Multi-parametric MRI combined with radiomics for the diagnosis and grading of endometrial fibrosis

Multi-parametric MRI combined with radiomics for the diagnosis and grading of endometrial fibrosis | 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 Multi-parametric MRI combined with radiomics for the diagnosis and grading of endometrial fibrosis Huanhuan Wang, Li Zhu, Hui Zhu, Jie Meng, Huanhuan Liang, Danyan Li, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4864304/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Accurate evaluation of endometrial fibrosis can help clinicians schedule individual treatment. This study aims to explore the application value of multi-parametric MRI (MP-MRI) combined with radiomics in the diagnosis and grading of endometrial fibrosis, aiming to construct models that can effectively distinguish endometrial fibrosis and compare the diagnostic performance of radiomics models established by different machine learning algorithms. Methods A total of 74 patients with severe endometrial fibrosis(SEF), 41 patients with mild to moderate fibrosis (MMEF)confirmed by hysteroscopy, and 40 healthy women of reproductive age were prospectively enrolled. All participants underwent T2 and DWI sequence scans during the periovulatory period. By freely delineating the volume of interest (VOI) of the endometrium in three subgroups, radiomic features were extracted and selected. Two feature selection methods and four machine learning classifiers were combined in pairs to establish five prediction models [model 1 (T2 + ADC + clinical data), model 2 (T2 + ADC), model 3 (T2), model 4 (ADC), and model 5 (clinical data)], resulting in a total of 40 classification models. The predictive performance of all models was evaluated using the area under the curve (AUC), F1 score, and accuracy (ACC). Results Among the 40 classification models, the "UFS-LR" model, which combined unsupervised feature selection (UFS) with the logistic regression (LR) classifier, performed the best, with an average AUC of 0.92 on the test set. Among the five models constructed via UFS-LR, model 1 exhibited the best performance, with average AUC, F1 score, and ACC values of 0.92, 0.79, and 0.81, respectively. The T2-related models had higher average AUC values than model 4 and model 5 did, especially in the MMEF and SEF groups. Among the optimal features selected from different models, T2-related features accounted for the largest number and had the highest weight. Conclusions Machine learning-based MP-MRI radiomics analysis exhibited excellent performance in grading endometrial fibrosis and has great potential for providing robust support for clinical diagnosis and treatment. Endometrium Fibrosis Magnetic Resonance Imaging Radiomics Machine Learning Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Endometrial fibrosis (EF) is the primary pathological feature of intrauterine adhesion (IUA), also known as Asherman syndrome[ 1 ]. This syndrome arises from damage to the endometrial basal layer, leading to the formation of fibrosis and adhesion within the uterus, ultimately resulting in menstrual abnormalities, infertility, or recurrent miscarriages[ 2 , 3 ]. IUA can be caused by various factors, among which induced abortion and postpartum curettage are important causes[ 4 ]. In recent years, with the frequent occurrence of intrauterine operations and the increase in reproductive system infections, the prevalence of IUA has shown an increasing trend, and IUA has become the second leading cause of secondary infertility in Chinese women[ 5 ]. This study adopted the scoring standard of the American Fertility Society (AFS), which categorizes patients into mild, moderate, and severe groups[ 6 ]. The treatment options differ for patients with varying severity. Patients with mild to moderate adhesions can undergo hysteroscopic adhesiolysis, whereas severe cases exhibit significant fibrosis of the endometrial basalis, posing greater challenges for surgical separation of adhesions and consequently leading to a notably lower pregnancy rate than mild to moderate adhesions do[ 7 , 8 ]. Therefore, accurately assessing the degree of endometrial fibrosis is crucial for assisting clinicians in selecting the optimal treatment plan and evaluating treatment outcomes. Currently, there are multiple methods for diagnosing intrauterine adhesions. Trans-vaginal ultrasound (TVS) is primarily used for initial clinical screening[ 9 ]. Prior to the invention of hysteroscopy, hysterosalpingography (HSG) was the preferred method for observing the uterine cavity, but it has a relatively high false-positive rate and lower diagnostic accuracy[ 8 , 10 ]. Hysteroscopy is considered the "gold standard" technique for the diagnosis and treatment of IUA[ 11 ]. Although it has high diagnostic accuracy, it is invasive and can increase the risk of re-adhesion. Furthermore, it can only visualize pathological changes on the surface of the uterine cavity and cannot clearly demonstrate lesions in the myometrium or basal layer. MRI offers a non-invasive, radiation-free, and multi-sequence, multi-directional approach to clearly visualize uterine contours and adhesion bands. With the development of imaging technology, MRI and its new technologies have emerged for the diagnosis and treatment of endometrial fibrosis[ 12 – 14 ]. T2WI and DWI are routinely acquired scanning sequences. On T2WI, the uterine cavity with adhesions appears as alternating high signals from normal endometrial tissue and low signals from areas of endometrial fibrosis[ 15 ]. Diffusion weighted imaging (DWI) is the most widely used method to non-invasively quantify water molecule diffusion by calculating the apparent diffusion coefficient (ADC) value. Studies have shown that when tissue fibrosis occurs, the apparent diffusion coefficient (ADC) decreases[ 16 – 18 ]. Traditional imaging assessment is subjective. Compared with traditional imaging, radiomics can further mine and analyze the information contained in imaging images in a noninvasive manner by extracting thousands of quantitative features in a high-throughput fashion[ 19 , 20 ]. Studies have shown that machine learning-based radiomics can be applied to tumor grading[ 21 , 22 ]. The extracellular matrix (ECM) accumulates in the damaged endometrial layer, causing excessive proliferation of collagen fibers. Consequently, endometrial regeneration is inhibited, and the endometrium is gradually replaced by connective tissue. The pathological basis for radiomic differences in varying degrees of endometrial fibrosis may be due to differences in the degree of fibrosis and water content. On this basis, we hypothesize that radiomics can be used to distinguish between different degrees of endometrial fibrosis. In recent years, increasing evidence has suggested that morphological and MRI signal alterations occur in the endometrium during fibrosis[ 12 , 13 ]. However, our study goes beyond these observations by utilizing MP-MRI, specifically T2WI and apparent diffusion coefficient (ADC) mapping, in conjunction with radiomics to develop a novel model capable of differentiating the degree of endometrial fibrosis, thereby enhancing diagnostic accuracy and providing more robust guidance for clinical diagnosis and treatment. Materials and methods Patients This study adhered to the Helsinki Declaration and was approved by the the Institutional Ethics Committee of Nanjing Drum Tower Hospital with the approval number: 2019-051-02. Written informed consent was obtained from all patients and healthy female participants. The AFS scoring criteria were used (see Additional file 1)[ 6 ]. A total score of 1–4 is considered mild, a score of 5–8 is considered moderate, and a score of 9–12 is considered severe. In this study, patients with severe IUA were defined as having severe endometrial fibrosis (SEF), whereas patients with mild to moderate IUA were classified as having moderate endometrial fibrosis (MMEF). Infertile patients and healthy volunteers were prospectively enrolled from June 2018 to December 2023. All patients were diagnosed with endometrial fibrosis through hysteroscopy and pathology. Hysteroscopy was performed within 7 days after MRI. The hysteroscopy and endometrial biopsy of the patient were performed by our experienced gynecologists in the late proliferative phase. The gynecologists assessed the degree of intrauterine adhesions according to the AFS scoring criteria and performed endometrial biopsy at the obvious fibrotic scar site via biopsy forceps. For infertile patients with a normal endometrium (such as those with tubal obstruction), endometrial biopsy was performed from the anterior wall, posterior wall, or fundus of the uterus as a histological control for a normal endometrium. The inclusion criteria for patients were as follows: 1) aged 20–40 years; 2) history of dilation and curettage (D&C) surgery; 3) clinical diagnosis of infertility (regular sexual activity without contraception for at least one year without conception); 4) endometrial fibrosis confirmed by hysteroscopy and pathological diagnosis; and 5) normal ovarian function. The inclusion criteria for healthy females were as follows: 1) aged 20–40 years; 2) normal menstrual cycle and menstrual flow; and 3) no history of uterine injury or disease, such as D&C and intrauterine infections. The exclusion criteria were as follows: 1) history of other severe uterine diseases, such as adenomyosis, endometrial tuberculosis, endometritis, endometrial hyperplasia or polyps, as well as severe uterine malformations confirmed by the ESHRE/ESGE consensus[ 23 ]; 2) MRI contraindications, such as cardiac pacemakers and cochlear implants; and 3) severe MRI artifacts that result in unclear visualization and affect data analysis. The flowchart for case inclusion and exclusion is shown in Fig. 1 . MRI examination method All participants underwent scanning using a 16-channel phased-array body coil on a 3.0T MRI scanner (Ingenia, Philips Medical Systems, Best, The Netherlands) in the supine position with the head first. All patients and healthy females were instructed to empty their bladders before the MRI scan. Owing to the morphological and microstructural changes that occur in the endometrium during the menstrual cycle, MRI examinations were performed during ovulation (confirmed by ultrasonography of the dominant follicle) to improve the consistency of the results. The MRI sequences included sagittal T2WI and DWI of the uterus, with a total scan time of approximately 6 minutes. The T2WI scan parameters were as follows: repetition time (TR)/echo time (TE) = 1700–5000 ms/100 ms; matrix size = 200×166; field of view (FOV) = 120 mm×120 mm; slice thickness = 3 mm; slice gap = 0.3 mm; voxel size = 0.60×0.72×3 mm 3 ; number of signal averages (NSA) = 1.1; scan duration = 3 minutes and 56 seconds. The DWI scan parameters were as follows: b-values of 0 and 1000 s/mm 2 , TR/TE = 6000 ms/90 ms; matrix size = 133×160; FOV = 240 mm×240 mm; slice thickness = 3 mm; slice gap = 0.3 mm; NSA = 2. The scan duration was 2 minutes and 6 seconds. Image Segmentation and Feature Extraction All the images were independently analyzed by two radiologists (WHH and LHH with 12 and 4 years of experience in pelvic MRI interpretation, respectively), who were blinded to the patients’ clinical information. The DWI images were loaded onto a workstation (Extended MR Workspace 2.6.3.4; Philips Medical Systems, Best, the Netherlands), and the ADC map was automatically generated via a monoexponential model[ 24 ]. The T2WI, DWI, and ADC maps were imported into 3D Slicer software ( www.slicer.org/ ) for segmentation of the volume of interest (VOI). T2WI and DWI were used as references to manually outline the contour of the endometrium layer by layer, covering as much of the endometrium as possible on each slice to accurately reflect the heterogeneity of the endometrium. This process automatically generates a 3D image (as shown in Fig. 2 ). Inter-observer Consistency Test Thirty patients were randomly selected according to their date from each of the three groups, and two