Development and validation of a radiomics deep learning signature from MRI-guided transrectal ultrasound combined with ultrasound imaging parameters for prostate cancer prediction

Development and validation of a radiomics deep learning signature from MRI-guided transrectal ultrasound combined with ultrasound imaging parameters for prostate cancer prediction | 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 Development and validation of a radiomics deep learning signature from MRI-guided transrectal ultrasound combined with ultrasound imaging parameters for prostate cancer prediction Senlin Bao, Qian Liu, Ran Sun, Yue Sun, Li Yang, Danyan Liang, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7653994/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 5 You are reading this latest preprint version Abstract Objective This study aimed to develop and validate an integrated radiomics deep learning signature combining MRI-guided transrectal ultrasound (TRUS) and contrast-enhanced ultrasound (CEUS) parameters for improved prediction of prostate cancer (PCa). Methods This bicentric retrospective study enrolled 443 patients with suspected PCa confirmed by histopathology. Each patient underwent multiparametric MRI and TRUS-guided CEUS. Radiomic and deep learning features were extracted from B-mode ultrasound images using an AI-based platform. Feature selection was performed using statistical and regression methods. Machine learning classifiers were developed and merged with clinical parameters into a combined model. Performance was evaluated via ROC analysis, calibration curves, and decision curve analysis. Results The SVM-based radiomics model achieved an area under the curve (AUC) of 0.936 (95% CI: 0.905–0.966) in the training cohort and 0.823 (95% CI: 0.721–0.925) in the validation cohort. The clinical model alone yielded an AUC of 0.856 (95% CI: 0.812–0.901) in the training cohort and 0.809 (95% CI: 0.708–0.910) in the validation cohort. The integrated radiomics-clinical model demonstrated superior performance, with an AUC of 0.956 (95% CI: 0.931–0.981) in the training cohort and 0.889 (95% CI: 0.813–0.966) in the validation cohort. DCA confirmed the clinical utility of the combined model across a wide range of threshold probabilities. Conclusion The integration of radiomics deep learning features with CEUS parameters and clinical risk factors significantly enhances the accuracy of PCa prediction. This non-invasive approach shows promise for supporting clinical decision-making and reducing unnecessary biopsies. Prostate cancer radiomics deep learning MRI-TRUS fusion contrast-enhanced ultrasound machine learning predictive model Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction Prostate cancer (PCa) is the foremost genitourinary cancer worldwide and the second most prevalent malignancy in the male population [ 1 ] . By 2024, it is estimated that the United States will have seen a total of 299,010 newly diagnosed cases of prostate cancer, with an anticipated 35,250 deaths attributable to the disease [ 2 ] . Therefore, accurate diagnosis of PCa is crucial for enhancing survival rates and prognosis in patients. Diagnosing prostate cancer is challenging due to the similarity of symptoms with other prostatitis-related conditions. Conventional approaches for the detection of prostate cancer principally encompass serum prostate-specific antigen (PSA) testing, digital rectal examination (DRE), and transrectal ultrasound (TRUS) guided needle biopsies. However, research has indicated that these methods exhibit low sensitivity and specificity [ 3 ] , and may also lead to complications such as infection, bleeding, and erectile dysfunction [ 4 , 5 ] . Many patients undergo unnecessary biopsies due to benign prostatic hyperplasia. Furthermore, there are numerous limitations associated with ultrasound-guided prostate puncture, and the primary drawbacks of transrectal or perineal prostate system puncture biopsy include false negatives, missed diagnosis of high-risk prostate cancer, and overdiagnosis [ 6 ] . Therefore, a non-invasive and accurate diagnosis of prostate cancer holds significant importance. MRI provides better anatomical resolution and higher accuracy in the diagnosis of PCa, but the system is not without its limitations with regard to the detection of angiogenesis in PCa, and may also misdiagnose or miss some csPCa with a proportion of 58% [ 7 , 8 ] . Consequently, the MRI diagnosis of PCa remains a clinical challenge that necessitates further