed the value of habitat analysis based on mpMRI for the evaluation of HGSOC. Recent studies have demonstrated that habitat analysis based on PET/CT or CT shows potential for predicting the prognosis of HGSOC patients [ 13 , 14 ]. However, the value of MRI-based habitat analysis in the prognosis of HGSOC warrants further investigation. The purpose of this study is to evaluate the effectiveness of mpMRI-based habitat analysis for predicting the prognosis of HGSOC patients and to develop a combined prognostic model incorporating clinical predictive indicators. MATERIALS AND METHODS Patient cohorts This retrospective cohort study was conducted in accordance with the ethical guidelines in the Helsinki Declaration. The institutional review committees at our center approved this study (reference number: KHLL2023-KY208), and written informed consent was waived. All consecutive patients who were histologically confirmed HGSOC between May 2015 and August 2023 were retrospectively collected for the present analysis. Inclusion criteria comprised individuals meeting the following stipulations: (1) a confirmed diagnosis of HGSOC by operation and pathology; (2) the standard treatment included primary debulking surgery or neoadjuvant chemotherapy (NACT) with interval debulking surgery, followed by postoperative platinum-based chemotherapy; (3) availability of comprehensive pelvic MRI data within a two-week interval preceding the initiation of treatment; (4) no intervention aside from NACT was conducted between the MRI and the surgical procedure; and (5) at least six months of follow-up documentation following postoperative chemotherapy. The exclusion criteria were delineated as follows: (1) patients with a history of other malignant tumors; (2) poor-quality MRI imaging or inadequate registration; (3) the largest diameter of the lesion was under 1 cm; and (4) missing essential clinical data. As a result, this study included a cohort of 503 patients recruited from the specified centers. Patients from Center A were divided into the training cohort (220 cases) and internal validation cohort (94 cases) in a ratio of 7:3. Patients from Centers B (129 cases), C (25 cases), and D (35 cases) were combined into an external test cohort (189 cases). Clinical parameters We collected the following clinical data from the patients: demographic information, treatment plan, laboratory examinations, surgical and histopathological findings, and follow-up records. Overall survival (OS) and progression-free survival (PFS) were defined as the duration from the initial treatment date to the occurrence of death or disease progression, respectively. According to the Response Evaluation Criteria in Solid Tumors (RECIST) guideline (version 1.1) [ 15 ], disease progression is characterized by a rise of at least 20% from the baseline in the total diameters of target lesions, an absolute increment of 5 mm or more, or the appearance of new lesions. Platinum resistance is determined by assessing whether disease progression occurs during the course of platinum-based chemotherapy or within a six-month period following the completion of such treatment [ 16 ]. Image acquisition and segmentation MRI was conducted utilizing either 1.5 T or 3.0 T scanners (GE Signa Pioneer 3.0 T, GE Signa HDxt 1.5 T/3.0 T, Philips Ingenia 1.5 T/3.0 T, and Siemens Prisma/Vida/Skyra 3.0 T). The pelvic MRI images that we need to analyze comprised axial T2-weighted imaging (T2WI), diffusion-weighted imaging (DWI) with a b-value of 1000 s/mm 2 , apparent diffusion coefficient (ADC) maps, and contrast-enhanced T1-weighted imaging (CE-T1WI) at the venous phase. Details of the parameters are presented in Supplementary Table S1 . Certain parameters were modified to better fit the individual needs of each patient. Image preprocessing and segmentation are shown in Supplementary Note 1. Habitat imaging and feature selection The habitat imaging was characterized using four distinct sequences, namely T2WI, DWI, ADC, and CE-T1WI. We recorded all intensity characteristics and procured a feature vector representing the characterization from the four sequences for each patient. Subsequently, the K-means module in the scikit-learn Python package ( https://scikit-learn.org/stable/index.html ) was utilized for clustering habitat sub-regions. Voxels were clustered using the K-means algorithm based on cohort with squared Euclidean distances as the similarity metric and visualized as spatial habitats. The Calinski-Harabasz score was employed to assess the efficacy of clustering setups to determine the optimal number of clusters, spanning from two to ten. The details for K-means clustering and Calinski-Harabasz score are shown in Supplementary Note 2. Details of feature extraction and selection are provided in Supplementary Note 3. Model development The habitat radiomics features were screened using univariate Cox regression and refined with Least Absolute Shrinkage and Selection Operator (LASSO)-Cox for dimensionality reduction, which eliminated irrelevant features by zeroing coefficients based on λ, optimized through 10-fold cross-validation to minimize mean standard error. The habitat models were developed through multivariate Cox regression analyses, utilizing the relevant features for OS and PFS, respectively. Figure 1 illustrates the overall workflow of the habitat models. We conducted univariate and multivariate Cox regression analyses to identify clinical predictors and build clinical models. The combined models were developed via multivariate Cox regression analyses, combining clinical predictors and habitat signatures. Statistical analysis The statistical analysis was conducted utilizing SPSS (version 26.0, IBM, New York, USA), Python (version 3.7.12), X-tile (version 3.6.1), and Onekey AI (version 4.8.4). Categorical variables were presented as counts and percentages, while continuous variables were presented as the median and interquartile range (IQR). Cox proportional hazards models with L2 regularization were used to build prognostic models for predicting OS and PFS. According to the optimal cut-off threshold determined by the X-tile software, the patients were divided into low- and high-risk groups. The performance of the different models was evaluated using Kaplan-Meier survival curves with a C-index. The log-rank test was employed to assess the statistical differences in survival between patients categorized as low-risk and high-risk. A P -value of less than 0.05 was deemed to statistical significance. Finally, the time-dependent receiver operator characteristic (ROC) curve with the area under the curve (AUC) was used to evaluate the predictive performance of different models at a specific time points. RESULTS Clinical parameters The clinical parameters of patients are summarized in Table 1 . Although the start dates varied, all patients were followed up until May 2024. The median OS and PFS were 33.4 (IQR: 16.7–49.3) months and 19.9 (IQR: 12.4–18.9) months for the training patients, 31.8 (IQR: 19.7–49.5) months and 18.5 (IQR: 10.5–30.0) months for the internal validation patients, and 27.3 (IQR: 14.7–38.8) months and 19.4 (IQR: 12.3–30.9) months for the external test patients, respectively. Univariate and multivariate Cox regression analyses of clinical parameters in the training cohort are shown in Table 2 . Clinical predictors of OS included neoadjuvant chemotherapy (NACT) ( P = 0.003, HR = 2.538 [1.372–4.694]), poly ADP-ribose polymerase inhibitor (PARPi) ( P = 0.045, HR = 0.439 [0.196–0.983]), and platinum resistance ( P < 0.001, HR = 3.902 [2.210–6.888]), while the predictors affecting PFS consisted of NACT ( P = 0.011, HR = 1.619 [1.105–2.452]), platinum resistance ( P < 0.001, HR = 7.706 [6.200-14.541]), and laterality of the lesion ( P = 0.028, HR = 1.567 [1.069–2.544]). Table 1 The clinical parameters of HGSOC patients in three cohorts. Parameter Training cohort (N = 220) Internal validation cohort (N = 94) External test cohort (N = 189) Age (years; median IQR) 54.0 (49.0–62.0) 53.5 (48.0–68.0) 55.0 (49.5–63.0) BMI (kg/m 2 ; median IQR) 22.2 (20.2–24.5) 22.5 (20.4–24.2) 23.1 (21.2–25.2) CA125 (U/ml; median IQR) 570.4 (159.2–1687.0) 604.9 (165.1-1392.5) 651.2 (212.0-2101.5) HE4 (pmol/L; median IQR) 365.0 (162.0-736.0) 365.0 (170.5-866.9) 365.0 (190.5-513.5) NACT No Yes 119 (54.1%) 101 (45.9%) 53 (56.4%) 41 (43.6%) 118 (62.4%) 71 (37.6%) PARPi No Yes 151 (68.6%) 69 (31.4%) 51 (54.3%) 43 (45.7%) 97 (51.3%) 92 (48.7%) Platinum resistant No Yes 171 (77.7%) 49 (22.3%) 68 (72.3%) 26 (27.7%) 161 (85.2%) 28 (14.8%) Lymph node dissection No Yes 74 (33.6%) 146 (66.4%) 132 (34.0%) 62 (66.0%) 113 (59.8%) 76 (40.2%) Laterality Unilateral Bilateral 75 (34.1%) 145 (65.9%) 23 (24.5%) 71 (75.5%) 75 (39.7%) 114 (60.3%) Residual tumor R0 R1 R2 211 (95.9%) 2 (0.9%) 7 (3.2%) 89 (94.7%) 2 (2.1%) 3 (3.2%) 172 (91.0%) 5 (2.6%) 12 (6.3%) FIGO stage I II III IV 10 (4.5%) 5 (2.3%) 119 (54.1%) 86 (39.1%) 3 (3.2%) 3 (3.2%) 59 (62.8%) 29 (30.9%) 8 (4.2%) 24 (12.7%) 121 (64.0%) 36 (19.0%) Diabetes No Yes 210 (95.5%) 10 (4.5%) 90 (95.7%) 4 (4.3%) 182 (96.3%) 7 (3.7%) Progression No Yes 113 (51.4%) 107 (48.6%) 41 (43.6%) 53 (56.4%) 115 (60.8%) 74 (39.2%) Death No Yes 170 (77.3%) 50 (22.7%) 68 (72.3%) 26 (27.7%) 168 (88.9%) 21 (11.1%) PFS (months; median IQR) 19.9 (12.4–18.9) 18.5 (10.5–30.0) 19.4 (12.3–30.9) OS (months; median IQR) 33.4 (16.7–49.3) 31.8 (19.7–49.5) 27.3 (14.7–38.8) HGSOC: high-grade serous ovarian cancer; BMI: body mass index; NACT: neoadjuvant chemotherapy; PARPi: poly ADP-ribose polymerase inhibitor; R0: no macroscopic tumor; R1: residual tumor < 1 cm; R2: residual tumor ≥ 1 cm; FIGO: Federation of Gynecology and Obstetrics; OC: ovarian cancer; PFS: progression-free survival; OS: overall survival; IQR: inter quartile range. * P < 0.05. Table 2 Univariate and multivariate Cox regression analyses of clinical parameters in the training cohort. Parameter OS PFS Univariate Multivariate Univariate Multivariate HR (95% CI) P HR (95% CI) P HR (95% CI) P HR (95% CI) P Age (years) 1.000 (0.975–1.026) 0.995 0.990 (0.971–1.009) 0.285 BMI (kg/m 2 ) 0.998 (0.934–1.066) 0.945 0.979 (0.933–1.027) 0.385 CA125 (U/ml) 1.000 (1.000–1.000) 0.503 1.000 (1.000–1.000) 0.284 HE4 (pmol/L) 1.000 (1.000–1.000) 0.842 1.000 (1.000–1.000) 0.577 NACT 2.135 (1.332–3.425) 0.002* 2.538 (1.372–4.694) 0.003* 2.177 (1.525–3.109) < 0.001* 1.619 (1.105–2.452) 0.011* PARPi 0.556 (0.320–0.967) 0.038* 0.439 (0.196–0.983) 0.045* 0.753 (0.508–1.117) 0.158 Platinum resistant 3.404 (2.076–5.579) < 0.001* 3.902 (2.210–6.888) < 0.001* 8.673 (5.818–12.928) < 0.001* 7.706 (6.200-14.541) < 0.001* Lymph node dissection 1.137 (0.696–1.857) 0.609 0.933 (0.652–1.337) 0.707 Diabetes 1.017 (0.319–3.240) 0.978 1.031 (0.453–2.346) 0.943 Laterality 1.638 (0.983–2.729) 0.058 1.841 (1.247–2.717) 0.002* 1.567 (1.069–2.544) 0.028* FIGO stage 1.276 (0.909–1.789) 0.159 1.283 (0.998–1.648) 0.051 Residual tumor 1.147 (0.664–1.980) 0.623 1.068 (0.684–1.669) 0.772 HR: Hazard risk; CI: confidence interval. * P < 0.05. Habitat imaging and feature selection The Calinski-Harabasz score indicated that the ideal quantity of clusters was four. Therefore, all lesions were divided into no more than four habitat sub-regions (Supplementary Figure S1 ). Supplementary Table S2 presents the mean signal intensity (SI) on different sequences, as well as the volume, and proportion of each habitat. Habitats 1 and 2 showed low ADC SIs and moderate or obvious enhancement. Habitat 3 had the highest SIs on both ADC maps and T2WI, while Habitat 4 presented the lowest SIs on ADC maps and CE-T1WI. In terms of sub-region proportions, Habitat 1 was the most predominant (34.7%), followed by habitat 3 (32.0%) . We extracted 1743 features from each sequence and improved the feature set by combining the features from four sequences, leading to a total of 27888 features, calculated as four sequences multiplied by four habitats and 1743 features. After feature reduction and LASSO-Cox selection (Supplementary Figure S2), 45 features and 62 features were retained to build models for predicting OS and PFS, respectively. For predicting OS, the top two features with the highest weight were derived from Habitat 2 on DWI and ADC maps (Supplementary Figure S3), and for predicting PFS, they were derived from Habitat 1 on T2WI and Habitat 2 on ADC maps (Supplementary Figure S4). Performance of different models Table 3 and Fig. 2 – 4 present the C-index and Kaplan-Meier survival curves of the different models, respectively. For predicting OS, the C-indexes of the clinical model, habitat model, and combined model in the training cohort were 0.776, 0.732, and 0.827, respectively (all P < 0.001); in the internal validation cohort, they were 0.713, 0.707, and 0.752, respectively (all P < 0.05); and in the external test cohort, they were 0.695 ( P = 0.065), 0.672 ( P < 0.001), and 0.745 ( P < 0.001), respectively. For predicting PFS, the C-indexes of the clinical model, habitat model, and combined model in the training cohort were 0.799, 0.791, and 0.874, respectively (all P < 0.001); in the internal validation cohort, they were 0.727, 0.627, and 0.784, respectively (all P < 0.05); in the external test cohort, they were 0.700, 0.641, and 0.754, respectively (all P < 0.05). In summary, for both OS and PFS predictions, the C-indexes of habitat models were lower than those of the clinical models, while the combined models demonstrated the highest C-indexes. Table 3 C-index of the different models. Cohort Clinical model Habitat model Combined model C-index P C-index P C-index P OS Training 0.776 < 0.001* 0.732 < 0.001* 0.827 < 0.001* Internal validation 0.713 < 0.001* 0.707 0.015* 0.752 < 0.001* External test 0.695 0.065 0.672 < 0.001* 0.745 < 0.001* PFS Training 0.799 < 0.001* 0.791 < 0.001* 0.874 < 0.001* Internal validation 0.727 0.011* 0.627 0.020* 0.784 < 0.001* External test 0.700 < 0.001* 0.641 0.002* 0.754 < 0.001* * P < 0.05. Table 4 and Fig. 5 presents the performance and ROC curves of the different models for predicting the risk of death or progression at 3 years, respectively. For predicting OS, the AUCs of the clinical model, habitat model, and combined model in the training cohort were 0.782 (0.693–0.871), 0.791 (0.706–0.877), and 0.856 (0.782–0.929), respectively; in the internal validation cohort, they were 0.759 (0.615–0.904), 0.760 (0.626–0.893), and 0.797 (0.666–0.928), respectively; and in the external test cohort, they were 0.760 (0.618–0.902), 0.747 (0.599–0.896), and 0.821 (0.705–0.937), respectively. For predicting PFS, the AUCs of the clinical model, habitat model, and combined model in the training cohort were 0.855 (0.798–0.912), 0.893 (0.846–0.941), and 0.948 (0.917–0.978), respectively; in the internal validation cohort, they were 0.778 (0.671–0.885), 0.660 (0.509–0.812), and 0.851 (0.759–0.944), respectively; and in the external test cohort, they were 0.668 (0.561–0.775), 0.788 (0.683–0.893), and 