physicians independently completed the VOI segmentation and feature extraction. The inter-class correlation coefficient (ICC) was adopted for consistency testing. Features with an ICC value greater than 0.75 were considered to have good reproducibility, whereas those with an ICC value of 0.75 or less were excluded. Feature Selection, Prediction Model Construction and Evaluation First, highly stable features were selected on the basis of the intra-group and inter-group consistency (ICC > 0.75) of feature extraction. To eliminate redundant features and enhance model efficiency, univariate feature selection (UFS) and recursive feature elimination (RFE) were employed for feature selection. Using logistic regression (LB), support vector machine (SVM), naïve bayes (NB), and Gaussian process classification (GPC) classifiers, 40 predictive models were established, including model 1 (T2 + ADC + clinical data), model 2 (T2 + ADC), model 3 (T2), model 4 (ADC), and model 5 (clinical data). The enrolled data were randomly stratified and divided into a train set (108 cases: 28 healthy women, 29 with MMEF, and 51 with SEF) and a test set (47 cases: 12 healthy women, 12 MMEF and 23 SEF) at a ratio of 7:3. The performance of each model in both the train and test sets was analyzed using receiver operating characteristic (ROC) curves. The average area under the curve (AUC), F1-score (the harmonic mean of recall and precision), and ACC were used to evaluate the performance of the different models. Statistical analysis The statistical analysis of this study was conducted using Python ( https://www.python.org/ ). Quantitative data that conformed to a normal distribution are expressed as the mean ± standard deviation. One-way ANOVA was used to compare differences in continuous variables among the three groups, and the chi-square test was used to compare differences in categorical variables among the three groups. ROC curves were plotted via Python software. The diagnostic performance of the models was evaluated using ROC curve analysis, and the AUC value, F1-score, and ACC were calculated. A P value < 0.05 was considered to indicate statistical significance. Results General Information A total of 74 patients with SFE, 41 patients with MMEF confirmed by hysteroscopy, and 40 healthy women were ultimately included. There were significant differences in age, body weight, endometrial thickness (EMT), and the length of the uterine cavity (LUC) among healthy women versus MMEF and healthy women versus SEF patients ( P < 0.05). Furthermore, significant differences in EMT and LUC were detected between patients with MMEF and those with SEF. The general information is presented below (Table 1 ). Table 1 General information of the enrolled cases Group Label Cases Age (y) Weight (kg) EMT (mm) LUC (mm) Healthy women 0 40 28.4 ± 3.3 52.0 ± 4.9 11.4 ± 1.9 39.9 ± 2.6 MMEF 1 41 32.7 ± 4.5 58.5 ± 8.7 7.1 ± 2.1 35.3 ± 4.5 SEF 2 74 32.9 ± 3.8 58.9 ± 7.7 6.1 ± 1.0 36.9 ± 3.8 EMT: endometrial thickness; LUC: length of uterine cavity; MMEF: mild to moderate endometrial fibrosis; SEF: severe endometrial fibrosis. Inter-observer Consistency Test and Feature Selection Based on the consistency test analysis results of the radiomic features extracted by the two physicians, features with ICC > 0.75 were retained. Ultimately, the numbers of selected features for T2 and ADC were 1055 and 1344, respectively. After UFS feature selection, the final numbers of features used to construct model 1 , model 2 , model 3 , model 4 , and model 5 were 16, 16, 16, 11, and 1, respectively. Following RFE feature selection, the final number of features used for model construction were 6, 6, 6, 6, and 4, respectively. Comparison of Different Radiomic Models The results indicated varying diagnostic efficiencies among the different predictive models. Among the 40 classification models, the “UFS-LR” model, which combines UFS with the LR classifier, performed the best, with an average AUC of 0.92 on the test set. Among the five models constructed via UFS-LR, model1, the ensemble model, exhibited the best performance, with average AUC, F1 score, and accuracy values of 0.92, 0.79, and 0.81, respectively (details in Table 2 ). The corresponding heatmap and confusion matrix are shown in Fig. 3 . The ROC curves of the test sets for the different models constructed using UFS-LR are shown in Fig. 4 . The average AUC values of the T2-related models (model 1 , model 2 , model 3 ) were higher than those of model 4 and model 5 , particularly in the MMEF and SEF groups. Model 1 enhances the diagnostic performance for differentiating between MMEF and SEF patients (as shown in Fig. 5 ). Table 2 Diagnostic performance of different models constructed using UFS-LR Model AUC F1-score ACC 0 1 2 Macro Micro 0 1 2 Macro Weighted Model 1 0.95 0.88 0.92 0.92 0.92 0.80 0.70 0.86 0.79 0.80 0.81 Model 2 0.96 0.88 0.92 0.92 0.92 0.83 0.60 0.84 0.75 0.77 0.79 Model 3 0.97 0.88 0.90 0.92 0.92 0.85 0.67 0.84 0.78 0.80 0.80 Model 4 0.90 0.63 0.85 0.79 0.83 0.67 0.13 0.81 0.54 0.60 0.66 Model 5 0.95 0.52 0.77 0.75 0.84 0.83 0.21 0.78 0.61 0.65 0.88 AUC: Area under the receiver operating characteristic curve; Macro: Macro-average; Micro: Micro-average; Weighted: Weighted-average; ACC: accuracy; 0: Healthy women; 1: Mild to moderate fibrosis (MMEF) ;2: Severe endometrial fibrosis (SEF). AUC: area under the receiver operating characteristic curve; Macro: macro-average; Micro: micro-average; Weighted: weighted-average; ACC: accuracy; 0: healthy women; 1: mild to moderate fibrosis (MMEF) ;2: severe endometrial fibrosis (SEF). Optimal features and weights selected by different models The optimal features and corresponding weights selected by machine learning methods vary. The optimal features of the combined models are detailed in see Additional file 2. The 16 features and their scores for model 1 constructed with UFS-LR are shown in Fig. 6 . Discussion To date, MRI studies on endometrial fibrosis have focused on morphological and signal changes[ 9 , 12 , 13 ]. This study aimed to develop a model that is capable of effectively diagnosing and distinguishing the degree of endometrial fibrosis by utilizing MP-MRI (T2 and ADC) in combination with radiomics and clinical parameters. Our results showed that the use of multi-parametric radiomic model enables accurate identification of healthy volunteers, MMEF and SEF patients, it can also improve diagnostic accuracy for MMEF and SEF patients facilitating the development of personalized treatment plans and enhancing fertility outcomes. Impact of different machine learning methods on predictive models Previous radiomics studies have often relied solely on a single feature selection and classifier method, without a consensus on the classifier that performs best. On the basis of different machine learning methods, we established multiple radiomic models to predict the degree of endometrial fibrosis. We employed the AUC, F1 score, and ACC to evaluate the performance of the different models. The results demonstrated varying diagnostic efficiencies among different prediction models. Notably, the UFS-LR model outperformed the other classification models. UFS selects the best features on the basis of univariate statistical tests, making it suitable for identifying features most relevant to the target variable. Applying an appropriate classifier can potentially enhance classifier performance. The LR method is widely used in classification tasks because of its low computational requirements and the ability to intuitively observe the predicted probabilities of samples. The results of all the machine learning methods indicate that the diagnostic performance of model 1 is greater than that of model 5 . The results indicated that the optimal classifier varied with different feature selections, which is consistent with the findings of Wang et al[ 25 ]. However, all the models exhibited excellent diagnostic efficacy, displaying consistent trends in their diagnostic performance across the different models. Therefore, when modeling with diverse radiomics features, attempting to utilize different selection methods and classifiers can enhance model performance and result reproducibility. Analysis of the value of radiomics in predicting endometrial fibrosis In clinical practice, the treatment approach varies according to the severity of fibrosis, and the “gold standard” for assessing EF is hysteroscopy[ 11 ]. Under hysteroscopy, the normal endometrial morphology is characterized by densely distributed glandular openings on the surface, a bright red color, a soft texture, and good elasticity. The uterine cavity appears regular in shape, forming an inverted triangle, with clearly visible bilateral tubal ostia. During hysteroscopy in patients, a pale endometrium and dense fibrous scar tissue covering the entire uterine cavity can be observed. The uterine cavity has lost its original softness, becoming rigid and inelastic. Transcervical resection of adhesion (TCRA) is the primary treatment method aimed at restoring the normal volume and shape of the endometrial cavity[ 7 , 26 ]. Postoperative estrogen therapy can stimulate endometrial regeneration and re-epithelialization of scar surfaces. Mild to moderate adhesions usually involve a small area or only the endometrial layer, making surgical separation and release of adhesions relatively easy. For patients with severe IUA, marked fibrosis of the endometrial basalis layer leads to localized or widespread decreased elasticity of the uterine wall, increasing the difficulty of surgical separation of adhesions and the risk of complications such as uterine perforation during surgery. Additionally, the postoperative recurrence rate is high, posing significant challenges in improving postoperative adhesion outcomes[ 27 ]. Recent reports have suggested that transplanting stem cells into the uterine cavity or injecting growth factors into the endometrium of patients with severe fibrosis can reduce fibrosis areas, increase glandular counts, promote angiogenesis, and enhance pregnancy rates, demonstrating the potential for treating severe endometrial fibrosis[ 28 ]. Additionally, studies have shown that ECM scaffolds (SIS, Small intestinal submucosa) combined with intrauterine balloon therapy can repair endometrial fibrosis and improve IUA, providing a novel treatment approach for improving pregnancy outcomes in patients with moderate-to-severe IUA-related infertility[ 29 ]. Therefore, accurate early assessment of the degree of endometrial fibrosis and treatment response is crucial for assisting clinicians in selecting optimal treatment regimens and evaluating outcomes. Model 1 demonstrated demonstrates the highest diagnostic performance (AUC = 0.92 in the test set), surpassing the diagnostic accuracy of model 5, which relies solely on clinical parameters (AUC = 0.75 in the test set). Notably, the diagnostic performance of the radiomics model in predicting MMEF and SEF is lower than that for healthy volunteers, which aligns with clinical observations. Purely clinical indicators exhibit limited effectiveness in distinguishing between MMEF and SEF. The combined machine learning radiomics model enhances the diagnostic accuracy in differentiating between MMEF and SEF, with model 1 and model 5, constructed via UFE-LR, achieving AUC values of 0.88 versus 0.52 and 0.92 versus 0.77, respectively. This approach can reduce subjectivity in clinical assessments of fibrosis severity, avoid excessive invasive hysteroscopy, and improve patient outcomes. The potential of radiomics in predicting endometrial fibrosis The morphological and functional information provided by MP-MRI is insufficient for a comprehensive and effective assessment of tissue biological characteristics, necessitating the exploration of new technologies and methods. Radiomics can mine quantitative image features from medical images, enabling the quantification of heterogeneity and microstructure information beyond the capability of the human visual system[ 19 ]. Its process includes image segmentation, feature extraction, feature selection, and model building and validation. Among these, image segmentation is a critical step in texture analysis technology, as it serves as the direct source for feature extraction[ 30 ]. The use of whole-lesion VOI delineation to acquire features involves more spatial information than features extracted on the basis of a single region of interest (ROI) and can characterize the heterogeneity of the entire tumor. This study adopted whole-lesion VOI delineation, avoiding the limitations of selecting partial lesions or extracting ROIs from a single slice. By extracting radiomic features from multiple sequences of MP-MRI, this method maximizes diversity and increases the number of features, thereby enhancing the completeness of lesion description, improving the accuracy of lesion heterogeneity characterization, and reducing sampling errors[ 31 ]. The features from different MP-MRI sequences reflect distinct biological information, and the comprehensive utilization of these image features allows for a more comprehensive quantitative assessment of lesion heterogeneity. T2WI serves as a fundamental sequence in MRI composition, capable of revealing a wealth of histopathological features such as water content, degree of fibrosis, necrosis, hemorrhage, etc.