investigation. TRUS can accurately assess the size, shape, structure, and abnormalities of the prostate, providing real-time imaging that aids in timely diagnosis. This is of significance for the early detection of prostate cancer and the evaluation of the extent of lesions. Therefore, TRUS has significant value in the diagnosis of prostate diseases, particularly in its outstanding performance in early screening and localization of prostate cancer [ 9 ] . As an emerging ultrasound technology, blood flow patterns can be dynamically visualized and even quantitatively analyzed by contrast agents [ 10 ] . Even at an early stage, CEUS can detect elevated blood flow resulting from new blood vessel formation associated to PCa [ 11 , 12 ] . In comparison with mp-MRI, CEUS is characterised by certain disadvantages, including the challenge of localising suspicious lesions due to suboptimal temporal and spatial resolutions [ 13 ] . However, studies have shown that CEUS when used in conjunction with MRI can enhance the diagnostic accuracy of suspicious lesiones [ 14 ] . Additionally, the intersection of radiomics and deep learning represents a burgeoning interdisciplinary domain that integrates medical imaging with computer science to extract extensive quantitative data from medical images, establishing connections between these features and the onset, progression, and prognosis of diseases, which has great potential for clinical diagnosis and treatment. It is acknowledged that preceding studies have progressively highlighted the utilisation of radiomics or deep learning techniques in the analysis of CT or MRI examinations [ 15 ] , but few studies have combined radiomics and deep learning methods with CEUS parameters, especially under the guidance of MRI [ 16 – 19 ] . Therefore, we postulated that the integration of CEUS parameters and ultrasound radiomics might result in a more precise diagnosis of prostate cancer. The objective of the present study is to employ artificial and deep learning feature extraction approaches to extract multi-dimensional features from MP-MRI-guided B-mode ultrasound images of prostate nodules for constructing a radiomics model, and to establish a novel prostate cancer prediction model combined with contrast-enhanced ultrasound parameters and clinical risk factors for clinical application. Materials and methods 1.Patients and data acquisition The present bicentric retrospective study was approved by the ethics committees of the relevant institutional reviewing boards at all collaborating hospitals, and the requirement for written consent was waived. A retrospective data collection was conducted on 504 patients with histopathologically-validated suspected prostate cancer at two hospitals between November 2019 and September 2023. In the end, 443 patients with prostate disease formed the study group. The training and internal testing cohorts of Hospital 1 (Inner Mongolia People's Hospital) comprised 270 and 68 patients, correspondingly. The external validation cohort of Hospital 2 (Tongliao People's Hospital) comprised 105 patients. The criteria that were utilised to determine the inclusion and exclusion of studies in this analysis are illustrated in the Fig. 1 . The following data is to be collated on every patient:Initial clinical and ultrasound image data, such as Age, prostate size [left and right diameter(LRD), upper and lower diameter (ULD), anteroposterior diameter(AD)], prostate volume(PV), total PSA (tPSA), free PSA (fPSA), peak systolic and diastolic blood flow(PS/PD), contrast agent entry time and full filling time(ET/FT), related angiographic parameters were retrieved from the medical documentation. The configuration of the study design and the pipeline are demonstrated in Fig. 2 . 2.Multiparametric magnetic resonance imaging (mp‑MRI) acquisition and interpretation Mp-MRI was carried out using a GE Signa HDxt 3.0 T magnet. The sequences included T2-weighted fast spin-echo imaging (T2WI), diffusion weighted MR Imaging (DWI) and dynamic contrast-enhanced MRI (DCE-MRI), with a high b value of 1500 s/mm 2 and a slice thickness of 3.0 mm. The analysis of MR images was conducted by two radiologists, who utilised the Prostate Imaging Reporting and Data System version 2.1 (PI-RADS v2.1) for this purpose [ 20 ] . If there were discrepancies, the PI-RADS score was finally determined by another senior radiologist. Images in Digital Imaging and Communications in Medicine (DICOM) format of mpMRI were uploaded to Volumetric navigation system