0.791 (0.696–0.885), respectively. Similarly, for both OS and PFS predictions, the combined models demonstrated the highest AUCs. Table 4 The performance of the different models in predicting 3-year risk of death or progression. Cohort Model AUC (95% CI) Accuracy Sensitivity Specificity PPV NPV OS Training Clinical 0.782 (0.693–0.871) 0.672 0.621 0.824 0.914 0.418 Habitat 0.791 (0.706–0.877) 0.730 0.718 0.765 0.902 0.473 Combined 0.856 (0.782–0.929) 0.781 0.777 0.794 0.920 0.540 Internal validation Clinical 0.759 (0.615–0.904) 0.759 0.850 0.556 0.810 0.625 Habitat 0.760 (0.626–0.893) 0.707 0.675 0.778 0.871 0.519 Combined 0.797 (0.666–0.928) 0.810 0.825 0.778 0.892 0.667 External test Clinical 0.760 (0.618–0.902) 0.757 0.778 0.636 0.925 0.333 Habitat 0.747 (0.599–0.896) 0.527 0.444 1.000 1.000 0.239 Combined 0.821 (0.705–0.937) 0.770 0.762 0.818 0.960 0.375 PFS Training Clinical 0.855 (0.798–0.912) 0.781 0.700 0.844 0.778 0.784 Habitat 0.893 (0.846–0.941) 0.838 0.871 0.811 0.782 0.890 Combined 0.948 (0.917–0.978) 0.863 0.900 0.833 0.808 0.915 Internal validation Clinical 0.778 (0.671–0.885) 0.677 0.737 0.652 0.467 0.857 Habitat 0.660 (0.509–0.812) 0.692 0.421 0.804 0.471 0.771 Combined 0.851 (0.759–0.944) 0.708 0.947 0.609 0.500 0.966 External test Clinical 0.668 (0.561–0.775) 0.602 0.750 0.530 0.436 0.814 Habitat 0.788 (0.683–0.893) 0.806 0.594 0.909 0.760 0.822 Combined 0.791 (0.696–0.885) 0.786 0.500 0.924 0.762 0.792 AUC: area under the curve; PPV: positive predictive value; NPV: negative predictive value. DISCUSSION In this study, we divided tumors into four habitats based on the K-means algorithm, constructed habitat models using the radiomics features of each habitat sub-region, and developed a combined model incorporating clinical predictors. We found that the predictive performance of the habitat models based on mpMRI did not surpass that of the clinical models. However, the combined models exhibited consistent and superior prognostic performance across all cohorts. This consistency highlights the reliability and clinical applicability of the combined models, which may assist healthcare professionals in developing personalized treatment strategies for HGSOC patients. In this study, NACT and platinum resistance were found to be predictors of OS and PFS. Due to tumor burden or genetic mutations induced by chemotherapy, some HGSOC lesions continued to grow after NACT, leading to a poor prognosis [ 17 ]. Elyashiv et al. [ 18 ] reported that platinum resistance was a significant prognostic indicator in females diagnosed with epithelial OC. Once platinum resistance occurred, both patients with primary platinum resistance and those with secondary platinum resistance had lower survival rates. In addition, we observed that maintenance therapy with PARPi was associated with OS, and the laterality of lesions influenced PFS. Yamada et al. [ 19 ] reported that the PFS of OC patients was longer in the unilateral group than in the bilateral group, confirming that tumor laterality could be an independent prognostic factor in OC patients. According to the American Society of Clinical Oncology (ASCO) guidelines [ 20 ], the application of PARPi after complete and partial responses to initial therapy or platinum-sensitive relapse therapy notably prolonged PFS in advanced OC patients. However, PARPi maintenance therapy did not significantly prolong OS, likely due to the presence of adverse effects and other factors [ 21 ]. In addition, the duration of maintenance treatment and its impact on prognosis require normalization and validation in real-world settings with large, multicenter sample sizes [ 22 ]. Beyond the factors mentioned above, residual tumor, FIGO stage, and CA125 levels, among others, are generally considered prognostic factors [ 5 , 23 ]. However, none of these indicators were identified as independent prognostic factors in our study. One possible explanation is that differences in patient distribution may have affected the results of the Cox regression analyses, such as the relatively small number of FIGO stage I-II and non-R0 patients. Additionally, individuals classified under the same FIGO stage often exhibit differing survival outcomes, and the clinical validity of indicators such as CA125 remains controversial [ 24 ]. The present study demonstrated that low-cost clinical models might serve as promising prognostic tools for forecasting survival outcomes in HGSOC patients. However, consistent with the findings of Huang et al. [ 25 ], the C-index of the clinical models was not sufficiently high in the external test cohort. This underscores the important of integrating imaging data and other methods to comprehensively evaluate the prognosis of OC patients. Radiomics, a crucial link connecting medical imaging with personalized medicine, enhances the diagnostic, prognostic, and predictive precision of cancer assessments through the extensive extraction and analysis of quantitative imaging features obtained from medical imaging techniques [ 26 ]. Due to the limited interpretability of radiomics, although many predictive models have been proposed, their connection to biologically relevant factors has rarely been elucidated [ 27 ]. Habitat radiomics combines the advantages of radiomics and spatial heterogeneity to successfully evaluate Ki-67 expression and PFS in HGSOC patients, and demonstrates that the habitat model can better stratify the prognosis in radiomics models [ 14 ]. According to the log-rank test, our habitat models also showed significant differences in stratification between the high- and low-risk groups across all cohorts. However, the C-indexes of our habitat models for predicting OS and PFS in the external test cohort were 0.672 and 0.641, respectively, which were similar to those in the previous study (C-index = 0.61) [ 28 ], indicating that image-based artificial intelligence analyses alone remain insufficient to accurately predict the prognosis of HGSOC patients. In addition, like conventional radiomics, the process of habitat radiomics also involves complicated steps such as feature extraction and reduction, so the studies of radiomics are still in the scientific research stage. At present, it is not possible to directly obtain the prognostic outcomes of patients simply by inputting images in clinical practice. However, with the development of artificial intelligence, this study may provide the basis for the development of prognostic software in the future. In addition to performing habitat imaging, the combination of radiological data with other data, such as clinical parameters, is also the alternative solutions to overcome the limitation of poor interpretability [ 27 ]. A series of high-quality studies have shown that the integration of multimodal data is beneficial for complementing tumor heterogeneity across multiple scales, providing complementary predictive information, and enhancing the predictive capability of unimodal models [ 29 – 31 ]. Boehm et al. [ 30 ] developed a machine learning integrated model by integrating clinical and CT features, which could effectively improve prognostic risk stratification in HGSOC patients. Bi et al. [ 12 ] concluded that a comprehensive nomogram combining habitat signature and clinical features had the best performance in identifying platinum-resistant patients with HGSOC, superior to both the clinical model and the habitat model. This study also found that the combined model had the best prognostic performance in HGSOC patients, regardless of whether it was C-indexes or time-dependent AUCs. In future clinical practice, when clinical parameters alone were not enough to accurately predict the prognosis of HGSOC patients, MRI-based habitat imaging should be added. Moreover, we hope to standardize and simplify the process of developing the combined models in the future. The specific steps include [ 32 ]: Initially, a standardized pelvic MRI scanning protocol will be implemented within a two-week window prior to the surgical intervention. Subsequently, by amalgamating data from the hospital information system alongside the Picture Archiving and Communication System, the integrated software will autonomously compute intratumoral heterogeneity metrics, while also delivering insights regarding risk stratification, potential prognostic outcomes, and individualized treatment options. Another recent study reported that the multimodal model based on CT for predicting prognosis of epithelial OC demonstrated a C-index of 0.64 in the external test set [ 28 ]. The C-indexes of our combined models based on mpMRI were around 0.75. Furthermore, Huang et al. [ 13 ] developed intratumoral and peritumoral radiomics models based on CT to predict the prognosis of HGSOC patients, with AUCs ranging from 0.552 to 0.777 in the internal test set. In present study, the AUCs of the combined models were above 0.79 in the internal validation and external test cohorts. These results indirectly may suggest that the performance of the combined model combining MRI habitat and clinical parameters in predicting the prognosis of HGSOC patients is superior and more stable than that of CT. Of course, preoperative CT and mpMRI data from more centers with larger samples should be included together for future analyses to verify this conclusion. Interestingly, we found that the top two features with the highest weight to the habitat models both contained features of Habitat 2 on the ADC maps. Recent studies emphasized the crucial role of the enhancing tumor or solid tumor, and the ADC values of these lesions might more accurately represent the heterogeneity of OC [ 33 – 35 ]. Habitats 1 and 2 showed low ADC SIs and moderate or obvious enhancement, indicating that the main component of these habitat was solid. Our present study suggested that the solid components of HGSOC lesions and the features derived from ADC maps have a greater impact on prognosis. Zhang et al. [ 7 ] suggested that the ADC values of functional tumor volume (the volume of solid mass) could be used to assess preoperative prognostic factors in epithelial OC, which further supported our results and speculation. However, the habitat sub-regions generated based on unsupervised clustering do not completely represent the actual pathological tissues and need to be prospectively verified by point-to-point analyses on MRI and pathological images. This study has some limitations. First, the retrospective design inherently carries the risk of selection bias, because only patients with available MRI data and those with valid follow-up data were included, and more patients who only underwent CT before surgery and patients lost to follow-up were not included. Additionally, we only incorporated four MRI sequences due to the limitations of retrospective design, and other functional MRI, such as diffusion kurtosis imaging, need to be explored further. Moreover, the lack of standardization in MRI equipment and parameters across different hospitals may lead to variability in model performance between different cohorts. Second, the follow-up duration may not be sufficient to comprehensively assess long-term outcomes, highlighting the need for future studies with larger sample sizes and extended follow-up periods to validate the prognostic significance of our models in diverse patient populations. Third, manual delineation remains the gold standard of image segmentation, but due to the large size and extensive range of HGSOC lesions, the workload for manual segmentation is substantial. As a result, the habitat radiomics features were not excluded through intraclass correlation coefficient testing. We are currently working on developing an automated and accurate tumor segmentation method to enhance the efficiency and stability of the model. In conclusion, this study highlights the critical role of combined models that integrate clinical features and habitat radiomics features in predicting OS and PFS in HGSOC patients. The findings may aid in clinical decision-making and foster personalized treatment strategies. Future research should focus on the prospective validation and standardization of MRI acquisition protocols and pathological validation of habitat biomarkers, as well as increasing follow-up time to further improve predictive accuracy and ultimately enhance the prognosis for HGSOC patients. Declarations Author Contribution Drafting of the manuscript: Q.B.Concept and design: Q.B. and CH.A.Acquisition, analysis, or interpretation of data: Q.B., K.M., Y.Liu., J.Y., A.Z., WW.S., Y.Lei., YZ.W. and Y.S.Critical revision of the manuscript for important intellectual content: HM.L. and JW.Q.Statistical analysis: Q.B. and K.M.Supervision: JWQ. All authors have accessed and verified the underlying data, and takes responsibility for the integrity of the data and the accuracy of the data analysis. Acknowledgement This work was supported by the National Natural Science Foundations of China [grant numbers 82460340, 82471943, 82471932, 82271940, 82160524]; Kunming University of Science and Technology & the First People's Hospital of Yunnan Province Joint Special Project on Medical Research [grant number KUST-KH2022027Y], the Basic Research on Application of Joint Special Funding of Science and Technology Department of Yunnan Province-Kunming Medical University [grant number 202301AY070001-084], Shanghai Jinshan District Health Committee [grant number JSZK2023A02], and Natural Science Foundation of Shanghai [grant number 22ZR1412500] Data Availability Due to the privacy of patients, the data related to patients cannot be available for public access but can be obtained from the corresponding author on reasonable request approved by the institutional review board of the four centers. References Siegel RL, Giaquinto AN, Jemal A (2024) Cancer statistics, 2024. Ca-Cancer J Clin 74:12–49. https://doi.org/10.3322/caac.21820 . Kurman RJ, Shih I (2016) The Dualistic Model of Ovarian Carcinogenesis: Revisited, Revised, and Expanded. Am J Pathol 186:733–747. https://doi.org/10.1016/j.ajpath.2015.11.011 . Bowtell DD, Bohm S, Ahmed AA, et al. (2015) Rethinking ovarian cancer II: reducing mortality from high-grade serous ovarian cancer. Nat Rev Cancer 15:668–679. https://doi.org/10.1038/nrc4019 . Gallagher DJ, Konner JA, Bell-McGuinn KM, et al. (2011) Survival in epithelial ovarian cancer: a multivariate analysis incorporating BRCA mutation status and platinum sensitivity. Ann Oncol 22:1127–1132. https://doi.org/10.1093/annonc/mdq577 . Irodi A, Rye T, Herbert K, et al. (2020) Patterns of clinicopathological features and outcome in epithelial ovarian cancer patients: 35 years of prospectively collected data. BJOG-Int J Obstet Gy 127:1409–1420. https://doi.org/10.1111/1471-0528.16264 . Daoud T, Sardana S, Stanietzky N, Klekers AR, Bhosale P, Morani AC (2022) Recent Imaging Updates and Advances in Gynecologic Malignancies. Cancers 14:5528. https://doi.org/10.3390/cancers14225528 . Zhang C, Ma L, Zhao Y, et al. (2024) Estimating pathological prognostic factors in epithelial ovarian cancers using apparent diffusion coefficients of functional tumor volume. Eur J Radiol 176:111514. https://doi.org/10.1016/j.ejrad.2024.111514 . Li H, Cai S, Deng L, et al. (2023) Prediction of platinum resistance for advanced high-grade serous ovarian carcinoma using MRI-based radiomics nomogram. Eur Radiol 33:5298–5308. https://doi.org/10.1007/s00330-023-09552-w . Liu L, Wan H, Liu L, Wang J, Tang Y, Cui S, Li Y (2023) Deep Learning Provides a New Magnetic Resonance Imaging-Based Prognostic Biomarker for Recurrence Prediction in High-Grade Serous Ovarian Cancer. Diagnostics 13:748. https://doi.org/10.3390/diagnostics13040748 . O'Connor JP, Rose CJ, Waterton JC, Carano RA, Parker GJ, Jackson A (2015) Imaging intratumor heterogeneity: role in therapy response, resistance, and clinical outcome. Clin Cancer Res 21:249–257. https://doi.org/10.1158/1078-0432.CCR-14-0990 . Dextraze K, Saha A, Kim D, et al. (2017) Spatial habitats from multiparametric MR imaging are associated with signaling pathway activities and survival in glioblastoma. Oncotarget 8:112992–113001. https://doi.org/10.18632/oncotarget.22947 . Bi Q, Miao K, Xu N, et al. (2024) Habitat Radiomics Based on MRI for Predicting Platinum Resistance in Patients with High-Grade Serous Ovarian Carcinoma: A Multicenter Study. Acad Radiol 31:2367–2380. https://doi.org/10.1016/j.acra.2023.11.038 . Huang X, Huang Y, Liu K, Zhang F, Zhu Z, Xu K, Li P (2024) Intratumoral and Peritumoral Radiomics for Predicting the Prognosis of High-grade Serous Ovarian Cancer Patients Receiving Platinum-Based Chemotherapy. Acad Radiol. https://doi.org/10.1016/j.acra.2024.09.001 . Wang X, Xu C, Grzegorzek M, Sun H (2022) Habitat radiomics analysis of pet/ct imaging in high-grade serous ovarian cancer: Application to Ki-67 status and progression-free survival. Front Physiol 13:948767. https://doi.org/10.3389/fphys.2022.948767 . Eisenhauer EA, Therasse P, Bogaerts J, et al. (2009) New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer 45:228–247. https://doi.org/10.1016/j.ejca.2008.10.026 . Wang T, Tang J, Yang H, et al. (2022) Effect of Apatinib Plus Pegylated Liposomal Doxorubicin vs Pegylated Liposomal Doxorubicin Alone on Platinum-Resistant Recurrent Ovarian Cancer: The APPROVE Randomized Clinical Trial. JAMA Oncol 8:1169–1176. https://doi.org/10.1001/jamaoncol.2022.2253 . Himoto Y, Cybulska P, Shitano F, et al. (2019) Does the method of primary treatment affect the pattern of first recurrence in high-grade serous ovarian cancer? Gynecol Oncol 155:192–200. https://doi.org/10.1016/j.ygyno.2019.08.011 . Elyashiv O, Aleohin N, Migdan Z, Leytes S, Peled O, Tal O, Levy T (2024) The Poor Prognosis of Acquired Secondary Platinum Resistance in Ovarian Cancer Patients. Cancers 16:641. https://doi.org/10.3390/cancers16030641 . Yamada Y, Mabuchi S, Kawahara N, Kawaguchi R (2021) Prognostic significance of tumor laterality in advanced ovarian cancer. Obstet Gynecol Sci 64:524–531. https://doi.org/10.5468/ogs.21176 . Tew WP, Lacchetti C, Ellis A, et al. (2020) PARP Inhibitors in the Management of Ovarian Cancer: ASCO Guideline. J Clin Oncol 38:3468–3493. https://doi.org/10.1200/JCO.20.01924 . Kim JH, Kim SI, Park EY, et al. (2023) Impact of postoperative residual disease on survival in epithelial ovarian cancer with consideration of recent frontline treatment advances: A systematic review and meta-analysis. Gynecol Oncol 179:24–32. https://doi.org/10.1016/j.ygyno.2023.10.018 . Tuninetti V, Marin-Jimenez JA, Valabrega G, Ghisoni E (2024) Long-term outcomes of PARP inhibitors in ovarian cancer: survival, adverse events, and post-progression insights. ESMO Open 9:103984. https://doi.org/10.1016/j.esmoop.2024.103984 . Salminen L, Nadeem N, Jain S, et al. (2020) A longitudinal analysis of CA125 glycoforms in the monitoring and follow up of high grade serous ovarian cancer. Gynecol Oncol 156:689–694. https://doi.org/10.1016/j.ygyno.2019.12.025 . Zhang M, Cheng S, Jin Y, Zhao Y, Wang Y (2021) Roles of CA125 in diagnosis, prediction, and oncogenesis of ovarian cancer. Bba-Rev Cancer 1875:188503. https://doi.org/10.1016/j.bbcan.2021.188503 . Huang W, Bao Y, Luo X, Yao L, Yuan L (2022) Novel prognostic nomograms to assess survival in high-grade serous ovarian carcinoma after surgery and chemotherapy: a retrospective cohort study from SEER database. Ann Transl Med 10:728. https://doi.org/10.21037/atm-21-4383 . Lambin P, Leijenaar R, Deist TM, et al. (2017) Radiomics: the bridge between medical imaging and personalized medicine. Nat Rev Clin Oncol 14:749–762. https://doi.org/10.1038/nrclinonc.2017.141 . Russo L, Bottazzi S, Sala E (2024) Artificial intelligence in female pelvic oncology: tailoring applications to clinical needs. Eur Radiol 34:4038–4040. https://doi.org/10.1007/s00330-023-10455-z . Huang X, Huang Y, Liu K, Zhang F, Zhu Z, Xu K, Li P (2024) Predicting prognosis for epithelial ovarian cancer patients receiving bevacizumab treatment with CT-based deep learning. NPJ Precis Oncol 8:202. https://doi.org/10.1038/s41698-024-00688-6 . Chen Z, Chen Y, Sun Y, et al. (2024) Predicting gastric cancer response to anti-HER2 therapy or anti-HER2 combined immunotherapy based on multi-modal data. Signal Transduct Tar 9:222. https://doi.org/10.1038/s41392-024-01932-y . Boehm KM, Aherne EA, Ellenson L, et al. (2022) Multimodal data integration using machine learning improves risk stratification of high-grade serous ovarian cancer. Nat Cancer 3:723–733. https://doi.org/10.1038/s43018-022-00388-9 . Sammut SJ, Crispin-Ortuzar M, Chin SF, et al. (2022) Multi-omic machine learning predictor of breast cancer therapy response. Nature 601:623–629. https://doi.org/10.1038/s41586-021-04278-5 . He D, Zhang X, Chang Z, Liu Z, Li B (2024) Survival time prediction in patients with high-grade serous ovarian cancer based on (18)F-FDG PET/CT- derived inter-tumor heterogeneity metrics. BMC Cancer 24:337. https://doi.org/10.1186/s12885-024-12087-y . Borde T, Nezami N, Laage GF, et al. (2022) Optimization of the BCLC Staging System for Locoregional Therapy for Hepatocellular Carcinoma by Using Quantitative Tumor Burden Imaging Biomarkers at MRI. Radiology 304:228–237. https://doi.org/10.1148/radiol.212426 . Li HM, Zhang R, Gu WY, et al. (2019) Whole solid tumour volume histogram analysis of the apparent diffusion coefficient for differentiating high-grade from low-grade serous ovarian carcinoma: correlation with Ki-67 proliferation status. Clin Radiol 74:918–925. https://doi.org/10.1016/j.crad.2019.07.019 . Mimura R, Kato F, Tha KK, et al. (2016) Comparison between borderline ovarian tumors and carcinomas using semi-automated histogram analysis of diffusion-weighted imaging: focusing on solid components. Jpn J Radiol 34:229–237. https://doi.org/10.1007/s11604-016-0518-6 . Additional Declarations No competing interests reported. Supplementary Files Supplementalmaterials.docx Cite Share Download PDF Status: Published Journal Publication published 29 May, 2025 Read the published version in Abdominal Radiology → Version 1 posted Editorial decision: Revision requested 01 May, 2025 Reviews received at journal 01 May, 2025 Reviewers agreed at journal 24 Apr, 2025 Reviews received at journal 19 Apr, 2025 Reviewers agreed at journal 20 Mar, 2025 Reviewers invited by journal 19 Mar, 2025 Editor assigned by journal 18 Mar, 2025 Submission checks completed at journal 18 Mar, 2025 First submitted to journal 17 Mar, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6245251","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":432953105,"identity":"bf7bc4a9-0f32-4fc1-93df-77287473b304","order_by":0,"name":"Qiu Bi","email":"","orcid":"","institution":"Jinshan Hospital of Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Qiu","middleName":"","lastName":"Bi","suffix":""},{"id":432953107,"identity":"cccf64fe-6bfe-4a62-8b82-eebf70d50f0e","order_by":1,"name":"Kun Miao","email":"","orcid":"","institution":"First People's Hospital of Yunnan Province","correspondingAuthor":false,"prefix":"","firstName":"Kun","middleName":"","lastName":"Miao","suffix":""},{"id":432953114,"identity":"fb1f9be5-feaf-42d5-b6fa-b701711c57ed","order_by":2,"name":"Yang Liu","email":"","orcid":"","institution":"First Affiliated Hospital of Chongqing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yang","middleName":"","lastName":"Liu","suffix":""},{"id":432953117,"identity":"95cd1977-2b55-41f0-990e-54d5f1bae08d","order_by":3,"name":"Jing Yang","email":"","orcid":"","institution":"First Affiliated Hospital of Chongqing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Jing","middleName":"","lastName":"Yang","suffix":""},{"id":432953118,"identity":"53597471-ed2d-42cd-a347-f175a4a26dc3","order_by":4,"name":"Ao Zhou","email":"","orcid":"","institution":"First Affiliated Hospital of Chongqing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Ao","middleName":"","lastName":"Zhou","suffix":""},{"id":432953120,"identity":"f2df47be-008a-4305-b68f-e2471fb9b416","order_by":5,"name":"Wenwei Shi","email":"","orcid":"","institution":"Zhongda Hospital Southeast University","correspondingAuthor":false,"prefix":"","firstName":"Wenwei","middleName":"","lastName":"Shi","suffix":""},{"id":432953121,"identity":"07dfeec3-8638-4e2c-9329-6483d7a44ccd","order_by":6,"name":"Ying Lei","email":"","orcid":"","institution":"Zhongda Hospital Southeast University","correspondingAuthor":false,"prefix":"","firstName":"Ying","middleName":"","lastName":"Lei","suffix":""},{"id":432953122,"identity":"f9f6adc9-96c0-49f3-8729-95b1291d9c9b","order_by":7,"name":"Yunzhu Wu","email":"","orcid":"","institution":"Zhongda Hospital Southeast University","correspondingAuthor":false,"prefix":"","firstName":"Yunzhu","middleName":"","lastName":"Wu","suffix":""},{"id":432953124,"identity":"63e7dd35-82d8-450f-bad5-74e157834891","order_by":8,"name":"Yang Song","email":"","orcid":"","institution":"Siemens Healthineers (China)","correspondingAuthor":false,"prefix":"","firstName":"Yang","middleName":"","lastName":"Song","suffix":""},{"id":432953125,"identity":"de0eae17-64b0-4b86-97d7-d9e1a4178c3e","order_by":9,"name":"Conghui Ai","email":"","orcid":"","institution":"Yunnan Cancer Hospital","correspondingAuthor":false,"prefix":"","firstName":"Conghui","middleName":"","lastName":"Ai","suffix":""},{"id":432953126,"identity":"113c9e49-82ce-464b-ae5e-ba2a56d1fa52","order_by":10,"name":"Haiming Li","email":"","orcid":"","institution":"Fudan University Shanghai Cancer Center","correspondingAuthor":false,"prefix":"","firstName":"Haiming","middleName":"","lastName":"Li","suffix":""},{"id":432953128,"identity":"d5670cab-ef8a-4192-a60d-0f17c1d8ba94","order_by":11,"name":"Jingwei Qiang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA9UlEQVRIiWNgGAWjYDCCAwwMBglAmo2BsQFI2fDw8zeQpiVNRnLGAcJakMFhG4OGBPw6+G4fYCh4uOOwHB97cuPngl/neQwYDjB++JiDW4vkuQQGg8Qzh43ZeB42S8/su81jztzALDlzG24tBmeAOLHtdmKbRGKDNG/PbR7LhgNszLxEaKkHamn+zdtzjsfgQAJxWhLYJBLbpHl+HCCsRfIMYwNQy3/DNp6Hbda8Dck8kjMONuP1C98Z5mOGP9vS5OXb0x/f5vljZ8/P33zww0c8WhgYGNsMIIwEEBss0oBPPQgwP4BrYfhDSPEoGAWjYBSMRAAA549T8RUPSI8AAAAASUVORK5CYII=","orcid":"","institution":"Jinshan Hospital of Fudan University","correspondingAuthor":true,"prefix":"","firstName":"Jingwei","middleName":"","lastName":"Qiang","suffix":""}],"badges":[],"createdAt":"2025-03-17 13:53:34","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6245251/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6245251/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s00261-025-05004-9","type":"published","date":"2025-05-29T15:57:01+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":79575125,"identity":"49ef8367-6083-4ef5-9df9-2371399402d5","added_by":"auto","created_at":"2025-03-31 11:13:13","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":3145798,"visible":true,"origin":"","legend":"\u003cp\u003eOverall workflow of habitat models. After image registration and segmentation, a K-means algorithm was used to generate habitats on mpMRI. Radiomics features were extracted from each habitat sub-region. After feature selection, habitat models were developed to predict prognoses in HGSOC patients. T2WI: T2-weighted imaging; CE-T1WI: contrast-enhanced T1-weighted imaging; DWI: diffusion-weighted imaging; ADC: apparent diffusion coefficient; MRI: magnetic resonance imaging; ROI: region of interest; ROC: receiver operator characteristic.\u003c/p\u003e","description":"","filename":"Fig1.png","url":"https://assets-eu.researchsquare.com/files/rs-6245251/v1/5b904c56c3630bd7eeda6e03.png"},{"id":79576435,"identity":"9519ce68-7870-48d1-b770-928e1a3344e9","added_by":"auto","created_at":"2025-03-31 11:21:13","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":2059357,"visible":true,"origin":"","legend":"\u003cp\u003eKaplan-Meier survival curves of Clinical models in three cohorts. According to the log-rank test, except for the external test group for predicting OS (\u003cem\u003eP\u003c/em\u003e = 0.065), the remaining \u003cem\u003eP\u003c/em\u003e-values were all less than 0.05. OS: overall survival.\u003c/p\u003e","description":"","filename":"Fig2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6245251/v1/1e906b8e8caa26417e032788.jpg"},{"id":79575132,"identity":"ca26b694-8533-419c-8cd1-cae6ca031526","added_by":"auto","created_at":"2025-03-31 11:13:13","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":5748696,"visible":true,"origin":"","legend":"\u003cp\u003eKaplan-Meier survival curves of Habitat models in three cohorts. According to the log-rank test, all \u003cem\u003eP\u003c/em\u003e-values were less than 0.05.\u003c/p\u003e","description":"","filename":"Fig3.png","url":"https://assets-eu.researchsquare.com/files/rs-6245251/v1/c88477c835515ea1eba8254d.png"},{"id":79576436,"identity":"1d7242c5-f931-4bd9-8d59-a7887a154f5e","added_by":"auto","created_at":"2025-03-31 11:21:13","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":2062177,"visible":true,"origin":"","legend":"\u003cp\u003eKaplan-Meier survival curves of Combined models in three cohorts. According to the log-rank test, all \u003cem\u003eP\u003c/em\u003e-values were less than 0.001.\u003c/p\u003e","description":"","filename":"Fig4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6245251/v1/4e6634a2d6b52d367a16c432.jpg"},{"id":79575133,"identity":"c55e901d-22b4-44b8-9c02-e286ab21cbad","added_by":"auto","created_at":"2025-03-31 11:13:13","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":2687297,"visible":true,"origin":"","legend":"\u003cp\u003eTime-dependent ROC curves of different models in three cohorts. For both OS and PFS predictions, the combined models demonstrated the highest AUCs. PFS: progression-free survival; AUC: area under the curve.