[ 12 ]. The ADC value quantitatively assesses the diffusion and restriction of water molecules within the target tissue, elucidating cell density and facilitating the differentiation of degrees of fibrosis. Studies have shown that the ADC decreases when tissue fibrosis occurs[ 13 ]. Consequently, this study employed T2WI and ADC sequences, with the potential for combining T2WI, DWI and clinical parameters to become a novel indicator for differentiating the degree of endometrial fibrosis. The T2-related models had higher average AUC values than model4 and model5 did, especially in the MMEF and SEF groups. A total of 16 features were ultimately selected through UFS-LR modeling model1, including 1 clinical feature (EMT), 5 ADC features, and 10 T2 features, with T2 features accounting for a larger proportion. Notably, the feature with the highest weight was the T2 wavelet feature (T2_wavelet-LLL_gldm_DependenceVariance). Different radiomics methods result in varying optimal features and weights, with T2-related features being the most numerous and having the greatest weight. The endometrium is lined by a luminal epithelium and contains tubular glands that radiate from the surface to the endometrial–myometrial interface[ 32 ]. The water content of the endometrium is an important component of its normal physiological function. T2WI can microscopically reflect this water content and the degree of fibrosis. In patients with fibrosis, the endometrial structure becomes disrupted and replaced by fibrous tissue, leading to decreased water content[ 33 , 34 ]. Therefore, T2 features may hold greater value in predicting endometrial fibrosis. The aforementioned studies underscore the significant role of radiomics features in enhancing the accuracy of grading endometrial fibrosis. Currently, there are multiple criteria for assessing the severity of intrauterine adhesions, and future endeavors may explore incorporating MRI features, particularly T2-related features, into scoring systems. This approach has the potential to contribute to the refinement of endometrial fibrosis scoring. Limitations First, the sample size was relatively small, and as this was a single-center study, there may be some selection bias. Second, image segmentation is based on manual delineation, which is currently the most commonly used research method but inevitably involves some subjectivity. In future studies, deep learning techniques, such as automatic or semi-automatic segmentation methods, can be employed to replace the current time-consuming and labor-intensive manual segmentation methods. Third, there are few clinical parameters of patients, and more clinical data can be incorporated into the analysis, for example, correlation studies combining clinical features and pathological characteristics. Finally, the inherent shortcomings of radiomics analysis techniques lead to poor reproducibility, requiring larger sample sizes and multi-center validation to achieve clinical translation. Nonetheless, this is sufficient to demonstrate the feasibility of using MP-MRI combined with radiomics for the quantitative assessment of endometrial fibrosis. Conclusion In summary, the radiomics model we established is feasible for differentiating healthy women, patients with MMEF and patients with SEF. This prospective study demonstrated the ability of T2 combined with ADC radiomics as a non-invasive imaging biomarker for the quantitative diagnosis and assessment of endometrial fibrosis. Compared with clinical parameters, morphological and signal changes, combined radiomics can provide additional microstructural information and heterogeneity, which not only aids clinicians in objectively diagnosing and assessing endometrial fibrosis but also facilitates the arrangement of timely and effective treatments, thereby offering certain assistance in clinical decision-making.In summary, the radiomics model we have established is feasible for differentiating healthy women, MMEF and SEF patients. This prospective study demonstrates the capability of T2 combined with ADC radiomics as a non-invasive imaging biomarker for quantitative diagnosis and assessment of endometrial fibrosis. Compared to clinical parameters, morphological and signal changes, combined radiomics can provide additional microstructural information and heterogeneity, which not only aids clinicians in objectively diagnosing and assessing endometrial fibrosis, but also facilitates the arrangement of timely and effective treatments, thereby offering certain assistance in clinical decision-making. Abbreviations MRI Magnetic resonance imaging MP-MRI Multi-parametric MRI SEF Severe endometrial fibrosis MMEF Mild to moderate fibrosis VOI Volume of interest DWI Diffusion weighted imaging ADC Apparent diffusion coefficient UFS Unsupervised feature selection LR Logistic regression RFE Recursive feature elimination SVM Support vector machine NB Naïve bayes GPC Gaussian process classification Declarations Ethics approval and consent to participate All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.This study was approved by the Institutional Ethics Committee of Nanjing Drum Tower Hospital, Nanjing, China (Ethic approval No.2019-051-02). Written informed consent was obtained from all patients and healthy female participants. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Funding This study was supported by the National Natural Science Foundation of China (81671751 and 81871410) and the Strategic Priority Research Program of the Chinese Academy of Sciences (XDA16040302). Author Contribution ZZY, HYL,LDY: Conceptualization, Methodology, Project administration Supervision, Writing - Review & Editing. WHH: Investigation, Data curation, Formal analysis, Visualization, Writing - Original Draft. ZL,ZH,LHH: Investigation, Visualization, Data Curation, Writing - Review & Editing.ZH, MJ,HJL: Resources, Investigation, Writing - Review & Editing.All authors reviewed the manuscript. Acknowledgments Not applicable. Data Availability The data that support the findings of this study are not openly available due to reasons of sensitivity and are available from the corresponding author upon reasonable request. Data are located in controlled access data storage at Nanjing Drum Tower Hospital. References Zhu Q, Yao S, Ye Z, et al. Ferroptosis contributes to endometrial fibrosis in intrauterine Adhesions. Free Radical Bio Med. 2023;205:151–62. Wallach EE, Schenker JG, Margalioth EJ. Intrauterine adhesions: An updated Appraisal. Fertil Steril. 1982;37(5):593–610. Asherman JG. Traumatic intra-uterine adhesions. BJOG. 1950;57(6):892–6. March CM. Management of Asherman’s syndrome. Reprod Biomed Online. 2011;23(1):63–76. Chinese Medical Association Obstetrics and Gynecology Branch. Chinese Expert Consensus on Clinical Diagnosis and Treatment of Intrauterine adhesions [J]. Chin J Obstet Gynecol. 2015;50(12):881–7. The American Fertility. Society classifications of adnexal adhesions, distal tubal occlusion, tubal occlusion secondary to tubal ligation, tubal pregnancies, Müllerian anomalies and intrauterine Adhesions. Fertil Steril. 1988;49(6):944–55. AAGL Advancing Minimally Invasive Gynecology Worldwide. AAGL practice report: Practice guidelines for management of intrauterine Synechiae. J Minim Invasive Gynecol. 2010;17(1):1–7. Yu D, Wong YM, Cheong Y, Xia E, Li TC. Asherman Syndrome–one century Later. Fertil Steril. 2008;89(4):759–79. Soares SR, dos Barbosa MM, Camargos AF. Diagnostic accuracy of sonohysterography, transvaginal sonography, and hysterosalpingography in patients with uterine cavity diseases. Fertil Steril. 2000;73(2):406–11. Acholonu UC, Silberzweig J, Stein DE, Keltz M. Hysterosalpingography versus sonohysterography for intrauterine abnormalities. JSLS. 2011;15(4):471–4. Yang JH, Chen MJ, Chen CD, Chen SU, Ho HN, Yang YS. Optimal waiting period for subsequent fertility treatment after various hysteroscopic surgeries. Fertil Steril. 2013;99(7):2092–6. Zhou N, Zhu H, Jiang P, et al. Quantification of Endometrial Fibrosis Using Noninvasive MRI T2 Mapping: Initial Findings. J Magn Reson Imaging. 2023;58(6):1703–13. Feng Y, Jiang P, Hu Q, et al. Initial Experience With Diffusion-Weighted Magnetic Resonance Imaging for the Evaluation of Endometrial Fibrosis. J Comput Assist Tomogr. 2021;45(3):383–8. Kang K, Wang A, Wu H. MRI for Diagnosing Intrauterine Adhesions. J Obstet Gynaecol Can. 2023;45(10):102168. Bacelar AC, Wilcock D, Powell M, Worthington BS. The value of MRI in the Assessment of Traumatic Intra-uterine adhesions (Asherman’s Syndrome). Clin Radiol. 1995;50(2):80–3. Jiang H, Chen J, Gao R, Huang Z, Wu M, Song B. Liver fibrosis staging with diffusion-weighted imaging: a systematic review and meta-analysis. Abdom Radiol (NY). 2017;42(2):490–501. Caruso A, Angriman I, Scarpa M, et al. Diffusion-weighted magnetic resonance for assessing fibrosis in Crohn’s disease. Abdom Radiol (NY). 2020;45(8):2327–35. Balci NC, Perman WH, Saglam S, Akisik F, Fattahi R, Bilgin M. Diffusion-weighted magnetic resonance imaging of the pancreas. Top Magn Reson Imaging. 2009;20(1):43–7. Kocak B, Baessler B, Bakas S, et al. CheckList for EvaluAtion of Radiomics research (CLEAR): A step-by-step reporting guideline for authors and reviewers endorsed by ESR and EuSoMII. Insights Imaging. 2023;14(1):75. Lambin P, Leijenaar RTH, Deist TM, et al. Radiomics: the bridge between medical imaging and personalized medicine. Nat Rev Clin Oncol. 2017;14(12):749–62. Hu J, Zhao Y, Li M, et al. Machine learning-based radiomics analysis in predicting the meningioma grade using multiparametric MRI. Eur J Radiol. 2020;10(131):109251. Kumar A, Jha AK, Agarwal JP, et al. Machine-Learning-Based Radiomics for Classifying Glioma Grade from Magnetic Resonance Images of the Brain. J Pers Med. 2023;13(6):920. Grimbizis GF, Gordts S, Di Spiezio Sardo A, et al. The ESHRE/ESGE consensus on the classification of female genital tract congenital Anomalies. Hum Reprod. 2013;28(8):2032–44. Kwee TC, Takahara T, Ochiai R, et al. Whole-body diffusion-weighted magnetic resonance Imaging. Eur J Radiol. 2009;70(3):409–17. Wang J, Chen J, Zhou R, Gao Y, Li J. Machine Learning-based multiparametric MRI radiomics for predicting poor responders after neoadjuvant chemoradiotherapy in rectal Cancer Patients. BMC Cancer. 2022;22(1):420. Salazar CA, Isaacson K, Morris S. A comprehensive review of Asherman’s syndrome: Causes, symptoms and treatment Options. Curr Opin Obstet Gynecol. 2017;29(4):249–56. Healy MW, Schexnayder B, Connell MT, et al. Intrauterine adhesion prevention after hysteroscopy: A systematic review and meta-Analysis. Am J Obstet Gynecol. 2016;215(3):267–75. Gharibeh N, Aghebati-Maleki L, Madani J, Pourakbari R, Yousefi M, Ahmadian Heris J. Cell-based therapy in thin endometrium and Asherman Syndrome. Stem Cell Res Ther. 2022;13(1):33. Pang WJ, Zhang Q, Ding HX, Sun NX, Li W. Effect of new biological patch in repairing intrauterine adhesion and improving clinical pregnancy outcome in infertile women: Study protocol for a randomized controlled Trial. Trials. 