in Logiq GE 9 R6 (GE HealthCare, USA). The motion and direction of the probe were transmitted to the fusion system by triangulation registration between a magnetic field generator and an electromagnetic tracking sensor on an IC5-9-D end-scan probe (2.4–9.6 MHz) equipped with a dedicated puncture frame. Targeted lesions were identified with reference to MRI. In patients exhibiting PI-RADS 2–5 scores, the lesion that attained the highest PI-RADS rating (or possessed the most substantial diameter) was designated as the CEUS target lesion. The motion of the ultrasound image presented by the probe was synchronised with the MRI image to ensure accurate acquisition of the two-dimensional ultrasound image. 3.Ultrasound imaging acquisition and image segmentation The size and blood flow of the prostate were observed and recorded by transrectal ultrasound. CEUS was performed on the largest cross-sectional section of the targeted lesion. A total of 2.4 ml of SonoVue (Bracco, Italy) was injected into the median cubital vein via a bolus injection, followed promptly by 5 ml of standard saline. Lesion development was observed continuously on fixed sections for 180s, and dynamic Dicom images were retained. One or two cores were taken from each target, and then the fusion system was withdrawn into the sagittal plane for a standardized biopsy protocol (basal, middle, or apical glands; Left or right, 12 points), using a biopsy gun and an 18G biopsy needle, with a penetration depth of 15 mm or 20 mm. The sample was placed in the specimen bottle. TIC curves were analyzed by calculating the video footage of CEUS with the imaging analysis software of the machine. Three ROIs are placed at the location where the contrast agent first enters, and the Wash-in equation [F(t) = A(1 - exp(-kt)) + B] is used to fit the curve to obtain relevant contrast parameters: Peak intensity average gradient (Grad), arrival time (ATM), time to peak(TtoP), Area under the curve (Area), fitting curve equation coefficient (A, B, k), fitting curve equation error (MSE). The average of the TIC parameters obtained from three measurements is included in the clinical information. The images were imported into ITK-SNAP(version 4.0.2), an open-source software program, and the ROIs were manually segmented in B-mode ultrasound images compared with mp-MRI images. The completed outline image was saved as a mask file in nii format. After 2 months, 50 patients were randomly selected, and the region of interest was redelineated by 2 radiologists. The intraclass correlation coefficient (ICC) has been utilised as a metric to assess the consistency of tumour delineation. 4.Radiomics feature extraction and deep learning features extraction In this study, the "One-key AI v3.2.2" platform ( http://www.medai.icu/ , which is based on Pytorch 1.8.0) was used to extract 1593 radiomics features and 2048 deep learning features from B-mode ultrasound images. The regions of interest were then cropped with the largest areas that contained the entire tumour. The grey-scale images were processed to ensure background information was standardised and minimise noise disruption prior to imaging. Employing this methodology, the input image was reduced to a dimension of 224 × 224 pixels through linear interpolation, whilst normalising the pixel intensity mean and variance to 0 and 1, thus ensuring consistency in data representation. Radiomics Extraction: The hand crafted features(HCR) can be classified into three categories: geometric features, intensity features, and texture features. These are used to depict the two-dimensional shape characteristics of tumors, the first-order statistical distribution of intensity within tumor voxels, and intensity patterns, i.e., the first-order and higher-order spatial distributions of intensity. Here, texture features are extracted using the gray-level co-occurrence matrix (GLCM), gray-level run length matrix (GLRLM), gray-level size zone matrix (GLSZM), and neighborhood gray-level difference matrix (NGTDM) methods. A full overview of the HCR features extracted in this project is available in the Pyradiomics manual ( http://pyradiomics.readthedocs.io ). The ROI with the maximum possible sagittal area was chosen for precise trimming before the DTL features were extracted. In this study, ResNet50 was used as the CNN model to extract deep learning features. ResNet50 was pre-trained on the ILSVRC-2012 database ( https://www.image-net.org/challenges/LSVRC/2012/ ). The slice with the largest tumour area was selected to depict every patient. Once the deep learning model had been trained, the features from the average pooled layer were taken as the deep learning features. 