\u003c/p\u003e","description":"","filename":"Fig5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6245251/v1/704d788815080ec804244aeb.jpg"},{"id":83782794,"identity":"e63be29c-462f-4523-9fd5-4ca37c361781","added_by":"auto","created_at":"2025-06-02 16:05:24","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":18957831,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6245251/v1/a14becce-5f9f-432f-861c-360065919f39.pdf"},{"id":79575129,"identity":"a2a05bc1-4ea2-4be0-af60-3bb657dd64ae","added_by":"auto","created_at":"2025-03-31 11:13:13","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":2245524,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementalmaterials.docx","url":"https://assets-eu.researchsquare.com/files/rs-6245251/v1/4053d2b0a16c16a53c4da111.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"mpMRI-based habitat analysis for predicting prognoses in patients with high-grade serous ovarian cancer: a multicenter study","fulltext":[{"header":"INTRODUCTION","content":"\u003cp\u003eOvarian cancer (OC) is acknowledged as one of the most lethal forms of gynecological cancer, ranking sixth in terms of mortality rates among women [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. High-grade serous ovarian cancer (HGSOC) is the predominant subtype, accounting for 70\u0026ndash;80% of all OC-related deaths [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Previous studies have demonstrated that clinicogenomic factors, like BRCA mutation status and the International Federation of Gynecology and Obstetrics (FIGO) stage, are valuable in predicting the prognosis of OC patients [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. However, these factors are neither consistently reliable as independent prognostic indicators nor adequately address the heterogeneity of clinical outcomes. Therefore, there is an urgent need to improve prognostic precision by integrating additional complementary indicators.\u003c/p\u003e \u003cp\u003eMRI is a commonly used non-invasive imaging method that offers several advantages over CT and ultrasound, including the absence of ionizing radiation, high soft tissue resolution, and multiparametric imaging capabilities, As a result, it is increasingly utilized in the assessment of OC [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. In particular, diffusion-weighted imaging (DWI), which leverages the apparent diffusion coefficient (ADC) metric, has shown significant predictive value for the prognosis of OC patients [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. With advancements in artificial intelligence, reseachers have recently developed MRI-based radiomics and deep learning models to predict the prognosis of OC patients with promising results [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Nevertheless, while these methodologies are capable of extracting general features of the lesion in a high-throughput manner, they fail to adequately capture the spatial heterogeneity that reflects the molecular biological characteristics of tumors. Habitat analysis is an unsupervised machine learning algorithm that can automatically identify sub-regions within different tissues by analyzing voxel similarity, thereby enabling the assessment of intratumoral heterogeneity [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Dextraze et al. [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e] found that spatial habitats based on multiparametric MRI (mpMRI) were associated with the prognosis of glioblastoma patients, and each prognostic related habitat had unique signaling pathway changes, suggesting that the association between image-derived phenotypic measurements and molecular characteristics. Since the tumor microenvironment was closely related to treatment resistance and survival prognosis, Bi et al. [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e] developed a mpMRI-based habitat radiomics model to predict platinum resistance in HGSOC patients, and the predictive performance of the habitat radiomics model was superior to that of conventional radiomics model and deep learning model due to the advantages of combining artificial intelligence and intratumoral heterogeneity, which preliminarily confirmed the value of habitat analysis based on mpMRI for the evaluation of HGSOC. Recent studies have demonstrated that habitat analysis based on PET/CT or CT shows potential for predicting the prognosis of HGSOC patients [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. However, the value of MRI-based habitat analysis in the prognosis of HGSOC warrants further investigation.\u003c/p\u003e \u003cp\u003eThe purpose of this study is to evaluate the effectiveness of mpMRI-based habitat analysis for predicting the prognosis of HGSOC patients and to develop a combined prognostic model incorporating clinical predictive indicators.\u003c/p\u003e"},{"header":"MATERIALS AND METHODS","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003ePatient cohorts\u003c/h2\u003e \u003cp\u003e This retrospective cohort study was conducted in accordance with the ethical guidelines in the Helsinki Declaration. The institutional review committees at our center approved this study (reference number: KHLL2023-KY208), and written informed consent was waived. All consecutive patients who were histologically confirmed HGSOC between May 2015 and August 2023 were retrospectively collected for the present analysis. Inclusion criteria comprised individuals meeting the following stipulations: (1) a confirmed diagnosis of HGSOC by operation and pathology; (2) the standard treatment included primary debulking surgery or neoadjuvant chemotherapy (NACT) with interval debulking surgery, followed by postoperative platinum-based chemotherapy; (3) availability of comprehensive pelvic MRI data within a two-week interval preceding the initiation of treatment; (4) no intervention aside from NACT was conducted between the MRI and the surgical procedure; and (5) at least six months of follow-up documentation following postoperative chemotherapy. The exclusion criteria were delineated as follows: (1) patients with a history of other malignant tumors; (2) poor-quality MRI imaging or inadequate registration; (3) the largest diameter of the lesion was under 1 cm; and (4) missing essential clinical data. As a result, this study included a cohort of 503 patients recruited from the specified centers. Patients from Center A were divided into the training cohort (220 cases) and internal validation cohort (94 cases) in a ratio of 7:3. Patients from Centers B (129 cases), C (25 cases), and D (35 cases) were combined into an external test cohort (189 cases).\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eClinical parameters\u003c/h3\u003e\n\u003cp\u003eWe collected the following clinical data from the patients: demographic information, treatment plan, laboratory examinations, surgical and histopathological findings, and follow-up records. Overall survival (OS) and progression-free survival (PFS) were defined as the duration from the initial treatment date to the occurrence of death or disease progression, respectively. According to the Response Evaluation Criteria in Solid Tumors (RECIST) guideline (version 1.1) [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e], disease progression is characterized by a rise of at least 20% from the baseline in the total diameters of target lesions, an absolute increment of 5 mm or more, or the appearance of new lesions. Platinum resistance is determined by assessing whether disease progression occurs during the course of platinum-based chemotherapy or within a six-month period following the completion of such treatment [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e].\u003c/p\u003e\n\u003ch3\u003eImage acquisition and segmentation\u003c/h3\u003e\n\u003cp\u003eMRI was conducted utilizing either 1.5 T or 3.0 T scanners (GE Signa Pioneer 3.0 T, GE Signa HDxt 1.5 T/3.0 T, Philips Ingenia 1.5 T/3.0 T, and Siemens Prisma/Vida/Skyra 3.0 T). The pelvic MRI images that we need to analyze comprised axial T2-weighted imaging (T2WI), diffusion-weighted imaging (DWI) with a b-value of 1000 s/mm\u003csup\u003e2\u003c/sup\u003e, apparent diffusion coefficient (ADC) maps, and contrast-enhanced T1-weighted imaging (CE-T1WI) at the venous phase. Details of the parameters are presented in Supplementary Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e. Certain parameters were modified to better fit the individual needs of each patient. Image preprocessing and segmentation are shown in Supplementary Note 1.\u003c/p\u003e\n\u003ch3\u003eHabitat imaging and feature selection\u003c/h3\u003e\n\u003cp\u003eThe habitat imaging was characterized using four distinct sequences, namely T2WI, DWI, ADC, and CE-T1WI. We recorded all intensity characteristics and procured a feature vector representing the characterization from the four sequences for each patient. Subsequently, the K-means module in the scikit-learn Python package (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://scikit-learn.org/stable/index.html\u003c/span\u003e\u003cspan address=\"https://scikit-learn.org/stable/index.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) was utilized for clustering habitat sub-regions. Voxels were clustered using the K-means algorithm based on cohort with squared Euclidean distances as the similarity metric and visualized as spatial habitats. The Calinski-Harabasz score was employed to assess the efficacy of clustering setups to determine the optimal number of clusters, spanning from two to ten. The details for K-means clustering and Calinski-Harabasz score are shown in Supplementary Note 2. Details of feature extraction and selection are provided in Supplementary Note 3.\u003c/p\u003e\n\u003ch3\u003eModel development\u003c/h3\u003e\n\u003cp\u003eThe habitat radiomics features were screened using univariate Cox regression and refined with Least Absolute Shrinkage and Selection Operator (LASSO)-Cox for dimensionality reduction, which eliminated irrelevant features by zeroing coefficients based on λ, optimized through 10-fold cross-validation to minimize mean standard error. The habitat models were developed through multivariate Cox regression analyses, utilizing the relevant features for OS and PFS, respectively. Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e illustrates the overall workflow of the habitat models. We conducted univariate and multivariate Cox regression analyses to identify clinical predictors and build clinical models. The combined models were developed via multivariate Cox regression analyses, combining clinical predictors and habitat signatures.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cp\u003eThe statistical analysis was conducted utilizing SPSS (version 26.0, IBM, New York, USA), Python (version 3.7.12), X-tile (version 3.6.1), and Onekey AI (version 4.8.4). Categorical variables were presented as counts and percentages, while continuous variables were presented as the median and interquartile range (IQR). Cox proportional hazards models with L2 regularization were used to build prognostic models for predicting OS and PFS. According to the optimal cut-off threshold determined by the X-tile software, the patients were divided into low- and high-risk groups. The performance of the different models was evaluated using Kaplan-Meier survival curves with a C-index. The log-rank test was employed to assess the statistical differences in survival between patients categorized as low-risk and high-risk. A \u003cem\u003eP\u003c/em\u003e-value of less than 0.05 was deemed to statistical significance. Finally, the time-dependent receiver operator characteristic (ROC) curve with the area under the curve (AUC) was used to evaluate the predictive performance of different models at a specific time points.\u003c/p\u003e \u003c/div\u003e"},{"header":"RESULTS","content":"\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003eClinical parameters\u003c/h2\u003e \u003cp\u003eThe clinical parameters of patients are summarized in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Although the start dates varied, all patients were followed up until May 2024. The median OS and PFS were 33.4 (IQR: 16.7\u0026ndash;49.3) months and 19.9 (IQR: 12.4\u0026ndash;18.9) months for the training patients, 31.8 (IQR: 19.7\u0026ndash;49.5) months and 18.5 (IQR: 10.5\u0026ndash;30.0) months for the internal validation patients, and 27.3 (IQR: 14.7\u0026ndash;38.8) months and 19.4 (IQR: 12.3\u0026ndash;30.9) months for the external test patients, respectively. Univariate and multivariate Cox regression analyses of clinical parameters in the training cohort are shown in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. Clinical predictors of OS included neoadjuvant chemotherapy (NACT) (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.003, HR\u0026thinsp;=\u0026thinsp;2.538 [1.372\u0026ndash;4.694]), poly ADP-ribose polymerase inhibitor (PARPi) (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.045, HR\u0026thinsp;=\u0026thinsp;0.439 [0.196\u0026ndash;0.983]), and platinum resistance (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001, HR\u0026thinsp;=\u0026thinsp;3.902 [2.210\u0026ndash;6.888]), while the predictors affecting PFS consisted of NACT (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.011, HR\u0026thinsp;=\u0026thinsp;1.619 [1.105\u0026ndash;2.452]), platinum resistance (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001, HR\u0026thinsp;=\u0026thinsp;7.706 [6.200-14.541]), and laterality of the lesion (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.028, HR\u0026thinsp;=\u0026thinsp;1.567 [1.069\u0026ndash;2.544]).\u003c/p\u003e \n\u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eThe clinical parameters of HGSOC patients in three cohorts.\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eParameter\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTraining\u0026nbsp;cohort\u003c/p\u003e\n \u003cp\u003e(N\u0026thinsp;=\u0026thinsp;220)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eInternal validation\u0026nbsp;cohort (N\u0026thinsp;=\u0026thinsp;94)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eExternal test\u0026nbsp;cohort\u003c/p\u003e\n \u003cp\u003e(N\u0026thinsp;=\u0026thinsp;189)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAge (years; median IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e54.0 (49.0\u0026ndash;62.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e53.5 (48.0\u0026ndash;68.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e55.0 (49.5\u0026ndash;63.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBMI (kg/m\u003csup\u003e2\u003c/sup\u003e; median IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e22.2 (20.2\u0026ndash;24.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e22.5 (20.4\u0026ndash;24.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e23.1 (21.2\u0026ndash;25.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCA125 (U/ml; median IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e570.4 (159.2\u0026ndash;1687.