2022;23(1):510. Zwanenburg A, Vallières M, Abdalah MA, et al. The Image Biomarker Standardization Initiative: Standardized Quantitative Radiomics for High-Throughput Image-based Phenotyping. Radiology. 2020;295(2):328–38. Rosenkrantz AB, Triolo MJ, Melamed J, Rusinek H, Taneja SS, Deng FM. Whole-lesion apparent diffusion coefficient metrics as a marker of percentage Gleason 4 component within Gleason 7 prostate cancer at radical Prostatectomy. J Magn Reson Imaging. 2015;41(3):708–14. Spencer TE, Hayashi K, Hu J, Carpenter KD. Comparative developmental biology of the mammalian uterus. Curr Top Dev Biol. 2005;68:85–122. Cameron IL, Ord VA, Fullerton GD. Characterization of proton NMR relaxation times in normal and pathological tissues by correlation with other tissue parameters. Magn Reson Imaging. 1984;2(2):97–106. De Kock I, Bos S, Delrue L, et al. MRI texture analysis of T2-weighted images is preferred over magnetization transfer imaging for readily longitudinal quantification of gut fibrosis. Eur Radiol. 2023;33(9):5943–52. Additional Declarations No competing interests reported. Supplementary Files Additionalfile1.doc Additionalfile2.doc Cite Share Download PDF Status: Posted Version 1 posted 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-4864304","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":351242009,"identity":"a3584ce4-c03c-4b37-aa01-eafff9230b56","order_by":0,"name":"Huanhuan Wang","email":"","orcid":"","institution":"Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School","correspondingAuthor":false,"prefix":"","firstName":"Huanhuan","middleName":"","lastName":"Wang","suffix":""},{"id":351242010,"identity":"7017eb4d-14c9-416a-8d39-30c61d3bc90e","order_by":1,"name":"Li Zhu","email":"","orcid":"","institution":"Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School","correspondingAuthor":false,"prefix":"","firstName":"Li","middleName":"","lastName":"Zhu","suffix":""},{"id":351242011,"identity":"b6d488cb-2fdf-4a5a-b01c-4b2fbe3e627c","order_by":2,"name":"Hui Zhu","email":"","orcid":"","institution":"Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School","correspondingAuthor":false,"prefix":"","firstName":"Hui","middleName":"","lastName":"Zhu","suffix":""},{"id":351242012,"identity":"ea87febd-be60-46c6-bb78-eef515d2940f","order_by":3,"name":"Jie Meng","email":"","orcid":"","institution":"Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School","correspondingAuthor":false,"prefix":"","firstName":"Jie","middleName":"","lastName":"Meng","suffix":""},{"id":351242013,"identity":"f96af016-db87-483c-b466-706731bc2dd2","order_by":4,"name":"Huanhuan Liang","email":"","orcid":"","institution":"Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School","correspondingAuthor":false,"prefix":"","firstName":"Huanhuan","middleName":"","lastName":"Liang","suffix":""},{"id":351242014,"identity":"1a2de20b-322d-43a8-b6ae-ece237738ac7","order_by":5,"name":"Danyan Li","email":"","orcid":"","institution":"Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School","correspondingAuthor":false,"prefix":"","firstName":"Danyan","middleName":"","lastName":"Li","suffix":""},{"id":351242015,"identity":"0dfaa97e-1694-44d9-bb35-e90fd84d2d93","order_by":6,"name":"Yali Hu","email":"","orcid":"","institution":"Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School","correspondingAuthor":false,"prefix":"","firstName":"Yali","middleName":"","lastName":"Hu","suffix":""},{"id":351242016,"identity":"64c3cace-02e7-4d3b-ba09-9f0c960faa79","order_by":7,"name":"Zhengyang Zhou","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAxklEQVRIiWNgGAWjYHACNiC2YWBsAFI8JGhJI13LYQiTKC3yM9KfPfjYdj6PeUYC44O3bQzy5oS0MM5ISDec2Xa7GMhgNpzbxmC4s4GAFmaJhGPSvG23ExtnJLABGQwJBgcIaGGTSGwDqjwH0sL+mygtPBLJIMMPgG1hJkqLBM8zNskZ55ITG3seNkvOOSdhuIGQFvn29GcSH8rsEje2Jx/88KbMRp6gLQwCCRDasAEcmRKE1AMBP9RQeSLUjoJRMApGwQgFAB3pPO02CdqXAAAAAElFTkSuQmCC","orcid":"","institution":"Nanjing Drum Tower Hospital, University of Chinese Medicine","correspondingAuthor":true,"prefix":"","firstName":"Zhengyang","middleName":"","lastName":"Zhou","suffix":""}],"badges":[],"createdAt":"2024-08-05 21:15:39","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4864304/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4864304/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":66370496,"identity":"11c24435-7b5e-457d-b0f1-2916df6bbaca","added_by":"auto","created_at":"2024-10-11 04:32:09","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1343082,"visible":true,"origin":"","legend":"\u003cp\u003eFlowchart of severe endometrial fibrosis (SEF) patients, mild to moderate endometrial fibrosis (MMEF) patients, and healthy women enrolled in this study.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-4864304/v1/20724e911c94130148da84be.png"},{"id":66371533,"identity":"02cbcad7-1a0a-48e2-95f5-5ab90d04d9bf","added_by":"auto","created_at":"2024-10-11 04:40:09","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":2931477,"visible":true,"origin":"","legend":"\u003cp\u003eWorkflow of radiomics. UFS: univariate feature selection; RFE: recursive feature elimination; LB: logistic regression; SVM: support vector machine; NB: naïve bayes; GPC: gaussian process classification.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-4864304/v1/33aca88bb2c80985ddb387d1.png"},{"id":66371530,"identity":"e0155f52-e333-45da-b268-f1bf23a88c03","added_by":"auto","created_at":"2024-10-11 04:40:09","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":185773,"visible":true,"origin":"","legend":"\u003cp\u003ea: Heatmap of model\u003csub\u003e1\u003c/sub\u003e constructed with different filtering and classifiers, b: Confusion matrix of model\u003csub\u003e1\u003c/sub\u003e constructed with UFS-LR (test set).\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-4864304/v1/f72a34cdbf30d9df29a82c39.png"},{"id":66370494,"identity":"e4d4d2bc-f173-4947-8d14-08f3d7f3cfc7","added_by":"auto","created_at":"2024-10-11 04:32:09","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":959384,"visible":true,"origin":"","legend":"\u003cp\u003eROC curves for different models constructed with UFS-LR (test set); a, b, c, d, and e represent the ROC curves for model\u003csub\u003e1\u003c/sub\u003e, model\u003csub\u003e2\u003c/sub\u003e, model\u003csub\u003e3\u003c/sub\u003e, model\u003csub\u003e4\u003c/sub\u003e, and model\u003csub\u003e5\u003c/sub\u003e respectively.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-4864304/v1/f19d63653757315f27aa3d50.png"},{"id":66370492,"identity":"4e1f5247-7eb7-4e49-be75-ff67bceda649","added_by":"auto","created_at":"2024-10-11 04:32:09","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":288336,"visible":true,"origin":"","legend":"\u003cp\u003eComparison of the diagnostic performance of model\u003csub\u003e1\u003c/sub\u003e and model\u003csub\u003e5\u003c/sub\u003e constructed with UFS-LR for MMEF and SEF patients.\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-4864304/v1/0f96de6310eca115ed32c62b.png"},{"id":66372664,"identity":"4303d66d-b848-4a6e-8962-daa1fe9650cb","added_by":"auto","created_at":"2024-10-11 04:56:09","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":304825,"visible":true,"origin":"","legend":"\u003cp\u003eFeatures and their importance in model\u003csub\u003e1\u003c/sub\u003e (T2+ADC+Clinical Data) constructed via UFS-LR.\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-4864304/v1/995c67e18e0123ca5e88ee2e.png"},{"id":68320322,"identity":"2ca4a02f-f5c2-4ec5-a7d3-ed15f427e82d","added_by":"auto","created_at":"2024-11-06 04:24:50","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":6951159,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4864304/v1/8eec1da6-439f-4846-8edf-ec1d4e4a5cfc.pdf"},{"id":66371765,"identity":"2f450e5f-58b0-4cbe-aed4-68c964444e92","added_by":"auto","created_at":"2024-10-11 04:48:09","extension":"doc","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":14848,"visible":true,"origin":"","legend":"","description":"","filename":"Additionalfile1.doc","url":"https://assets-eu.researchsquare.com/files/rs-4864304/v1/de7ffbbc69753489ff95ec2e.doc"},{"id":66371535,"identity":"f2ff4e75-988d-4435-8452-0a9f2c8ef199","added_by":"auto","created_at":"2024-10-11 04:40:09","extension":"doc","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":33280,"visible":true,"origin":"","legend":"","description":"","filename":"Additionalfile2.doc","url":"https://assets-eu.researchsquare.com/files/rs-4864304/v1/ada623bf12565f8d04fbca5e.doc"}],"financialInterests":"No competing interests reported.","formattedTitle":"Multi-parametric MRI combined with radiomics for the diagnosis and grading of endometrial fibrosis","fulltext":[{"header":"Introduction","content":"\u003cp\u003eEndometrial fibrosis (EF) is the primary pathological feature of intrauterine adhesion (IUA), also known as Asherman syndrome[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. This syndrome arises from damage to the endometrial basal layer, leading to the formation of fibrosis and adhesion within the uterus, ultimately resulting in menstrual abnormalities, infertility, or recurrent miscarriages[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. IUA can be caused by various factors, among which induced abortion and postpartum curettage are important causes[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. In recent years, with the frequent occurrence of intrauterine operations and the increase in reproductive system infections, the prevalence of IUA has shown an increasing trend, and IUA has become the second leading cause of secondary infertility in Chinese women[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. This study adopted the scoring standard of the American Fertility Society (AFS), which categorizes patients into mild, moderate, and severe groups[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. The treatment options differ for patients with varying severity. Patients with mild to moderate adhesions can undergo hysteroscopic adhesiolysis, whereas severe cases exhibit significant fibrosis of the endometrial basalis, posing greater challenges for surgical separation of adhesions and consequently leading to a notably lower pregnancy rate than mild to moderate adhesions do[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Therefore, accurately assessing the degree of endometrial fibrosis is crucial for assisting clinicians in selecting the optimal treatment plan and evaluating treatment outcomes.\u003c/p\u003e \u003cp\u003eCurrently, there are multiple methods for diagnosing intrauterine adhesions. Trans-vaginal ultrasound (TVS) is primarily used for initial clinical screening[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Prior to the invention of hysteroscopy, hysterosalpingography (HSG) was the preferred method for observing the uterine cavity, but it has a relatively high false-positive rate and lower diagnostic accuracy[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Hysteroscopy is considered the \"gold standard\" technique for the diagnosis and treatment of IUA[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Although it has high diagnostic accuracy, it is invasive and can increase the risk of re-adhesion. Furthermore, it can only visualize pathological changes on the surface of the uterine cavity and cannot clearly demonstrate lesions in the myometrium or basal layer. MRI offers a non-invasive, radiation-free, and multi-sequence, multi-directional approach to clearly visualize uterine contours and adhesion bands. With the development of imaging technology, MRI and its new technologies have emerged for the diagnosis and treatment of endometrial fibrosis[\u003cspan additionalcitationids=\"CR13\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. T2WI and DWI are routinely acquired scanning sequences. On T2WI, the uterine cavity with adhesions appears as alternating high signals from normal endometrial tissue and low signals from areas of endometrial fibrosis[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Diffusion weighted imaging (DWI) is the most widely used method to non-invasively quantify water molecule diffusion by calculating the apparent diffusion coefficient (ADC) value. Studies have shown that when tissue fibrosis occurs, the apparent diffusion coefficient (ADC) decreases[\u003cspan additionalcitationids=\"CR17\" citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eTraditional imaging assessment is subjective. Compared with traditional imaging, radiomics can further mine and analyze the information contained in imaging images in a noninvasive manner by extracting thousands of quantitative features in a high-throughput fashion[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. Studies have shown that machine learning-based radiomics can be applied to tumor grading[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. The extracellular matrix (ECM) accumulates in the damaged endometrial layer, causing excessive proliferation of collagen fibers. Consequently, endometrial regeneration is inhibited, and the endometrium is gradually replaced by connective tissue. The pathological basis for radiomic differences in varying degrees of endometrial fibrosis may be due to differences in the degree of fibrosis and water content. On this basis, we hypothesize that radiomics can be used to distinguish between different degrees of endometrial fibrosis.