5.Feature selection and model construction We performed statistical screening of radiomic features using the Mann-Whitney U test. Only features with a significance level of p < 0.05 were retained. Subsequently, Spearman’s rank correlation analysis was applied to evaluate inter-feature relationships among those with high reproducibility. To mitigate multicollinearity, feature pairs with a correlation coefficient exceeding 0.9 were identified, and a greedy recursive elimination approach was adopted to iteratively remove the most redundant feature from each highly correlated pair. This process preserved feature diversity while minimizing redundancy. Ultimately, 23 features were selected for further analysis.A least absolute shrinkage and selection operator (LASSO) regression model was then implemented on the training cohort to construct a predictive signature. LASSO incorporates L1 regularization, which drives the coefficients of less relevant features toward zero under the constraint of a tuning parameter λ. The optimal value of λ was determined via 10-fold cross-validation using the minimum error criterion. Features with non-zero coefficients were retained and linearly combined, weighted by their respective coefficients, to compute a radiomics score for each patient. All LASSO modeling procedures were conducted using the scikit-learn package in Python. Following feature selection via LASSO, the retained features were used to train multiple machine learning classifiers—including logistic regression (LR), support vector machine (SVM), random forest (RF), and extreme gradient boosting (XGBoost)—for predictive model development. The final radiomic signature was derived through 5-fold cross-validation.To integrate the radiomic signature with clinical risk factors, a radiomics nomogram was developed on the validation cohort using multivariable logistic regression. This combined tool visually represents the contribution of both imaging and clinical variables. Calibration curves were plotted to evaluate the agreement between nomogram-predicted probabilities and actual observed outcomes. To handle the high-dimensional nature of deep learning-derived features (originally 2048 dimensions), we employed principal component analysis (PCA) for dimensionality reduction. By projecting the feature space onto 32 principal components, we improved model generalizability and reduced the risk of overfitting. In this work, we fused handcrafted radiomic features with deep learning-based features to leverage complementary information and enhance predictive performance. An early fusion strategy was adopted to merge feature subsets from both handcrafted and deep learning sources into a unified representation. Feature selection was then performed on this combined set following a procedure consistent with that used for handcrafted features alone. 6.The building of the clinical model and Radiomics-Clinical model The development process of clinical biomarkers shares a striking resemblance with that of radioactive markers. Our model incorporates a comprehensive array of risk factors, encompassing both clinical features—such as age, tPSA, and fPSA—and imaging parameters, including LRD, ULD, AD, PV, PS, PD, ET, FT, and associated contrast-enhanced ultrasound metrics. Initially, relevant features for model construction were identified using foundational statistical analyses (p-value < 0.05). Subsequently, we utilized an identical suite of machine learning models for the construction of the radiomic model. In order to maintain a level playing field for comparisons, we rigorously applied a 5-fold cross-validation strategy and evaluated our model on an independent test cohort. A radiomic-clinical integrated model was developed by combining the radiomic signature with clinical predictors. The diagnostic performance of this combined model was assessed in the test cohort. ROC curves were generated to evaluate its predictive accuracy. Calibration curves were plotted to visualize the agreement between predicted probabilities and actual outcomes, and the Hosmer-Lemeshow test was applied to statistically assess model fit. Decision curve analysis (DCA) was performed to quantify the clinical utility of the model across different threshold probabilities. 