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e604.9 (165.1-1392.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e651.2 (212.0-2101.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHE4 (pmol/L; median IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e365.0 (162.0-736.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e365.0 (170.5-866.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e365.0 (190.5-513.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNACT\u003c/p\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e119 (54.1%)\u003c/p\u003e\n \u003cp\u003e101 (45.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e53 (56.4%)\u003c/p\u003e\n \u003cp\u003e41 (43.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e118 (62.4%)\u003c/p\u003e\n \u003cp\u003e71 (37.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePARPi\u003c/p\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e151 (68.6%)\u003c/p\u003e\n \u003cp\u003e69 (31.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e51 (54.3%)\u003c/p\u003e\n \u003cp\u003e43 (45.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e97 (51.3%)\u003c/p\u003e\n \u003cp\u003e92 (48.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePlatinum resistant\u003c/p\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e171 (77.7%)\u003c/p\u003e\n \u003cp\u003e49 (22.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e68 (72.3%)\u003c/p\u003e\n \u003cp\u003e26 (27.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e161 (85.2%)\u003c/p\u003e\n \u003cp\u003e28 (14.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLymph node dissection\u003c/p\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e74 (33.6%)\u003c/p\u003e\n \u003cp\u003e146 (66.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e132 (34.0%)\u003c/p\u003e\n \u003cp\u003e62 (66.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e113 (59.8%)\u003c/p\u003e\n \u003cp\u003e76 (40.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLaterality\u003c/p\u003e\n \u003cp\u003eUnilateral\u003c/p\u003e\n \u003cp\u003eBilateral\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e75 (34.1%)\u003c/p\u003e\n \u003cp\u003e145 (65.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e23 (24.5%)\u003c/p\u003e\n \u003cp\u003e71 (75.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e75 (39.7%)\u003c/p\u003e\n \u003cp\u003e114 (60.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eResidual tumor\u003c/p\u003e\n \u003cp\u003eR0\u003c/p\u003e\n \u003cp\u003eR1\u003c/p\u003e\n \u003cp\u003eR2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e211 (95.9%)\u003c/p\u003e\n \u003cp\u003e2 (0.9%)\u003c/p\u003e\n \u003cp\u003e7 (3.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e89 (94.7%)\u003c/p\u003e\n \u003cp\u003e2 (2.1%)\u003c/p\u003e\n \u003cp\u003e3 (3.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e172 (91.0%)\u003c/p\u003e\n \u003cp\u003e5 (2.6%)\u003c/p\u003e\n \u003cp\u003e12 (6.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFIGO\u0026nbsp;stage\u003c/p\u003e\n \u003cp\u003eI\u003c/p\u003e\n \u003cp\u003eII\u003c/p\u003e\n \u003cp\u003eIII\u003c/p\u003e\n \u003cp\u003eIV\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e10 (4.5%)\u003c/p\u003e\n \u003cp\u003e5 (2.3%)\u003c/p\u003e\n \u003cp\u003e119 (54.1%)\u003c/p\u003e\n \u003cp\u003e86 (39.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e3 (3.2%)\u003c/p\u003e\n \u003cp\u003e3 (3.2%)\u003c/p\u003e\n \u003cp\u003e59 (62.8%)\u003c/p\u003e\n \u003cp\u003e29 (30.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e8 (4.2%)\u003c/p\u003e\n \u003cp\u003e24 (12.7%)\u003c/p\u003e\n \u003cp\u003e121 (64.0%)\u003c/p\u003e\n \u003cp\u003e36 (19.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDiabetes\u003c/p\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e210 (95.5%)\u003c/p\u003e\n \u003cp\u003e10 (4.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e90 (95.7%)\u003c/p\u003e\n \u003cp\u003e4 (4.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e182 (96.3%)\u003c/p\u003e\n \u003cp\u003e7 (3.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eProgression\u003c/p\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e113 (51.4%)\u003c/p\u003e\n \u003cp\u003e107 (48.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e41 (43.6%)\u003c/p\u003e\n \u003cp\u003e53 (56.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e115 (60.8%)\u003c/p\u003e\n \u003cp\u003e74 (39.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDeath\u003c/p\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e170 (77.3%)\u003c/p\u003e\n \u003cp\u003e50 (22.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e68 (72.3%)\u003c/p\u003e\n \u003cp\u003e26 (27.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e168 (88.9%)\u003c/p\u003e\n \u003cp\u003e21 (11.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePFS (months; median IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e19.9 (12.4\u0026ndash;18.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e18.5 (10.5\u0026ndash;30.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e19.4 (12.3\u0026ndash;30.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOS (months; median IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e33.4 (16.7\u0026ndash;49.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e31.8 (19.7\u0026ndash;49.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e27.3 (14.7\u0026ndash;38.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003ctfoot\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"4\"\u003eHGSOC: high-grade serous ovarian cancer; BMI: body mass index; NACT: neoadjuvant chemotherapy; PARPi: poly ADP-ribose polymerase inhibitor; R0: no macroscopic tumor; R1: residual tumor\u0026thinsp;\u0026lt;\u0026thinsp;1 cm; R2: residual tumor\u0026thinsp;\u0026ge;\u0026thinsp;1 cm; FIGO: Federation of Gynecology and Obstetrics; OC: ovarian cancer; PFS: progression-free survival; OS: overall survival; IQR: inter quartile range.\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"4\"\u003e* \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tfoot\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eUnivariate and multivariate Cox regression analyses of clinical parameters in the training cohort.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"9\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003eParameter\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"4\" nameend=\"c5\" namest=\"c2\"\u003e \u003cp\u003eOS\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"4\" nameend=\"c9\" namest=\"c6\"\u003e \u003cp\u003ePFS\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003eUnivariate\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e \u003cp\u003eMultivariate\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003eUnivariate\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c9\" namest=\"c8\"\u003e \u003cp\u003eMultivariate\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHR (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eHR (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eHR (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eHR (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge (years)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1.000 (0.975\u0026ndash;1.026)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.995\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.990 (0.971\u0026ndash;1.009)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.285\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBMI (kg/m\u003csup\u003e2\u003c/sup\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.998 (0.934\u0026ndash;1.066)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.945\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.979 (0.933\u0026ndash;1.027)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.385\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCA125 (U/ml)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1.000 (1.000\u0026ndash;1.000)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.503\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.000 (1.000\u0026ndash;1.000)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.284\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHE4 (pmol/L)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1.000 (1.000\u0026ndash;1.000)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.842\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.000 (1.000\u0026ndash;1.000)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.577\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNACT\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2.135 (1.332\u0026ndash;3.425)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.002*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2.538 (1.372\u0026ndash;4.694)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.003*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2.177 (1.525\u0026ndash;3.109)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e1.619 (1.105\u0026ndash;2.452)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.011*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePARPi\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.556 (0.320\u0026ndash;0.967)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.038*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.439 (0.196\u0026ndash;0.983)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.045*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.753 (0.508\u0026ndash;1.117)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.158\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePlatinum resistant\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.404 (2.076\u0026ndash;5.579)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.902 (2.210\u0026ndash;6.888)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e8.673 (5.818\u0026ndash;12.928)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e7.706 (6.200-14.541)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLymph node dissection\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1.137 (0.696\u0026ndash;1.857)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.609\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.933 (0.652\u0026ndash;1.337)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.707\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDiabetes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1.017 (0.319\u0026ndash;3.240)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.978\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.031 (0.453\u0026ndash;2.346)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.943\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLaterality\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1.638 (0.983\u0026ndash;2.729)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.058\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.841 (1.247\u0026ndash;2.717)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.002*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e1.567 (1.069\u0026ndash;2.544)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.028*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFIGO\u0026nbsp;stage\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1.276 (0.909\u0026ndash;1.789)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.159\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.283 (0.998\u0026ndash;1.648)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.051\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eResidual tumor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1.147 (0.664\u0026ndash;1.980)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.623\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.068 (0.684\u0026ndash;1.669)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.772\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"9\"\u003eHR: Hazard risk; CI: confidence interval.\u003c/td\u003e\u003c/tr\u003e \u003ctr\u003e\u003ctd colspan=\"9\"\u003e* \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eHabitat imaging and feature selection\u003c/h2\u003e \u003cp\u003eThe Calinski-Harabasz score indicated that the ideal quantity of clusters was four. Therefore, all lesions were divided into no more than four habitat sub-regions (Supplementary Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). Supplementary Table S2 presents the mean signal intensity (SI) on different sequences, as well as the volume, and proportion of each habitat. Habitats 1 and 2 showed low ADC SIs and moderate or obvious enhancement. Habitat 3 had the highest SIs on both ADC maps and T2WI, while Habitat 4 presented the lowest SIs on ADC maps and CE-T1WI. In terms of sub-region proportions, Habitat 1 was the most predominant (34.7%), followed by habitat 3 (32.0%) .\u003c/p\u003e \u003cp\u003eWe extracted 1743 features from each sequence and improved the feature set by combining the features from four sequences, leading to a total of 27888 features, calculated as four sequences multiplied by four habitats and 1743 features. After feature reduction and LASSO-Cox selection (Supplementary Figure S2), 45 features and 62 features were retained to build models for predicting OS and PFS, respectively. For predicting OS, the top two features with the highest weight were derived from Habitat 2 on DWI and ADC maps (Supplementary Figure S3), and for predicting PFS, they were derived from Habitat 1 on T2WI and Habitat 2 on ADC maps (Supplementary Figure S4).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003ePerformance of different models\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e present the C-index and Kaplan-Meier survival curves of the different models, respectively. For predicting OS, the C-indexes of the clinical model, habitat model, and combined model in the training cohort were 0.776, 0.732, and 0.827, respectively (all \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001); in the internal validation cohort, they were 0.713, 0.707, and 0.752, respectively (all \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05); and in the external test cohort, they were 0.695 (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.065), 0.672 (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and 0.745 (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), respectively. For predicting PFS, the C-indexes of the clinical model, habitat model, and combined model in the training cohort were 0.799, 0.791, and 0.874, respectively (all \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001); in the internal validation cohort, they were 0.727, 0.627, and 0.784, respectively (all \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05); in the external test cohort, they were 0.700, 0.641, and 0.754, respectively (all \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05). In summary, for both OS and PFS predictions, the C-indexes of habitat models were lower than those of the clinical models, while the combined models demonstrated the highest C-indexes.