\u003c/p\u003e \u003cp\u003eIn recent years, increasing evidence has suggested that morphological and MRI signal alterations occur in the endometrium during fibrosis[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. However, our study goes beyond these observations by utilizing MP-MRI, specifically T2WI and apparent diffusion coefficient (ADC) mapping, in conjunction with radiomics to develop a novel model capable of differentiating the degree of endometrial fibrosis, thereby enhancing diagnostic accuracy and providing more robust guidance for clinical diagnosis and treatment.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003ePatients\u003c/h2\u003e \u003cp\u003e This study adhered to the Helsinki Declaration and was approved by the the Institutional Ethics Committee of Nanjing Drum Tower Hospital with the approval number: 2019-051-02. Written informed consent was obtained from all patients and healthy female participants. The AFS scoring criteria were used (see Additional file 1)[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. A total score of 1\u0026ndash;4 is considered mild, a score of 5\u0026ndash;8 is considered moderate, and a score of 9\u0026ndash;12 is considered severe. In this study, patients with severe IUA were defined as having severe endometrial fibrosis (SEF), whereas patients with mild to moderate IUA were classified as having moderate endometrial fibrosis (MMEF). Infertile patients and healthy volunteers were prospectively enrolled from June 2018 to December 2023. All patients were diagnosed with endometrial fibrosis through hysteroscopy and pathology. Hysteroscopy was performed within 7 days after MRI. The hysteroscopy and endometrial biopsy of the patient were performed by our experienced gynecologists in the late proliferative phase. The gynecologists assessed the degree of intrauterine adhesions according to the AFS scoring criteria and performed endometrial biopsy at the obvious fibrotic scar site via biopsy forceps. For infertile patients with a normal endometrium (such as those with tubal obstruction), endometrial biopsy was performed from the anterior wall, posterior wall, or fundus of the uterus as a histological control for a normal endometrium. The inclusion criteria for patients were as follows: 1) aged 20\u0026ndash;40 years; 2) history of dilation and curettage (D\u0026amp;C) surgery; 3) clinical diagnosis of infertility (regular sexual activity without contraception for at least one year without conception); 4) endometrial fibrosis confirmed by hysteroscopy and pathological diagnosis; and 5) normal ovarian function. The inclusion criteria for healthy females were as follows: 1) aged 20\u0026ndash;40 years; 2) normal menstrual cycle and menstrual flow; and 3) no history of uterine injury or disease, such as D\u0026amp;C and intrauterine infections. The exclusion criteria were as follows: 1) history of other severe uterine diseases, such as adenomyosis, endometrial tuberculosis, endometritis, endometrial hyperplasia or polyps, as well as severe uterine malformations confirmed by the ESHRE/ESGE consensus[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]; 2) MRI contraindications, such as cardiac pacemakers and cochlear implants; and 3) severe MRI artifacts that result in unclear visualization and affect data analysis. The flowchart for case inclusion and exclusion is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eMRI examination method\u003c/h2\u003e \u003cp\u003eAll participants underwent scanning using a 16-channel phased-array body coil on a 3.0T MRI scanner (Ingenia, Philips Medical Systems, Best, The Netherlands) in the supine position with the head first. All patients and healthy females were instructed to empty their bladders before the MRI scan. Owing to the morphological and microstructural changes that occur in the endometrium during the menstrual cycle, MRI examinations were performed during ovulation (confirmed by ultrasonography of the dominant follicle) to improve the consistency of the results. The MRI sequences included sagittal T2WI and DWI of the uterus, with a total scan time of approximately 6 minutes. The T2WI scan parameters were as follows: repetition time (TR)/echo time (TE)\u0026thinsp;=\u0026thinsp;1700\u0026ndash;5000 ms/100 ms; matrix size\u0026thinsp;=\u0026thinsp;200\u0026times;166; field of view (FOV)\u0026thinsp;=\u0026thinsp;120 mm\u0026times;120 mm; slice thickness\u0026thinsp;=\u0026thinsp;3 mm; slice gap\u0026thinsp;=\u0026thinsp;0.3 mm; voxel size\u0026thinsp;=\u0026thinsp;0.60\u0026times;0.72\u0026times;3 mm\u003csup\u003e3\u003c/sup\u003e; number of signal averages (NSA)\u0026thinsp;=\u0026thinsp;1.1; scan duration\u0026thinsp;=\u0026thinsp;3 minutes and 56 seconds. The DWI scan parameters were as follows: b-values of 0 and 1000 s/mm\u003csup\u003e2\u003c/sup\u003e, TR/TE\u0026thinsp;=\u0026thinsp;6000 ms/90 ms; matrix size\u0026thinsp;=\u0026thinsp;133\u0026times;160; FOV\u0026thinsp;=\u0026thinsp;240 mm\u0026times;240 mm; slice thickness\u0026thinsp;=\u0026thinsp;3 mm; slice gap\u0026thinsp;=\u0026thinsp;0.3 mm; NSA\u0026thinsp;=\u0026thinsp;2. The scan duration was 2 minutes and 6 seconds.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003eImage Segmentation and Feature Extraction\u003c/h2\u003e \u003cp\u003eAll the images were independently analyzed by two radiologists (WHH and LHH with 12 and 4 years of experience in pelvic MRI interpretation, respectively), who were blinded to the patients\u0026rsquo; clinical information. The DWI images were loaded onto a workstation (Extended MR Workspace 2.6.3.4; Philips Medical Systems, Best, the Netherlands), and the ADC map was automatically generated via a monoexponential model[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. The T2WI, DWI, and ADC maps were imported into 3D Slicer software (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ewww.slicer.org/\u003c/a\u003e\u003c/span\u003e\u003cspan address=\"http://www.slicer.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) for segmentation of the volume of interest (VOI). T2WI and DWI were used as references to manually outline the contour of the endometrium layer by layer, covering as much of the endometrium as possible on each slice to accurately reflect the heterogeneity of the endometrium. This process automatically generates a 3D image (as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003eInter-observer Consistency Test\u003c/h2\u003e \u003cp\u003eThirty patients were randomly selected according to their date from each of the three groups, and two physicians independently completed the VOI segmentation and feature extraction. The inter-class correlation coefficient (ICC) was adopted for consistency testing. Features with an ICC value greater than 0.75 were considered to have good reproducibility, whereas those with an ICC value of 0.75 or less were excluded.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eFeature Selection, Prediction Model Construction and Evaluation\u003c/h2\u003e \u003cp\u003eFirst, highly stable features were selected on the basis of the intra-group and inter-group consistency (ICC\u0026thinsp;\u0026gt;\u0026thinsp;0.75) of feature extraction. To eliminate redundant features and enhance model efficiency, univariate feature selection (UFS) and recursive feature elimination (RFE) were employed for feature selection. Using logistic regression (LB), support vector machine (SVM), na\u0026iuml;ve bayes (NB), and Gaussian process classification (GPC) classifiers, 40 predictive models were established, including model\u003csub\u003e1\u003c/sub\u003e (T2\u0026thinsp;+\u0026thinsp;ADC\u0026thinsp;+\u0026thinsp;clinical data), model\u003csub\u003e2\u003c/sub\u003e (T2\u0026thinsp;+\u0026thinsp;ADC), model\u003csub\u003e3\u003c/sub\u003e (T2), model\u003csub\u003e4\u003c/sub\u003e (ADC), and model\u003csub\u003e5\u003c/sub\u003e (clinical data). The enrolled data were randomly stratified and divided into a train set (108 cases: 28 healthy women, 29 with MMEF, and 51 with SEF) and a test set (47 cases: 12 healthy women, 12 MMEF and 23 SEF) at a ratio of 7:3. The performance of each model in both the train and test sets was analyzed using receiver operating characteristic (ROC) curves. The average area under the curve (AUC), F1-score (the harmonic mean of recall and precision), and ACC were used to evaluate the performance of the different models.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cp\u003eThe statistical analysis of this study was conducted using Python (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.python.org/\u003c/span\u003e\u003cspan address=\"https://www.python.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). Quantitative data that conformed to a normal distribution are expressed as the mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation. One-way ANOVA was used to compare differences in continuous variables among the three groups, and the chi-square test was used to compare differences in categorical variables among the three groups. ROC curves were plotted via Python software. The diagnostic performance of the models was evaluated using ROC curve analysis, and the AUC value, F1-score, and ACC were calculated. A P value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered to indicate statistical significance.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003eGeneral Information\u003c/h2\u003e \u003cp\u003eA total of 74 patients with SFE, 41 patients with MMEF confirmed by hysteroscopy, and 40 healthy women were ultimately included. There were significant differences in age, body weight, endometrial thickness (EMT), and the length of the uterine cavity (LUC) among healthy women versus MMEF and healthy women versus SEF patients (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Furthermore, significant differences in EMT and LUC were detected between patients with MMEF and those with SEF. The general information is presented below (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eGeneral information of the enrolled cases\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\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=\"\u0026plusmn;\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGroup\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLabel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCases\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAge (y)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eWeight (kg)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eEMT (mm)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eLUC (mm)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHealthy women\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e40\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c4\"\u003e \u003cp\u003e28.4\u0026thinsp;\u0026plusmn;\u0026thinsp;3.