7.Statistical analysis All statistical analyses were conducted using SPSS (version 23.0; IBM Corp., Armonk, NY, USA) and Python (version 3.7). Continuous variables were presented as mean ± standard deviation (SD), while categorical variables were summarized as frequency and percentage. Group comparisons for continuous variables were performed using the Student’s t-test or the Mann–Whitney U test, as appropriate, and the chi-square test was used for categorical variables. A 95% confidence interval (CI) was estimated through bootstrapping with 2000 iterations. Differences in sensitivity and specificity between the deep learning-based radiomics (DLR) model and radiologists were assessed using McNemar’s test. A two-sided p-value less than 0.05 was considered statistically significant. Results 1.Patient characteristics A total of 338 participants were enrolled based on the predefined inclusion and exclusion criteria, which comprised 188 patients diagnosed with prostate cancer and 150 controls without the disease. These participants were then randomly allocated into a training cohort containing 270 individuals and a validation cohort consisting of 68 subjects.In the training group, there were statistically significant differences between the benign group and the malignant group in age, PV, tPSA, fPSA, LRD, PD, RI and CEUS parameter AREA (P 0.05). Table 1 shows the clinical characteristics of patients in the training cohort and testing cohort. Table 1 Baseline Patient Characteristics in Training cohort and Test cohort. Training cohort(n = 270) Testing cohort(n = 68) Validation cohort(n = 108) Benign group (n = 120) Malignant group (n = 150) P Benign group(n = 30) Malignant group(n = 38) P Benign group(n = 46) Malignant group(n = 62) P tPSA(ng/m) 15.34 ± 13.92 71.88 ± 260.42 < 0.001 19.71 ± 19.16 69.21 ± 71.54 < 0.001 19.27 ± 19.40 48.10 ± 36.68 < 0.001 fPSA(ng/ml) 2.31 ± 4.76 14.24 ± 25.50 < 0.001 3.22 ± 3.57 16.74 ± 18.36 < 0.001 2.89 ± 3.50 12.68 ± 16.70 < 0.001 Age(years) 65.89 ± 6.34 71.17 ± 6.12 < 0.001 68.67 ± 7.47 70.87 ± 9.34 0.418 65.49 ± 7.01 70.08 ± 6.90 < 0.001 LRD(mm) 53.48 ± 6.25 51.73 ± 6.32 0.009 55.27 ± 5.83 54.29 ± 8.17 0.347 54.13 ± 5.72 52.30 ± 7.50 0.034 ULD(mm) 42.51 ± 8.84 41.32 ± 9.28 0.088 46.20 ± 8.43 44.00 ± 11.85 0.223 42.91 ± 7.58 41.30 ± 10.53 0.103 AD(mm) 44.66 ± 9.50 44.13 ± 8.49 0.659 46.67 ± 11.91 47.63 ± 11.73 0.877 45.94 ± 9.55 44.02 ± 8.79 0.281 PV(cm3) 56.34 ± 27.89 52.36 ± 27.64 0.179 65.60 ± 28.98 66.02 ± 42.36 0.429 58.62 ± 25.09 54.26 ± 35.82 0.104 PS(cm/s) 26.38 ± 13.09 27.57 ± 12.67 0.288 34.90 ± 17.84 33.05 ± 19.80 0.268 25.13 ± 12.95 28.21 ± 14.94 0.264 PD(cm/s) 7.62 ± 3.01 7.16 ± 3.69 0.037 9.50 ± 5.15 8.34 ± 4.98 0.299 8.00 ± 3.93 7.08 ± 3.89 0.209 RI 0.69 ± 0.09 0.72 ± 0.10 0.003 0.72 ± 0.09 0.72 ± 0.10 0.990 0.67 ± 0.08 0.72 ± 0.10 0.006 ET(s) 16.47 ± 4.07 17.28 ± 4.20 0.092 17.50 ± 4.30 16.63 ± 3.86 0.921 17.04 ± 4.49 16.98 ± 4.28 0.511 FT(s) 29.99 ± 7.87 30.42 ± 7.38 0.430 29.57 ± 7.13 29.87 ± 8.18 0.946 30.83 ± 7.88 31.20 ± 7.82 0.587 A 24.26 ± 3.65 24.76 ± 4.94 0.819 23.30 ± 2.98 24.05 ± 3.00 0.478 24.01 ± 3.49 24.93 ± 3.78 0.197 B -64.49 ± 2.84 -64.75 ± 3.24 0.441 -63.81 ± 2.28 -64.29 ± 3.35 0.926 -64.06 ± 3.01 -64.93 ± 3.11 0.086 k 0.29 ± 0.10 0.27 ± 0.10 0.176 0.28 ± 0.10 0.28 ± 0.11 1.000 0.29 ± 0.10 0.26 ± 0.09 0.106 MSE 1.16 ± 0.81 1.12 ± 0.66 0.890 1.05 ± 0.48 1.23 ± 0.71 0.634 0.99 ± 0.45 1.19 ± 0.60 0.101 TtoPK(s) 11.35 ± 4.43 10.75 ± 3.16 0.420 11.99 ± 4.04 10.32 ± 2.99 0.157 11.26 ± 4.37 10.66 ± 2.88 0.838 AREA 199.95 ± 80.92 175.24 ± 61.65 0.008 211.49 ± 100.68 163.12 ± 48.64 0.006 194.92 ± 70.64 171.95 ± 49.98 0.045 Grad 2.15 ± 0.68 2.15 ± 0.60 0.974 2.05 ± 0.78 2.24 ± 0.66 0.248 2.19 ± 0.73 2.18 ± 0.57 0.951 ATM(s) 15.65 ± 4.29 16.45 ± 3.78 0.064 16.92 ± 4.78 15.98 ± 3.19 0.595 15.95 ± 4.77 16.03 ± 3.34 0.490 Data are mean ± standard deviation (range) at the time of examination. PSA prostate-specific antigen ,tPSA total PSA, fPSA free PSA, LRD prostate left and right diameter, ULD prostate upper and lower diameter, AD prostate anteroposterior diameter, PV prostate volume ,PS/PD peak systolic and diastolic blood flow, ET/FT contrast agent entry time and full filling time, A/B/k fitting curve equation coefficient, MSE fitting curve equation error, TtoPK time to peak, AREA area under the curve, Grad peak intensity average gradient, ATM arrival time. SD standard deviation. 