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eC-index of the different models.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"8\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eCohort\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e \u003cp\u003eClinical model\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003eHabitat model\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c8\" namest=\"c7\"\u003e \u003cp\u003eCombined model\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eC-index\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eC-index\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eC-index\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003eOS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTraining\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.776\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.732\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.827\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eInternal validation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.713\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.707\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.015*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.752\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eExternal test\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.695\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.065\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.672\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.745\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003ePFS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTraining\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.799\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.791\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.874\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eInternal validation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.727\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.011*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.627\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.020*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.784\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eExternal test\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.700\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.641\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.002*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.754\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"8\"\u003e* \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e presents the performance and ROC curves of the different models for predicting the risk of death or progression at 3 years, respectively. For predicting OS, the AUCs of the clinical model, habitat model, and combined model in the training cohort were 0.782 (0.693\u0026ndash;0.871), 0.791 (0.706\u0026ndash;0.877), and 0.856 (0.782\u0026ndash;0.929), respectively; in the internal validation cohort, they were 0.759 (0.615\u0026ndash;0.904), 0.760 (0.626\u0026ndash;0.893), and 0.797 (0.666\u0026ndash;0.928), respectively; and in the external test cohort, they were 0.760 (0.618\u0026ndash;0.902), 0.747 (0.599\u0026ndash;0.896), and 0.821 (0.705\u0026ndash;0.937), respectively. For predicting PFS, the AUCs of the clinical model, habitat model, and combined model in the training cohort were 0.855 (0.798\u0026ndash;0.912), 0.893 (0.846\u0026ndash;0.941), and 0.948 (0.917\u0026ndash;0.978), respectively; in the internal validation cohort, they were 0.778 (0.671\u0026ndash;0.885), 0.660 (0.509\u0026ndash;0.812), and 0.851 (0.759\u0026ndash;0.944), respectively; and in the external test cohort, they were 0.668 (0.561\u0026ndash;0.775), 0.788 (0.683\u0026ndash;0.893), and 0.791 (0.696\u0026ndash;0.885), respectively. Similarly, for both OS and PFS predictions, the combined models demonstrated the highest AUCs.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eThe performance of the different models in predicting 3-year risk of death or progression.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"9\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCohort\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAUC (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eAccuracy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eSensitivity\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eSpecificity\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003ePPV\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003eNPV\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"8\" rowspan=\"9\"\u003e \u003cp\u003eOS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003eTraining\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eClinical\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.782 (0.693\u0026ndash;0.871)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.672\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.621\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.824\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.914\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.418\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHabitat\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.791 (0.706\u0026ndash;0.877)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.730\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.718\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.765\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.902\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.473\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCombined\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.856 (0.782\u0026ndash;0.929)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.781\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.777\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.794\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.920\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.540\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003eInternal validation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eClinical\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.759 (0.615\u0026ndash;0.904)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.759\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.850\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.556\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.810\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.625\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHabitat\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.760 (0.626\u0026ndash;0.893)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.707\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.675\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.778\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.871\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.519\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCombined\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.797 (0.666\u0026ndash;0.928)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.810\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.825\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.778\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.892\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.667\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003eExternal test\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eClinical\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.760 (0.618\u0026ndash;0.902)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.757\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.778\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.636\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.925\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.333\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHabitat\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.747 (0.599\u0026ndash;0.896)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.527\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.444\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.239\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCombined\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.821 (0.705\u0026ndash;0.937)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.770\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.762\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.818\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.960\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.375\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"8\" rowspan=\"9\"\u003e \u003cp\u003ePFS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003eTraining\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eClinical\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.855 (0.798\u0026ndash;0.912)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.781\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.700\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.844\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.778\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.784\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHabitat\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.893 (0.846\u0026ndash;0.941)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.838\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.871\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.811\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.782\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.890\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCombined\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.948 (0.917\u0026ndash;0.978)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.863\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.900\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.833\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.808\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.915\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003eInternal validation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eClinical\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.778 (0.671\u0026ndash;0.885)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.677\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.737\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.652\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.467\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.857\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHabitat\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.660 (0.509\u0026ndash;0.812)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.692\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.421\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.804\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.471\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.771\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCombined\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.851 (0.759\u0026ndash;0.944)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.708\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.947\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.609\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.500\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.966\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003eExternal test\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eClinical\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.668 (0.561\u0026ndash;0.775)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.602\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.750\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.530\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.436\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.814\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHabitat\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.788 (0.683\u0026ndash;0.893)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.806\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.594\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.909\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.760\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.822\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCombined\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.791 (0.696\u0026ndash;0.885)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.786\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.500\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.924\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.762\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.792\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"9\"\u003eAUC: area under the curve; PPV: positive predictive value; NPV: negative predictive value.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"DISCUSSION","content":"\u003cp\u003eIn this study, we divided tumors into four habitats based on the K-means algorithm, constructed habitat models using the radiomics features of each habitat sub-region, and developed a combined model incorporating clinical predictors. We found that the predictive performance of the habitat models based on mpMRI did not surpass that of the clinical models. However, the combined models exhibited consistent and superior prognostic performance across all cohorts. This consistency highlights the reliability and clinical applicability of the combined models, which may assist healthcare professionals in developing personalized treatment strategies for HGSOC patients.