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c5\"\u003e \u003cp\u003e52.0\u0026thinsp;\u0026plusmn;\u0026thinsp;4.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c6\"\u003e \u003cp\u003e11.4\u0026thinsp;\u0026plusmn;\u0026thinsp;1.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e39.9\u0026thinsp;\u0026plusmn;\u0026thinsp;2.6\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMMEF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e41\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c4\"\u003e \u003cp\u003e32.7\u0026thinsp;\u0026plusmn;\u0026thinsp;4.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c5\"\u003e \u003cp\u003e58.5\u0026thinsp;\u0026plusmn;\u0026thinsp;8.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c6\"\u003e \u003cp\u003e7.1\u0026thinsp;\u0026plusmn;\u0026thinsp;2.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e35.3\u0026thinsp;\u0026plusmn;\u0026thinsp;4.5\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSEF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e74\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c4\"\u003e \u003cp\u003e32.9\u0026thinsp;\u0026plusmn;\u0026thinsp;3.8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c5\"\u003e \u003cp\u003e58.9\u0026thinsp;\u0026plusmn;\u0026thinsp;7.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c6\"\u003e \u003cp\u003e6.1\u0026thinsp;\u0026plusmn;\u0026thinsp;1.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e36.9\u0026thinsp;\u0026plusmn;\u0026thinsp;3.8\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eEMT: endometrial thickness; LUC: length of uterine cavity; MMEF: mild to moderate endometrial fibrosis; SEF: severe endometrial fibrosis.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eInter-observer Consistency Test and Feature Selection\u003c/h2\u003e \u003cp\u003eBased on the consistency test analysis results of the radiomic features extracted by the two physicians, features with ICC\u0026thinsp;\u0026gt;\u0026thinsp;0.75 were retained. Ultimately, the numbers of selected features for T2 and ADC were 1055 and 1344, respectively. After UFS feature selection, the final numbers of features used to construct model\u003csub\u003e1\u003c/sub\u003e, model\u003csub\u003e2\u003c/sub\u003e, model\u003csub\u003e3\u003c/sub\u003e, model\u003csub\u003e4\u003c/sub\u003e, and model\u003csub\u003e5\u003c/sub\u003e were 16, 16, 16, 11, and 1, respectively. Following RFE feature selection, the final number of features used for model construction were 6, 6, 6, 6, and 4, respectively.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eComparison of Different Radiomic Models\u003c/h2\u003e \u003cp\u003eThe results indicated varying diagnostic efficiencies among the different predictive models. Among the 40 classification models, the \u0026ldquo;UFS-LR\u0026rdquo; model, which combines UFS with the LR classifier, performed the best, with an average AUC of 0.92 on the test set. Among the five models constructed via UFS-LR, model1, the ensemble model, exhibited the best performance, with average AUC, F1 score, and accuracy values of 0.92, 0.79, and 0.81, respectively (details in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e2\u003c/span\u003e). The corresponding heatmap and confusion matrix are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e. The ROC curves of the test sets for the different models constructed using UFS-LR are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e. The average AUC values of the T2-related models (model\u003csub\u003e1\u003c/sub\u003e, model\u003csub\u003e2\u003c/sub\u003e, model\u003csub\u003e3\u003c/sub\u003e) were higher than those of model\u003csub\u003e4\u003c/sub\u003e and model\u003csub\u003e5\u003c/sub\u003e, particularly in the MMEF and SEF groups. Model\u003csub\u003e1\u003c/sub\u003e enhances the diagnostic performance for differentiating between MMEF and SEF patients (as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e).\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 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDiagnostic performance of different models constructed using UFS-LR\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"14\"\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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c10\" colnum=\"10\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c11\" colnum=\"11\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c12\" colnum=\"12\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c13\" colnum=\"13\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c14\" colnum=\"14\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"5\" nameend=\"c6\" namest=\"c2\"\u003e \u003cp\u003eAUC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"5\" nameend=\"c12\" namest=\"c8\"\u003e \u003cp\u003eF1-score\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c13\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c14\"\u003e \u003cp\u003eACC\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMacro\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eMicro\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003eMacro\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003eWeighted\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c14\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003csub\u003e1\u003c/sub\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.95\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.88\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.79\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c14\"\u003e \u003cp\u003e0.81\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003csub\u003e2\u003c/sub\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.88\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.83\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.60\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.84\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c14\"\u003e \u003cp\u003e0.79\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003csub\u003e3\u003c/sub\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.97\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.88\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.90\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.67\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.84\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.78\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c14\"\u003e \u003cp\u003e0.80\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003csub\u003e4\u003c/sub\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.90\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.63\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.79\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.83\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.67\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.13\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.81\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.54\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.60\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c14\"\u003e \u003cp\u003e0.66\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003csub\u003e5\u003c/sub\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.95\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.52\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.84\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.83\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.78\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.61\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.65\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c14\"\u003e \u003cp\u003e0.88\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"14\"\u003eAUC: Area under the receiver operating characteristic curve; Macro: Macro-average; Micro: Micro-average; Weighted: Weighted-average; ACC: accuracy; 0: Healthy women; 1: Mild to moderate fibrosis (MMEF) ;2: Severe endometrial fibrosis (SEF).\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eAUC: area under the receiver operating characteristic curve; Macro: macro-average; Micro: micro-average; Weighted: weighted-average; ACC: accuracy; 0: healthy women; 1: mild to moderate fibrosis (MMEF) ;2: severe endometrial fibrosis (SEF).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eOptimal features and weights selected by different models\u003c/h2\u003e \u003cp\u003eThe optimal features and corresponding weights selected by machine learning methods vary. The optimal features of the combined models are detailed in see Additional file 2. The 16 features and their scores for model 1 constructed with UFS-LR are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eTo date, MRI studies on endometrial fibrosis have focused on morphological and signal changes[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. This study aimed to develop a model that is capable of effectively diagnosing and distinguishing the degree of endometrial fibrosis by utilizing MP-MRI (T2 and ADC) in combination with radiomics and clinical parameters. Our results showed that the use of multi-parametric radiomic model enables accurate identification of healthy volunteers, MMEF and SEF patients, it can also improve diagnostic accuracy for MMEF and SEF patients facilitating the development of personalized treatment plans and enhancing fertility outcomes.\u003c/p\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eImpact of different machine learning methods on predictive models\u003c/h2\u003e \u003cp\u003ePrevious radiomics studies have often relied solely on a single feature selection and classifier method, without a consensus on the classifier that performs best. On the basis of different machine learning methods, we established multiple radiomic models to predict the degree of endometrial fibrosis. We employed the AUC, F1 score, and ACC to evaluate the performance of the different models. The results demonstrated varying diagnostic efficiencies among different prediction models. Notably, the UFS-LR model outperformed the other classification models. UFS selects the best features on the basis of univariate statistical tests, making it suitable for identifying features most relevant to the target variable. Applying an appropriate classifier can potentially enhance classifier performance. The LR method is widely used in classification tasks because of its low computational requirements and the ability to intuitively observe the predicted probabilities of samples. The results of all the machine learning methods indicate that the diagnostic performance of model\u003csub\u003e1\u003c/sub\u003e is greater than that of model\u003csub\u003e5\u003c/sub\u003e. The results indicated that the optimal classifier varied with different feature selections, which is consistent with the findings of Wang et al[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. However, all the models exhibited excellent diagnostic efficacy, displaying consistent trends in their diagnostic performance across the different models. Therefore, when modeling with diverse radiomics features, attempting to utilize different selection methods and classifiers can enhance model performance and result reproducibility.