2.Establishment and performance of the radiomics model The original 3306 features incorporating both radiological and deep learning aspects were meticulously pruned to a concise set of 34 highly relevant features. This selection process yielded a refined dataset consisting of 29 radiological characteristics and 5 attributes derived from deep learning, which collectively informed the development of a novel multi-parameter Rad-score. The process of screening radiomics features and the detail of the features shows in Fig. 3 . Radiomics machine learning models were developed using all selected features, with the optimal model determined through comparison with LR, SVM, KNN, and MLP classifiers. Notably, the SVM algorithm yielded the highest AUC value within the training cohort, reaching 0.936 (95% confidence interval: 0.905–0.966), with corresponding sensitivities and specificities of 0.787 and 0.950, respectively. In the validation cohort, the SVM model demonstrated an AUC of 0.823 (95% CI: 0.721–0.925), with sensitivities and specificities of 0.684 and 0.833, respectively (refer to Table 2 and Fig. 4 for detailed statistics). Table 2 Differences between various machine learning models. Model name Accuracy AUC 95% CI Sensitivity Specificity PPV NPV Precision Recall F1 Threshold Task LR 0.793 0.868 0.825–0.910 0.847 0.725 0.794 0.791 0.794 0.847 0.819 0.454 Training 0.872 0.921 0.844–0.998 0.929 0.789 0.867 0.882 0.867 0.929 0.897 0.551 Testing 0.735 0.765 0.648–0.882 0.868 0.567 0.717 0.773 0.717 0.868 0.786 0.400 Validation SVM 0.859 0.936 0.905–0.966 0.787 0.95 0.952 0.781 0.952 0.787 0.861 0.669 Training 0.745 0.803 0.678–0.928 0.821 0.632 0.767 0.706 0.767 0.821 0.793 0.5 Testing 0.75 0.823 0.721–0.925 0.684 0.833 0.839 0.676 0.839 0.684 0.754 0.628 Validation KNN 0.681 0.822 0.775–0.869 0.500 0.908 0.872 0.592 0.872 0.500 0.636 0.600 Training 0.574 0.742 0.602–0.883 0.321 0.947 0.9 0.486 0.9 0.321 0.474 0.6 Testing 0.529 0.723 0.603–0.844 0.184 0.967 0.875 0.483 0.875 0.184 0.304 0.800 Validation MLP 0.833 0.902 0.866–0.938 0.853 0.808 0.848 0.815 0.848 0.853 0.85 0.554 Training 0.766 0.814 0.689–0.939 0.786 0.737 0.815 0.7 0.815 0.786 0.8 0.463 Testing 0.779 0.844 0.749–0.939 0.763 0.8 0.829 0.727 0.829 0.763 0.795 0.560 Validation AUC, area under the curve; PPV, positive predictive value; NPV, negative predictive value; F1 = 2 x (precision x recall)/(precision + recall); LR, logistic regression; SVM, support vector machine; KNN, K‑nearest neighbors; MLP, multi‑layer perceptron. AUC, area under the curve; PPV, positive predictive value; NPV, negative predictive value; F1 = 2 x (precision x recall)/(precision + recall); 3.Establishment and performance of the clinical model and RadiomicsClinical model The clinical model was constructed based on significant features, as indicated by P-values (< 0.05) from the training cohort, including clinical parameters tPSA, fPSA, Age, and imaging characteristics LRD, PD, RI, and ARE, which collectively define the clinical signatures. The refined clinical model yielded an AUC of 0.746 (95%CI 0.652–0.840) with a sensitivity and specificity of 0.693 and 0.883, respectively, in the training dataset. Furthermore, the model demonstrated an AUC of 0.809 (95%CI 0.708–0.910) in the validation cohort, with corresponding sensitivities and specificities of 0.605 and 0.900 (refer to Table 3 and Fig. 5 for detailed results). Table 3 Predictive Performance of Three Models in the Training Cohort and Validation Cohort. Signature Training Cohort (n = 270) AUC (95% CI) Sensitivity* Specificity* Testing cohort(n = 68) AUC (95% CI) Sensitivity* Specificity* Validation Cohort (n = 108) AUC (95% CI) Sensitivity* Specificity* Clinic Signature 0.856(0.812–0.901) 0.693 0.883 0.917(0.826–1.000) 0.893 0.895 0.809 (0.708–0.910) 0.605 0.900 Rad Signature 0.936(0.905–0.966) 0.787 0.951 0.803(0.678–0.928) 0.821 0.632 0.823 (0.721–0.925) 0.684 0.833 Radiomics-clinical 0.956(0.931–0.981) 0.893 0.925 0.951(0.881–1.000) 0.893 0.947 0.889 (0.813–0.966) 0.737 0.933 *Balanced sensitivity and specificity at the cutoff yielding the largest Youden index value. The radiomics-clinical model yielded an AUC of 0.956 (95% CI 0.931–0.981) in the training cohort, with corresponding sensitivities and specificities of 0.893 and 0.925. In the validation cohort, the model demonstrated an AUC of 0.889 (95% CI 0.813–0.966), with sensitivities and specificities of 0.737 and 0.933, respectively. (Refer to Table 3 and Fig. 5 for details.) Subsequently, a nomogram was developed based on the radiomics-clinical model. (Please see Fig. 7 for the nomogram.) The Decision Curve Analysis (DCA) revealed that the model confers substantial benefits in the majority of scenarios. (Fig. 6 provides