\u003c/p\u003e \u003cp\u003eIn this study, NACT and platinum resistance were found to be predictors of OS and PFS. Due to tumor burden or genetic mutations induced by chemotherapy, some HGSOC lesions continued to grow after NACT, leading to a poor prognosis [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. Elyashiv et al. [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e] reported that platinum resistance was a significant prognostic indicator in females diagnosed with epithelial OC. Once platinum resistance occurred, both patients with primary platinum resistance and those with secondary platinum resistance had lower survival rates. In addition, we observed that maintenance therapy with PARPi was associated with OS, and the laterality of lesions influenced PFS. Yamada et al. [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e] reported that the PFS of OC patients was longer in the unilateral group than in the bilateral group, confirming that tumor laterality could be an independent prognostic factor in OC patients. According to the American Society of Clinical Oncology (ASCO) guidelines [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e], the application of PARPi after complete and partial responses to initial therapy or platinum-sensitive relapse therapy notably prolonged PFS in advanced OC patients. However, PARPi maintenance therapy did not significantly prolong OS, likely due to the presence of adverse effects and other factors [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. In addition, the duration of maintenance treatment and its impact on prognosis require normalization and validation in real-world settings with large, multicenter sample sizes [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Beyond the factors mentioned above, residual tumor, FIGO stage, and CA125 levels, among others, are generally considered prognostic factors [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. However, none of these indicators were identified as independent prognostic factors in our study. One possible explanation is that differences in patient distribution may have affected the results of the Cox regression analyses, such as the relatively small number of FIGO stage I-II and non-R0 patients. Additionally, individuals classified under the same FIGO stage often exhibit differing survival outcomes, and the clinical validity of indicators such as CA125 remains controversial [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. The present study demonstrated that low-cost clinical models might serve as promising prognostic tools for forecasting survival outcomes in HGSOC patients. However, consistent with the findings of Huang et al. [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e], the C-index of the clinical models was not sufficiently high in the external test cohort. This underscores the important of integrating imaging data and other methods to comprehensively evaluate the prognosis of OC patients.\u003c/p\u003e \u003cp\u003eRadiomics, a crucial link connecting medical imaging with personalized medicine, enhances the diagnostic, prognostic, and predictive precision of cancer assessments through the extensive extraction and analysis of quantitative imaging features obtained from medical imaging techniques [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. Due to the limited interpretability of radiomics, although many predictive models have been proposed, their connection to biologically relevant factors has rarely been elucidated [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. Habitat radiomics combines the advantages of radiomics and spatial heterogeneity to successfully evaluate Ki-67 expression and PFS in HGSOC patients, and demonstrates that the habitat model can better stratify the prognosis in radiomics models [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. According to the log-rank test, our habitat models also showed significant differences in stratification between the high- and low-risk groups across all cohorts. However, the C-indexes of our habitat models for predicting OS and PFS in the external test cohort were 0.672 and 0.641, respectively, which were similar to those in the previous study (C-index\u0026thinsp;=\u0026thinsp;0.61) [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e], indicating that image-based artificial intelligence analyses alone remain insufficient to accurately predict the prognosis of HGSOC patients. In addition, like conventional radiomics, the process of habitat radiomics also involves complicated steps such as feature extraction and reduction, so the studies of radiomics are still in the scientific research stage. At present, it is not possible to directly obtain the prognostic outcomes of patients simply by inputting images in clinical practice. However, with the development of artificial intelligence, this study may provide the basis for the development of prognostic software in the future.\u003c/p\u003e \u003cp\u003eIn addition to performing habitat imaging, the combination of radiological data with other data, such as clinical parameters, is also the alternative solutions to overcome the limitation of poor interpretability [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. A series of high-quality studies have shown that the integration of multimodal data is beneficial for complementing tumor heterogeneity across multiple scales, providing complementary predictive information, and enhancing the predictive capability of unimodal models [\u003cspan additionalcitationids=\"CR30\" citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. Boehm et al. [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e] developed a machine learning integrated model by integrating clinical and CT features, which could effectively improve prognostic risk stratification in HGSOC patients. Bi et al. [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e] concluded that a comprehensive nomogram combining habitat signature and clinical features had the best performance in identifying platinum-resistant patients with HGSOC, superior to both the clinical model and the habitat model. This study also found that the combined model had the best prognostic performance in HGSOC patients, regardless of whether it was C-indexes or time-dependent AUCs. In future clinical practice, when clinical parameters alone were not enough to accurately predict the prognosis of HGSOC patients, MRI-based habitat imaging should be added. Moreover, we hope to standardize and simplify the process of developing the combined models in the future. The specific steps include [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]: Initially, a standardized pelvic MRI scanning protocol will be implemented within a two-week window prior to the surgical intervention. Subsequently, by amalgamating data from the hospital information system alongside the Picture Archiving and Communication System, the integrated software will autonomously compute intratumoral heterogeneity metrics, while also delivering insights regarding risk stratification, potential prognostic outcomes, and individualized treatment options.\u003c/p\u003e \u003cp\u003eAnother recent study reported that the multimodal model based on CT for predicting prognosis of epithelial OC demonstrated a C-index of 0.64 in the external test set [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. The C-indexes of our combined models based on mpMRI were around 0.75. Furthermore, Huang et al. [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e] developed intratumoral and peritumoral radiomics models based on CT to predict the prognosis of HGSOC patients, with AUCs ranging from 0.552 to 0.777 in the internal test set. In present study, the AUCs of the combined models were above 0.79 in the internal validation and external test cohorts. These results indirectly may suggest that the performance of the combined model combining MRI habitat and clinical parameters in predicting the prognosis of HGSOC patients is superior and more stable than that of CT. Of course, preoperative CT and mpMRI data from more centers with larger samples should be included together for future analyses to verify this conclusion.\u003c/p\u003e \u003cp\u003eInterestingly, we found that the top two features with the highest weight to the habitat models both contained features of Habitat 2 on the ADC maps. Recent studies emphasized the crucial role of the enhancing tumor or solid tumor, and the ADC values of these lesions might more accurately represent the heterogeneity of OC [\u003cspan additionalcitationids=\"CR34\" citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. Habitats 1 and 2 showed low ADC SIs and moderate or obvious enhancement, indicating that the main component of these habitat was solid. Our present study suggested that the solid components of HGSOC lesions and the features derived from ADC maps have a greater impact on prognosis. Zhang et al. [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e] suggested that the ADC values of functional tumor volume (the volume of solid mass) could be used to assess preoperative prognostic factors in epithelial OC, which further supported our results and speculation. However, the habitat sub-regions generated based on unsupervised clustering do not completely represent the actual pathological tissues and need to be prospectively verified by point-to-point analyses on MRI and pathological images.\u003c/p\u003e \u003cp\u003eThis study has some limitations. First, the retrospective design inherently carries the risk of selection bias, because only patients with available MRI data and those with valid follow-up data were included, and more patients who only underwent CT before surgery and patients lost to follow-up were not included. Additionally, we only incorporated four MRI sequences due to the limitations of retrospective design, and other functional MRI, such as diffusion kurtosis imaging, need to be explored further. Moreover, the lack of standardization in MRI equipment and parameters across different hospitals may lead to variability in model performance between different cohorts. Second, the follow-up duration may not be sufficient to comprehensively assess long-term outcomes, highlighting the need for future studies with larger sample sizes and extended follow-up periods to validate the prognostic significance of our models in diverse patient populations. Third, manual delineation remains the gold standard of image segmentation, but due to the large size and extensive range of HGSOC lesions, the workload for manual segmentation is substantial. As a result, the habitat radiomics features were not excluded through intraclass correlation coefficient testing. We are currently working on developing an automated and accurate tumor segmentation method to enhance the efficiency and stability of the model.\u003c/p\u003e \u003cp\u003eIn conclusion, this study highlights the critical role of combined models that integrate clinical features and habitat radiomics features in predicting OS and PFS in HGSOC patients. The findings may aid in clinical decision-making and foster personalized treatment strategies. Future research should focus on the prospective validation and standardization of MRI acquisition protocols and pathological validation of habitat biomarkers, as well as increasing follow-up time to further improve predictive accuracy and ultimately enhance the prognosis for HGSOC patients.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eDrafting of the manuscript: Q.B.Concept and design: Q.B. and CH.A.Acquisition, analysis, or interpretation of data: Q.B., K.M., Y.Liu., J.Y., A.Z., WW.S., Y.Lei., YZ.W. and Y.S.Critical revision of the manuscript for important intellectual content: HM.L. and JW.Q.Statistical analysis: Q.B. and K.M.Supervision: JWQ. All authors have accessed and verified the underlying data, and takes responsibility for the integrity of the data and the accuracy of the data analysis.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eThis work was supported by the National Natural Science Foundations of China [grant numbers 82460340, 82471943, 82471932, 82271940, 82160524]; Kunming University of Science and Technology \u0026amp; the First People's Hospital of Yunnan Province Joint Special Project on Medical Research [grant number KUST-KH2022027Y], the Basic Research on Application of Joint Special Funding of Science and Technology Department of Yunnan Province-Kunming Medical University [grant number 202301AY070001-084], Shanghai Jinshan District Health Committee [grant number JSZK2023A02], and Natural Science Foundation of Shanghai [grant number 22ZR1412500]\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eDue to the privacy of patients, the data related to patients cannot be available for public access but can be obtained from the corresponding author on reasonable request approved by the institutional review board of the four centers.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eSiegel RL, Giaquinto AN, Jemal A (2024) Cancer statistics, 2024. Ca-Cancer J Clin 74:12\u0026ndash;49. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3322/caac.21820\u003c/span\u003e\u003cspan address=\"10.3322/caac.21820\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKurman RJ, Shih I (2016) The Dualistic Model of Ovarian Carcinogenesis: Revisited, Revised, and Expanded. Am J Pathol 186:733\u0026ndash;747. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ajpath.2015.11.011\u003c/span\u003e\u003cspan address=\"10.1016/j.ajpath.2015.11.011\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBowtell DD, Bohm S, Ahmed AA, et al. (2015) Rethinking ovarian cancer II: reducing mortality from high-grade serous ovarian cancer. Nat Rev Cancer 15:668\u0026ndash;679. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/nrc4019\u003c/span\u003e\u003cspan address=\"10.1038/nrc4019\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGallagher DJ, Konner JA, Bell-McGuinn KM, et al. (2011) Survival in epithelial ovarian cancer: a multivariate analysis incorporating BRCA mutation status and platinum sensitivity. Ann Oncol 22:1127\u0026ndash;1132. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/annonc/mdq577\u003c/span\u003e\u003cspan address=\"10.1093/annonc/mdq577\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIrodi A, Rye T, Herbert K, et al. (2020) Patterns of clinicopathological features and outcome in epithelial ovarian cancer patients: 35 years of prospectively collected data. BJOG-Int J Obstet Gy 127:1409\u0026ndash;1420. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/1471-0528.16264\u003c/span\u003e\u003cspan address=\"10.1111/1471-0528.16264\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDaoud T, Sardana S, Stanietzky N, Klekers AR, Bhosale P, Morani AC (2022) Recent Imaging Updates and Advances in Gynecologic Malignancies. Cancers 14:5528. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/cancers14225528\u003c/span\u003e\u003cspan address=\"10.3390/cancers14225528\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang C, Ma L, Zhao Y, et al. (2024) Estimating pathological prognostic factors in epithelial ovarian cancers using apparent diffusion coefficients of functional tumor volume. Eur J Radiol 176:111514. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ejrad.2024.111514\u003c/span\u003e\u003cspan address=\"10.1016/j.ejrad.2024.111514\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi H, Cai S, Deng L, et al. (2023) Prediction of platinum resistance for advanced high-grade serous ovarian carcinoma using MRI-based radiomics nomogram. Eur Radiol 33:5298\u0026ndash;5308. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00330-023-09552-w\u003c/span\u003e\u003cspan address=\"10.1007/s00330-023-09552-w\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu L, Wan H, Liu L, Wang J, Tang Y, Cui S, Li Y (2023) Deep Learning Provides a New Magnetic Resonance Imaging-Based Prognostic Biomarker for Recurrence Prediction in High-Grade Serous Ovarian Cancer. Diagnostics 13:748. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/diagnostics13040748\u003c/span\u003e\u003cspan address=\"10.3390/diagnostics13040748\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eO'Connor JP, Rose CJ, Waterton JC, Carano RA, Parker GJ, Jackson A (2015) Imaging intratumor heterogeneity: role in therapy response, resistance, and clinical outcome. Clin Cancer Res 21:249\u0026ndash;257. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1158/1078-0432.CCR-14-0990\u003c/span\u003e\u003cspan address=\"10.1158/1078-0432.CCR-14-0990\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDextraze K, Saha A, Kim D, et al. (2017) Spatial habitats from multiparametric MR imaging are associated with signaling pathway activities and survival in glioblastoma. Oncotarget 8:112992\u0026ndash;113001. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.18632/oncotarget.22947\u003c/span\u003e\u003cspan address=\"10.18632/oncotarget.22947\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBi Q, Miao K, Xu N, et al. (2024) Habitat Radiomics Based on MRI for Predicting Platinum Resistance in Patients with High-Grade Serous Ovarian Carcinoma: A Multicenter Study. Acad Radiol 31:2367\u0026ndash;2380. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.acra.2023.11.038\u003c/span\u003e\u003cspan address=\"10.1016/j.acra.2023.11.038\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuang X, Huang Y, Liu K, Zhang F, Zhu Z, Xu K, Li P (2024) Intratumoral and Peritumoral Radiomics for Predicting the Prognosis of High-grade Serous Ovarian Cancer Patients Receiving Platinum-Based Chemotherapy. Acad Radiol. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.acra.2024.09.001\u003c/span\u003e\u003cspan address=\"10.1016/j.acra.2024.09.001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang X, Xu C, Grzegorzek M, Sun H (2022) Habitat radiomics analysis of pet/ct imaging in high-grade serous ovarian cancer: Application to Ki-67 status and progression-free survival. Front Physiol 13:948767. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fphys.2022.948767\u003c/span\u003e\u003cspan address=\"10.3389/fphys.2022.948767\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEisenhauer EA, Therasse P, Bogaerts J, et al. (2009) New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer 45:228\u0026ndash;247. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ejca.2008.10.026\u003c/span\u003e\u003cspan address=\"10.1016/j.ejca.2008.10.026\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang T, Tang J, Yang H, et al. (2022) Effect of Apatinib Plus Pegylated Liposomal Doxorubicin vs Pegylated Liposomal Doxorubicin Alone on Platinum-Resistant Recurrent Ovarian Cancer: The APPROVE Randomized Clinical Trial. JAMA Oncol 8:1169\u0026ndash;1176. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1001/jamaoncol.2022.2253\u003c/span\u003e\u003cspan address=\"10.1001/jamaoncol.2022.2253\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHimoto Y, Cybulska P, Shitano F, et al. (2019) Does the method of primary treatment affect the pattern of first recurrence in high-grade serous ovarian cancer? Gynecol Oncol 155:192\u0026ndash;200. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ygyno.2019.08.011\u003c/span\u003e\u003cspan address=\"10.1016/j.ygyno.2019.08.011\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eElyashiv O, Aleohin N, Migdan Z, Leytes S, Peled O, Tal O, Levy T (2024) The Poor Prognosis of Acquired Secondary Platinum Resistance in Ovarian Cancer Patients. Cancers 16:641. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/cancers16030641\u003c/span\u003e\u003cspan address=\"10.3390/cancers16030641\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYamada Y, Mabuchi S, Kawahara N, Kawaguchi R (2021) Prognostic significance of tumor laterality in advanced ovarian cancer. Obstet Gynecol Sci 64:524\u0026ndash;531. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5468/ogs.21176\u003c/span\u003e\u003cspan address=\"10.5468/ogs.21176\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTew WP, Lacchetti C, Ellis A, et al. (2020) PARP Inhibitors in the Management of Ovarian Cancer: ASCO Guideline. J Clin Oncol 38:3468\u0026ndash;3493. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1200/JCO.20.01924\u003c/span\u003e\u003cspan address=\"10.1200/JCO.20.01924\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim JH, Kim SI, Park EY, et al. (2023) Impact of postoperative residual disease on survival in epithelial ovarian cancer with consideration of recent frontline treatment advances: A systematic review and meta-analysis. Gynecol Oncol 179:24\u0026ndash;32. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ygyno.2023.10.018\u003c/span\u003e\u003cspan address=\"10.1016/j.ygyno.2023.10.018\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTuninetti V, Marin-Jimenez JA, Valabrega G, Ghisoni E (2024) Long-term outcomes of PARP inhibitors in ovarian cancer: survival, adverse events, and post-progression insights. ESMO Open 9:103984. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.esmoop.2024.103984\u003c/span\u003e\u003cspan address=\"10.1016/j.esmoop.2024.103984\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSalminen L, Nadeem N, Jain S, et al. (2020) A longitudinal analysis of CA125 glycoforms in the monitoring and follow up of high grade serous ovarian cancer. Gynecol Oncol 156:689\u0026ndash;694. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ygyno.2019.12.025\u003c/span\u003e\u003cspan address=\"10.1016/j.ygyno.2019.12.025\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang M, Cheng S, Jin Y, Zhao Y, Wang Y (2021) Roles of CA125 in diagnosis, prediction, and oncogenesis of ovarian cancer. Bba-Rev Cancer 1875:188503. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.bbcan.2021.188503\u003c/span\u003e\u003cspan address=\"10.1016/j.bbcan.2021.188503\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuang W, Bao Y, Luo X, Yao L, Yuan L (2022) Novel prognostic nomograms to assess survival in high-grade serous ovarian carcinoma after surgery and chemotherapy: a retrospective cohort study from SEER database. Ann Transl Med 10:728. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.21037/atm-21-4383\u003c/span\u003e\u003cspan address=\"10.21037/atm-21-4383\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLambin P, Leijenaar R, Deist TM, et al. (2017) Radiomics: the bridge between medical imaging and personalized medicine. Nat Rev Clin Oncol 14:749\u0026ndash;762. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/nrclinonc.2017.141\u003c/span\u003e\u003cspan address=\"10.1038/nrclinonc.2017.141\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRusso L, Bottazzi S, Sala E (2024) Artificial intelligence in female pelvic oncology: tailoring applications to clinical needs. Eur Radiol 34:4038\u0026ndash;4040. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00330-023-10455-z\u003c/span\u003e\u003cspan address=\"10.1007/s00330-023-10455-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuang X, Huang Y, Liu K, Zhang F, Zhu Z, Xu K, Li P (2024) Predicting prognosis for epithelial ovarian cancer patients receiving bevacizumab treatment with CT-based deep learning. NPJ Precis Oncol 8:202. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41698-024-00688-6\u003c/span\u003e\u003cspan address=\"10.1038/s41698-024-00688-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen Z, Chen Y, Sun Y, et al. (2024) Predicting gastric cancer response to anti-HER2 therapy or anti-HER2 combined immunotherapy based on multi-modal data. Signal Transduct Tar 9:222. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41392-024-01932-y\u003c/span\u003e\u003cspan address=\"10.1038/s41392-024-01932-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoehm KM, Aherne EA, Ellenson L, et al. (2022) Multimodal data integration using machine learning improves risk stratification of high-grade serous ovarian cancer. Nat Cancer 3:723\u0026ndash;733. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s43018-022-00388-9\u003c/span\u003e\u003cspan address=\"10.1038/s43018-022-00388-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSammut SJ, Crispin-Ortuzar M, Chin SF, et al. (2022) Multi-omic machine learning predictor of breast cancer therapy response. Nature 601:623\u0026ndash;629. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41586-021-04278-5\u003c/span\u003e\u003cspan address=\"10.1038/s41586-021-04278-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHe D, Zhang X, Chang Z, Liu Z, Li B (2024) Survival time prediction in patients with high-grade serous ovarian cancer based on (18)F-FDG PET/CT- derived inter-tumor heterogeneity metrics. BMC Cancer 24:337. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12885-024-12087-y\u003c/span\u003e\u003cspan address=\"10.1186/s12885-024-12087-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBorde T, Nezami N, Laage GF, et al. (2022) Optimization of the BCLC Staging System for Locoregional Therapy for Hepatocellular Carcinoma by Using Quantitative Tumor Burden Imaging Biomarkers at MRI. Radiology 304:228\u0026ndash;237. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1148/radiol.212426\u003c/span\u003e\u003cspan address=\"10.1148/radiol.212426\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi HM, Zhang R, Gu WY, et al. (2019) Whole solid tumour volume histogram analysis of the apparent diffusion coefficient for differentiating high-grade from low-grade serous ovarian carcinoma: correlation with Ki-67 proliferation status. Clin Radiol 74:918\u0026ndash;925. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.crad.2019.07.019\u003c/span\u003e\u003cspan address=\"10.1016/j.crad.2019.07.019\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMimura R, Kato F, Tha KK, et al. (2016) Comparison between borderline ovarian tumors and carcinomas using semi-automated histogram analysis of diffusion-weighted imaging: focusing on solid components. Jpn J Radiol 34:229\u0026ndash;237. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11604-016-0518-6\u003c/span\u003e\u003cspan address=\"10.1007/s11604-016-0518-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"abdominal-radiology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"aima","sideBox":"Learn more about [Abdominal Radiology](http://link.springer.com/journal/261)","snPcode":"261","submissionUrl":"https://submission.springernature.com/new-submission/261/3","title":"Abdominal Radiology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"ovarian cancer, prognosis, MRI, habitat, radiomics","lastPublishedDoi":"10.21203/rs.3.rs-6245251/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6245251/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003ePurpose\u003c/h2\u003e \u003cp\u003eTo evaluate the value of multiparametric MRI (mpMRI)-based habitat analysis for predicting prognoses in patients with high-grade serous ovarian cancer (HGSOC), and to develop combined models by integrating habitat analysis with clinical predictors.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eThis retrospective study included 503 HGSOC patients from four centers. A K-means algorithm was used to identify voxel clusters and generate habitats on mpMRI. Radiomics features were extracted from each habitat sub-region. After feature selection, habitat models were developed to predict overall survival (OS) and progression-free survival (PFS). Cox regression analyses were performed to identify clinical predictors and construct clinical models. Combined models were developed by integrating habitat signatures with clinical predictors. Model performance was evaluated using C-index and time-dependent receiver operating characteristic area under the curves (AUCs).\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eCompared with the clinical models (OS: 0.713 and 0.695; PFS: 0.727 and 0.700) and habitat models (OS: 0.707 and 0.672; PFS: 0.627 and 0.641), the combined models integrating habitat features and clinical independent predictors such as neoadjuvant chemotherapy (OS: 0.752 and 0.745; PFS: 0.784 and 0.754) achieved the highest C-indexes for predicting OS and PFS in the internal validation cohort and external test cohort. The combined models also achieved the highest AUCs in all cohorts.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eThe habitat models based on mpMRI demonstrated potential value in predicting the prognoses of HGSOC patients, but no significant advantages over the clinical models. The combined models were expected to improve the prognoses from the level of individual clinical characteristics and habitat features reflecting intratumoral heterogeneity.\u003c/p\u003e","manuscriptTitle":"mpMRI-based habitat analysis for predicting prognoses in patients with high-grade serous ovarian cancer: a multicenter study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-03-31 11:13:08","doi":"10.21203/rs.3.rs-6245251/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-05-01T13:04:18+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-05-01T09:01:46+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"187186905017940038827589232152295938109","date":"2025-04-25T01:46:32+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-04-19T14:13:25+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"198235675716546454789832107801671121600","date":"2025-03-20T18:15:37+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-03-19T23:52:15+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-03-18T10:35:12+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-03-18T10:30:57+00:00","index":"","fulltext":""},{"type":"submitted","content":"Abdominal Radiology","date":"2025-03-17T13:42:38+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"abdominal-radiology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"aima","sideBox":"Learn more about [Abdominal Radiology](http://link.springer.com/journal/261)","snPcode":"261","submissionUrl":"https://submission.springernature.com/new-submission/261/3","title":"Abdominal Radiology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"99b16dc4-bc0c-479c-8c43-c6f267df08c9","owner":[],"postedDate":"March 31st, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-06-02T15:59:02+00:00","versionOfRecord":{"articleIdentity":"rs-6245251","link":"https://doi.org/10.1007/s00261-025-05004-9","journal":{"identity":"abdominal-radiology","isVorOnly":false,"title":"Abdominal Radiology"},"publishedOn":"2025-05-29 15:57:01","publishedOnDateReadable":"May 29th, 2025"},"versionCreatedAt":"2025-03-31 11:13:08","video":"","vorDoi":"10.1007/s00261-025-05004-9","vorDoiUrl":"https://doi.org/10.1007/s00261-025-05004-9","workflowStages":[]},"version":"v1","identity":"rs-6245251","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6245251","identity":"rs-6245251","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}