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eAnalysis of the value of radiomics in predicting endometrial fibrosis\u003c/h2\u003e \u003cp\u003eIn clinical practice, the treatment approach varies according to the severity of fibrosis, and the \u0026ldquo;gold standard\u0026rdquo; for assessing EF is hysteroscopy[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Under hysteroscopy, the normal endometrial morphology is characterized by densely distributed glandular openings on the surface, a bright red color, a soft texture, and good elasticity. The uterine cavity appears regular in shape, forming an inverted triangle, with clearly visible bilateral tubal ostia. During hysteroscopy in patients, a pale endometrium and dense fibrous scar tissue covering the entire uterine cavity can be observed. The uterine cavity has lost its original softness, becoming rigid and inelastic. Transcervical resection of adhesion (TCRA) is the primary treatment method aimed at restoring the normal volume and shape of the endometrial cavity[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. Postoperative estrogen therapy can stimulate endometrial regeneration and re-epithelialization of scar surfaces. Mild to moderate adhesions usually involve a small area or only the endometrial layer, making surgical separation and release of adhesions relatively easy. For patients with severe IUA, marked fibrosis of the endometrial basalis layer leads to localized or widespread decreased elasticity of the uterine wall, increasing the difficulty of surgical separation of adhesions and the risk of complications such as uterine perforation during surgery. Additionally, the postoperative recurrence rate is high, posing significant challenges in improving postoperative adhesion outcomes[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. Recent reports have suggested that transplanting stem cells into the uterine cavity or injecting growth factors into the endometrium of patients with severe fibrosis can reduce fibrosis areas, increase glandular counts, promote angiogenesis, and enhance pregnancy rates, demonstrating the potential for treating severe endometrial fibrosis[\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. Additionally, studies have shown that ECM scaffolds (SIS, Small intestinal submucosa) combined with intrauterine balloon therapy can repair endometrial fibrosis and improve IUA, providing a novel treatment approach for improving pregnancy outcomes in patients with moderate-to-severe IUA-related infertility[\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. Therefore, accurate early assessment of the degree of endometrial fibrosis and treatment response is crucial for assisting clinicians in selecting optimal treatment regimens and evaluating outcomes. Model\u003csub\u003e1\u003c/sub\u003e demonstrated demonstrates the highest diagnostic performance (AUC\u0026thinsp;=\u0026thinsp;0.92 in the test set), surpassing the diagnostic accuracy of model 5, which relies solely on clinical parameters (AUC\u0026thinsp;=\u0026thinsp;0.75 in the test set). Notably, the diagnostic performance of the radiomics model in predicting MMEF and SEF is lower than that for healthy volunteers, which aligns with clinical observations. Purely clinical indicators exhibit limited effectiveness in distinguishing between MMEF and SEF. The combined machine learning radiomics model enhances the diagnostic accuracy in differentiating between MMEF and SEF, with model 1 and model 5, constructed via UFE-LR, achieving AUC values of 0.88 versus 0.52 and 0.92 versus 0.77, respectively. This approach can reduce subjectivity in clinical assessments of fibrosis severity, avoid excessive invasive hysteroscopy, and improve patient outcomes.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eThe potential of radiomics in predicting endometrial fibrosis\u003c/h2\u003e \u003cp\u003eThe morphological and functional information provided by MP-MRI is insufficient for a comprehensive and effective assessment of tissue biological characteristics, necessitating the exploration of new technologies and methods. Radiomics can mine quantitative image features from medical images, enabling the quantification of heterogeneity and microstructure information beyond the capability of the human visual system[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. Its process includes image segmentation, feature extraction, feature selection, and model building and validation. Among these, image segmentation is a critical step in texture analysis technology, as it serves as the direct source for feature extraction[\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. The use of whole-lesion VOI delineation to acquire features involves more spatial information than features extracted on the basis of a single region of interest (ROI) and can characterize the heterogeneity of the entire tumor. This study adopted whole-lesion VOI delineation, avoiding the limitations of selecting partial lesions or extracting ROIs from a single slice. By extracting radiomic features from multiple sequences of MP-MRI, this method maximizes diversity and increases the number of features, thereby enhancing the completeness of lesion description, improving the accuracy of lesion heterogeneity characterization, and reducing sampling errors[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThe features from different MP-MRI sequences reflect distinct biological information, and the comprehensive utilization of these image features allows for a more comprehensive quantitative assessment of lesion heterogeneity. T2WI serves as a fundamental sequence in MRI composition, capable of revealing a wealth of histopathological features such as water content, degree of fibrosis, necrosis, hemorrhage, etc.[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. The ADC value quantitatively assesses the diffusion and restriction of water molecules within the target tissue, elucidating cell density and facilitating the differentiation of degrees of fibrosis. Studies have shown that the ADC decreases when tissue fibrosis occurs[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Consequently, this study employed T2WI and ADC sequences, with the potential for combining T2WI, DWI and clinical parameters to become a novel indicator for differentiating the degree of endometrial fibrosis. The T2-related models had higher average AUC values than model4 and model5 did, especially in the MMEF and SEF groups. A total of 16 features were ultimately selected through UFS-LR modeling model1, including 1 clinical feature (EMT), 5 ADC features, and 10 T2 features, with T2 features accounting for a larger proportion. Notably, the feature with the highest weight was the T2 wavelet feature (T2_wavelet-LLL_gldm_DependenceVariance). Different radiomics methods result in varying optimal features and weights, with T2-related features being the most numerous and having the greatest weight. The endometrium is lined by a luminal epithelium and contains tubular glands that radiate from the surface to the endometrial\u0026ndash;myometrial interface[\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. The water content of the endometrium is an important component of its normal physiological function. T2WI can microscopically reflect this water content and the degree of fibrosis. In patients with fibrosis, the endometrial structure becomes disrupted and replaced by fibrous tissue, leading to decreased water content[\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]. Therefore, T2 features may hold greater value in predicting endometrial fibrosis.\u003c/p\u003e \u003cp\u003eThe aforementioned studies underscore the significant role of radiomics features in enhancing the accuracy of grading endometrial fibrosis. Currently, there are multiple criteria for assessing the severity of intrauterine adhesions, and future endeavors may explore incorporating MRI features, particularly T2-related features, into scoring systems. This approach has the potential to contribute to the refinement of endometrial fibrosis scoring.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eLimitations\u003c/h2\u003e \u003cp\u003eFirst, the sample size was relatively small, and as this was a single-center study, there may be some selection bias. Second, image segmentation is based on manual delineation, which is currently the most commonly used research method but inevitably involves some subjectivity. In future studies, deep learning techniques, such as automatic or semi-automatic segmentation methods, can be employed to replace the current time-consuming and labor-intensive manual segmentation methods. Third, there are few clinical parameters of patients, and more clinical data can be incorporated into the analysis, for example, correlation studies combining clinical features and pathological characteristics. Finally, the inherent shortcomings of radiomics analysis techniques lead to poor reproducibility, requiring larger sample sizes and multi-center validation to achieve clinical translation. Nonetheless, this is sufficient to demonstrate the feasibility of using MP-MRI combined with radiomics for the quantitative assessment of endometrial fibrosis.\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn summary, the radiomics model we established is feasible for differentiating healthy women, patients with MMEF and patients with SEF. This prospective study demonstrated the ability of T2 combined with ADC radiomics as a non-invasive imaging biomarker for the quantitative diagnosis and assessment of endometrial fibrosis. Compared with clinical parameters, morphological and signal changes, combined radiomics can provide additional microstructural information and heterogeneity, which not only aids clinicians in objectively diagnosing and assessing endometrial fibrosis but also facilitates the arrangement of timely and effective treatments, thereby offering certain assistance in clinical decision-making.In summary, the radiomics model we have established is feasible for differentiating healthy women, MMEF and SEF patients. This prospective study demonstrates the capability of T2 combined with ADC radiomics as a non-invasive imaging biomarker for quantitative diagnosis and assessment of endometrial fibrosis. Compared to clinical parameters, morphological and signal changes, combined radiomics can provide additional microstructural information and heterogeneity, which not only aids clinicians in objectively diagnosing and assessing endometrial fibrosis, but also facilitates the arrangement of timely and effective treatments, thereby offering certain assistance in clinical decision-making.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eMRI Magnetic resonance imaging\u003c/p\u003e\n\u003cp\u003eMP-MRI Multi-parametric MRI\u003c/p\u003e\n\u003cp\u003eSEF Severe endometrial fibrosis\u003c/p\u003e\n\u003cp\u003eMMEF Mild to moderate fibrosis\u003c/p\u003e\n\u003cp\u003eVOI Volume of interest\u003c/p\u003e\n\u003cp\u003eDWI Diffusion weighted imaging\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eADC Apparent diffusion coefficient\u003c/p\u003e\n\u003cp\u003eUFS Unsupervised feature selection\u003c/p\u003e\n\u003cp\u003eLR Logistic regression\u003c/p\u003e\n\u003cp\u003eRFE Recursive feature elimination\u003c/p\u003e\n\u003cp\u003eSVM Support vector machine\u003c/p\u003e\n\u003cp\u003eNB Na\u0026iuml;ve bayes\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eGPC Gaussian process classification\u003c/p\u003e"},{"header":"Declarations","content":" \u003ch2\u003eEthics approval and consent to participate\u003c/strong\u003e \u003cp\u003e All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.This study was approved by the Institutional Ethics Committee of Nanjing Drum Tower Hospital, Nanjing, China (Ethic approval No.2019-051-02). Written informed consent was obtained from all patients and healthy female participants.\u003c/p\u003e\u003ch2\u003eConsent for publication\u003c/strong\u003e \u003cp\u003eNot applicable.\u003c/p\u003e \u003ch2\u003eCompeting interests\u003c/strong\u003e \u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e \u003ch2\u003eFunding\u003c/h2\u003e \u003cp\u003eThis study was supported by the National Natural Science Foundation of China (81671751 and 81871410) and the Strategic Priority Research Program of the Chinese Academy of Sciences (XDA16040302).\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eZZY, HYL,LDY: Conceptualization, Methodology, Project administration Supervision, Writing - Review \u0026amp; Editing. WHH: Investigation, Data curation, Formal analysis, Visualization, Writing - Original Draft. ZL,ZH,LHH: Investigation, Visualization, Data Curation, Writing - Review \u0026amp; Editing.ZH, MJ,HJL: Resources, Investigation, Writing - Review \u0026amp; Editing.All authors reviewed the manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgments\u003c/h2\u003e \u003cp\u003eNot applicable.