further insights into these benefits.) Delong's examination of the comparison model elucidated a significant disparity in the AUC between the radiomic-clinical model and the sole clinical model within the testing cohort (p < 0.001). Conversely, no significant differentiation was observed between the radiomic-clinical model and the radiomic model (p = 0.123). Discussion TRUS is the most commonly used method in clinical diagnosis of prostate cancer, which has the advantages of high imaging resolution, non-invasive, strong real-time imaging ability, high sensitivity and strong specificity. Recently, there have been study [ 21 ] that simply establish radiomics models of transrectal ultrasound to predict prostate cancer, and good results have been achieved. Compared with conventional ultrasound, CEUS can perform real-time dynamic imaging, observe actual blood flow, and evaluate blood perfusion, thus realizing clear visualization of tissue and lesion microcirculation perfusion, and significantly improving the diagnostic efficiency of prostate cancer [ 10 , 22 ] . If the data extracted from static B-model images can be combined with CEUS parameters, it is believed that the diagnosis of prostate cancer will be further improved. Mining deep information from traditional imaging and improving diagnosis and treatment methods are the main directions to conquer prostate cancer. As a non-invasive new technology that can extract features with high throughput, imaging omics has emerged at the moment, showing great advantages in tumor phenotyping, treatment decision-making and prognosis analysis. At present, the imaging omics of prostate cancer mainly focuses on the diagnosis and differential diagnosis of CT, MRI and PET images, tumor staging, efficacy evaluation and prognosis prediction. Yong Wang et al. [ 23 ] used clinical data CEUS parameters MRI images to establish a model for predicting biochemical recurrence of prostate cancer (PCa) after surgery and to verify its clinical validity, which can help improve its prognosis. Chunyu Li et al. [ 24 ] established an imaging model based on multi-parameter MRI images to distinguish prostate cancer from prostate hyperplasia, thereby helping clinicians to make better clinical treatment decisions and reduce unnecessary prostate biopsies. In terms of ultrasound imaging of prostate cancer, Wei Ou MD et al. [ 25 ] developed and validated the imaging score to predict prostate cancer before biopsy in order to reduce unnecessary prostate cancer biopsies. However, they only conducted a single imaging study, without applying the blood perfusion characteristics and deep learning methods of prostate cancer. In a prior investigation into the diagnosis of prostate cancer, Ya Sun et al. [ 19 ] developed a machine learning model that integrated pyradiomic radiomic features with B-mode transrectal ultrasound and contrast-enhanced ultrasound. This innovative model aimed to enhance the detection of peripheral prostate cancer and supply empirical evaluation metrics. In our present study, our target mapping based on CEUS demonstrates a higher degree of accuracy when compared to the direct mapping on the poorly defined B-mode image. And we extended this approach by incorporating a deep learning methodology, performing feature fusion prior to model development, and integrating clinical risk factors to construct a hybrid model. Our findings corroborate that the fusion of deep learning methodologies can significantly enhance the diagnostic accuracy and clinical utility of differentiating PCa. In the present investigation, we developed three predictive models—each grounded in radiomic model, clinical model, and a hybrid radiomic-clinical model, respectively. Notably, the integrated radiomic-clinical model exhibited superior predictive performance, demonstrating a commendable accuracy in prognostication for prostate cancer. Additionally, a nomogram was constructed to assist urologists and sonographers in clinical decision-making. Radiomics-based B-medel TRUS image analysis has yielded commendable predictive outcomes, with the resulting model markedly outperforming both the clinical model and the standalone radiomics model. This suggests a promising role for radiomics in the prognostication of prostate cancer, and underscores the superiority of integrated models when compared to their monolithic counterparts. Certainly, several constraints are inherent to our investigation. Initially, this study collected patients from only two centers, and enrollment of