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe data that support the findings of this study are not openly available due to reasons of sensitivity and are available from the corresponding author upon reasonable request. Data are located in controlled access data storage at Nanjing Drum Tower Hospital.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eZhu Q, Yao S, Ye Z, et al. Ferroptosis contributes to endometrial fibrosis in intrauterine Adhesions. Free Radical Bio Med. 2023;205:151\u0026ndash;62.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWallach EE, Schenker JG, Margalioth EJ. Intrauterine adhesions: An updated Appraisal. Fertil Steril. 1982;37(5):593\u0026ndash;610.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAsherman JG. Traumatic intra-uterine adhesions. BJOG. 1950;57(6):892\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMarch CM. Management of Asherman\u0026rsquo;s syndrome. Reprod Biomed Online. 2011;23(1):63\u0026ndash;76.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChinese Medical Association Obstetrics and Gynecology Branch. Chinese Expert Consensus on Clinical Diagnosis and Treatment of Intrauterine adhesions [J]. Chin J Obstet Gynecol. 2015;50(12):881\u0026ndash;7.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThe American Fertility. Society classifications of adnexal adhesions, distal tubal occlusion, tubal occlusion secondary to tubal ligation, tubal pregnancies, M\u0026uuml;llerian anomalies and intrauterine Adhesions. Fertil Steril. 1988;49(6):944\u0026ndash;55.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAAGL Advancing Minimally Invasive Gynecology Worldwide. AAGL practice report: Practice guidelines for management of intrauterine Synechiae. J Minim Invasive Gynecol. 2010;17(1):1\u0026ndash;7.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYu D, Wong YM, Cheong Y, Xia E, Li TC. Asherman Syndrome\u0026ndash;one century Later. Fertil Steril. 2008;89(4):759\u0026ndash;79.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSoares SR, dos Barbosa MM, Camargos AF. Diagnostic accuracy of sonohysterography, transvaginal sonography, and hysterosalpingography in patients with uterine cavity diseases. Fertil Steril. 2000;73(2):406\u0026ndash;11.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAcholonu UC, Silberzweig J, Stein DE, Keltz M. Hysterosalpingography versus sonohysterography for intrauterine abnormalities. JSLS. 2011;15(4):471\u0026ndash;4.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang JH, Chen MJ, Chen CD, Chen SU, Ho HN, Yang YS. Optimal waiting period for subsequent fertility treatment after various hysteroscopic surgeries. Fertil Steril. 2013;99(7):2092\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou N, Zhu H, Jiang P, et al. Quantification of Endometrial Fibrosis Using Noninvasive MRI T2 Mapping: Initial Findings. J Magn Reson Imaging. 2023;58(6):1703\u0026ndash;13.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFeng Y, Jiang P, Hu Q, et al. Initial Experience With Diffusion-Weighted Magnetic Resonance Imaging for the Evaluation of Endometrial Fibrosis. J Comput Assist Tomogr. 2021;45(3):383\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKang K, Wang A, Wu H. MRI for Diagnosing Intrauterine Adhesions. J Obstet Gynaecol Can. 2023;45(10):102168.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBacelar AC, Wilcock D, Powell M, Worthington BS. The value of MRI in the Assessment of Traumatic Intra-uterine adhesions (Asherman\u0026rsquo;s Syndrome). Clin Radiol. 1995;50(2):80\u0026ndash;3.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJiang H, Chen J, Gao R, Huang Z, Wu M, Song B. Liver fibrosis staging with diffusion-weighted imaging: a systematic review and meta-analysis. Abdom Radiol (NY). 2017;42(2):490\u0026ndash;501.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCaruso A, Angriman I, Scarpa M, et al. Diffusion-weighted magnetic resonance for assessing fibrosis in Crohn\u0026rsquo;s disease. Abdom Radiol (NY). 2020;45(8):2327\u0026ndash;35.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBalci NC, Perman WH, Saglam S, Akisik F, Fattahi R, Bilgin M. Diffusion-weighted magnetic resonance imaging of the pancreas. Top Magn Reson Imaging. 2009;20(1):43\u0026ndash;7.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKocak B, Baessler B, Bakas S, et al. CheckList for EvaluAtion of Radiomics research (CLEAR): A step-by-step reporting guideline for authors and reviewers endorsed by ESR and EuSoMII. Insights Imaging. 2023;14(1):75.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLambin P, Leijenaar RTH, Deist TM, et al. Radiomics: the bridge between medical imaging and personalized medicine. Nat Rev Clin Oncol. 2017;14(12):749\u0026ndash;62.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu J, Zhao Y, Li M, et al. Machine learning-based radiomics analysis in predicting the meningioma grade using multiparametric MRI. Eur J Radiol. 2020;10(131):109251.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKumar A, Jha AK, Agarwal JP, et al. Machine-Learning-Based Radiomics for Classifying Glioma Grade from Magnetic Resonance Images of the Brain. J Pers Med. 2023;13(6):920.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrimbizis GF, Gordts S, Di Spiezio Sardo A, et al. The ESHRE/ESGE consensus on the classification of female genital tract congenital Anomalies. Hum Reprod. 2013;28(8):2032\u0026ndash;44.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKwee TC, Takahara T, Ochiai R, et al. Whole-body diffusion-weighted magnetic resonance Imaging. Eur J Radiol. 2009;70(3):409\u0026ndash;17.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang J, Chen J, Zhou R, Gao Y, Li J. Machine Learning-based multiparametric MRI radiomics for predicting poor responders after neoadjuvant chemoradiotherapy in rectal Cancer Patients. BMC Cancer. 2022;22(1):420.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSalazar CA, Isaacson K, Morris S. A comprehensive review of Asherman\u0026rsquo;s syndrome: Causes, symptoms and treatment Options. Curr Opin Obstet Gynecol. 2017;29(4):249\u0026ndash;56.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHealy MW, Schexnayder B, Connell MT, et al. Intrauterine adhesion prevention after hysteroscopy: A systematic review and meta-Analysis. Am J Obstet Gynecol. 2016;215(3):267\u0026ndash;75.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGharibeh N, Aghebati-Maleki L, Madani J, Pourakbari R, Yousefi M, Ahmadian Heris J. Cell-based therapy in thin endometrium and Asherman Syndrome. Stem Cell Res Ther. 2022;13(1):33.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePang WJ, Zhang Q, Ding HX, Sun NX, Li W. Effect of new biological patch in repairing intrauterine adhesion and improving clinical pregnancy outcome in infertile women: Study protocol for a randomized controlled Trial. Trials. 2022;23(1):510.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZwanenburg A, Valli\u0026egrave;res M, Abdalah MA, et al. The Image Biomarker Standardization Initiative: Standardized Quantitative Radiomics for High-Throughput Image-based Phenotyping. Radiology. 2020;295(2):328\u0026ndash;38.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRosenkrantz AB, Triolo MJ, Melamed J, Rusinek H, Taneja SS, Deng FM. Whole-lesion apparent diffusion coefficient metrics as a marker of percentage Gleason 4 component within Gleason 7 prostate cancer at radical Prostatectomy. J Magn Reson Imaging. 2015;41(3):708\u0026ndash;14.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSpencer TE, Hayashi K, Hu J, Carpenter KD. Comparative developmental biology of the mammalian uterus. Curr Top Dev Biol. 2005;68:85\u0026ndash;122.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCameron IL, Ord VA, Fullerton GD. Characterization of proton NMR relaxation times in normal and pathological tissues by correlation with other tissue parameters. Magn Reson Imaging. 1984;2(2):97\u0026ndash;106.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDe Kock I, Bos S, Delrue L, et al. MRI texture analysis of T2-weighted images is preferred over magnetization transfer imaging for readily longitudinal quantification of gut fibrosis. Eur Radiol. 2023;33(9):5943\u0026ndash;52.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Endometrium, Fibrosis, Magnetic Resonance Imaging, Radiomics, Machine Learning","lastPublishedDoi":"10.21203/rs.3.rs-4864304/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4864304/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e \u003cb\u003eBackground\u003c/b\u003e Accurate evaluation of endometrial fibrosis can help clinicians schedule individual treatment. This study aims to explore the application value of multi-parametric MRI (MP-MRI) combined with radiomics in the diagnosis and grading of endometrial fibrosis, aiming to construct models that can effectively distinguish endometrial fibrosis and compare the diagnostic performance of radiomics models established by different machine learning algorithms.\u003c/p\u003e \u003cp\u003e \u003cb\u003eMethods\u003c/b\u003e A total of 74 patients with severe endometrial fibrosis(SEF), 41 patients with mild to moderate fibrosis (MMEF)confirmed by hysteroscopy, and 40 healthy women of reproductive age were prospectively enrolled. All participants underwent T2 and DWI sequence scans during the periovulatory period. By freely delineating the volume of interest (VOI) of the endometrium in three subgroups, radiomic features were extracted and selected. Two feature selection methods and four machine learning classifiers were combined in pairs to establish five prediction models [model\u003csub\u003e1\u003c/sub\u003e (T2\u0026thinsp;+\u0026thinsp;ADC\u0026thinsp;+\u0026thinsp;clinical data), model\u003csub\u003e2\u003c/sub\u003e (T2\u0026thinsp;+\u0026thinsp;ADC), model\u003csub\u003e3\u003c/sub\u003e (T2), model\u003csub\u003e4\u003c/sub\u003e (ADC), and model\u003csub\u003e5\u003c/sub\u003e (clinical data)], resulting in a total of 40 classification models. The predictive performance of all models was evaluated using the area under the curve (AUC), F1 score, and accuracy (ACC).\u003c/p\u003e \u003cp\u003e \u003cb\u003eResults\u003c/b\u003e Among the 40 classification models, the \"UFS-LR\" model, which combined unsupervised feature selection (UFS) with the logistic regression (LR) classifier, performed the best, with an average AUC of 0.92 on the test set. Among the five models constructed via UFS-LR, model\u003csub\u003e1\u003c/sub\u003e exhibited the best performance, with average AUC, F1 score, and ACC values of 0.92, 0.79, and 0.81, respectively. The T2-related models had higher average AUC values than model\u003csub\u003e4\u003c/sub\u003e and model\u003csub\u003e5\u003c/sub\u003e did, especially in the MMEF and SEF groups. Among the optimal features selected from different models, T2-related features accounted for the largest number and had the highest weight.\u003c/p\u003e \u003cp\u003e \u003cb\u003eConclusions\u003c/b\u003e Machine learning-based MP-MRI radiomics analysis exhibited excellent performance in grading endometrial fibrosis and has great potential for providing robust support for clinical diagnosis and treatment.\u003c/p\u003e","manuscriptTitle":"Multi-parametric MRI combined with radiomics for the diagnosis and grading of endometrial fibrosis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-10-11 04:32:04","doi":"10.21203/rs.3.rs-4864304/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"dd5313af-5c92-4370-af31-b4a95d727b8c","owner":[],"postedDate":"October 11th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-11-06T04:24:05+00:00","versionOfRecord":[],"versionCreatedAt":"2024-10-11 04:32:04","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4864304","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4864304","identity":"rs-4864304","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}