patients from more centers is needed to improve the generalizability of the model in clinical application. Secondly, the extant evidence necessitates corroboration via extensive, prospective, randomized controlled trials. Thirdly, manual lesion segmentation is time-consuming and easy to be affected by the experience of the image reader. In the future, it is necessary to explore automatic or semi-automatic methods to describe the lesions more accurately. Moreover, our inquiry is exclusively predicated on machine learning and deep learning methodologies, excluding more cutting-edge techniques such as deep transfer learning. Subsequent to this iteration of research, we intend to incorporate deep transfer learning approaches into our methodology. Conclusion The radiomics-clinical machine learning model based on B-model combined with traditional radiomics and deep learning can be used for preoperative prediction of prostate cancer patients, providing a new and promising method for urological surgeons to select suitable prostate cancer patients for treatment. Abbreviations TRUS: Transrectal ultrasound; CEUS: Contrast-enhanced ultrasound; PCa: Prostate cancer; PSA: Prostate-specific antigen; DRE: Digital rectal examination; LRD: left and right diameter; ULD: upper and lower diameter; AD: anteroposterior diameter; PV: prostate volume; tPSA total PSA; fPSA: free PSA; PS/PD: peak systolic and diastolic blood flow; ET/FT: contrast agent entry time and full filling time; mp‑MRI: Multiparametric magnetic resonance imaging; T2WI: 2-weighted fast spin-echo imaging; DWI: diffusion weighted MR Imaging; DCE-MRI: dynamic contrast-enhanced MRI; DICOM: Digital Imaging and Communications in Medicine; Grad: Peak intensity average gradient; ATM: arrival time; TtoP: time to peak; Area: Area under the curve; ICC: intraclass correlation coefficient; HCR: hand crafted features; GLCM: gray-level co-occurrence matrix; GLRLM: gray-level run length matrix; GLSZM: gray-level size zone matrix; NGTDM: neighborhood gray-level difference matrix; LASSO: least absolute shrinkage and selection operator; LR: logistic regression; SVM: support vector machine; RF: random forest; XGBoost: extreme gradient boosting; PCA: principal component analysis; DCA: Decision curve analysis Declarations Ethics approval and consent to participate This study complies with the 1964 Helsinki Declaration and its subsequent amendments or equivalent ethical standards. It has been approved by the Ethics Committee of Inner Mongolia Autonomous Region People's Hospital (Approval Number: SC-07/01KT2024022). Consent for publication Not applicable. Availability of data and materials All data generated or analyzed during this study are included in this published article. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Funding This study was supported by the Science and Technology Program of the Joint Fund of Scientific Research for the Public Hospitals of Inner Mongolia Academy of Medical Sciences(2023GLLH0020). Authors’ Contributions SB：conceived and designed the analysis, collected the data, performed the analysis, wrote the paper. QL: conceived and designed the analysis, collected the data, performed the analysis, wrote the paper. HH: conceived and designed the analysis. RS: collected the data. YS: collected the data. LY: collected the data, contributed data or analysis tools. DL: contributed data or analysis tools. XR: Contributed data or analysis tools. WL: performed the analysis. HL: performed the analysis. Acknowledgment Certain experiments were conducted on the Onekey AI platform, for which we are grateful for the support and assistance provided by Onekey AI and its development team in this scientific research endeavor. References Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2024;74(3):229–63. Siegel RL, Giaquinto AN, Jemal A. Cancer statistics, 2024. CA Cancer J Clin. 2024;74(1):12–49. Falagario UG, Sanguedolce F, Dovey Z, et al. Prostate cancer biomarkers: a practical review based on different clinical scenarios. Crit Rev Clin Lab Sci. 2022;59(5):297–308. Islam MA, Alam SM, Reza AM. Urosepsis and Bacteriuria in Patients Undergoing TRUS-Guided Prostate Biopsy. Mymensingh Med J. 2023;32(2):330–7. Morelli M, Sampogna G, Molteni S, et al. The impact of prostate biopsy on erectile and ejaculatory function: A prospective study. Arch Ital Urol Androl. 2022;94(4):420–3. Jacewicz M, Rud E, Lauritzen P, Baco E. Non-infectious adverse events of transperineal prostate biopsies performed under local anaesthesia. BJU Int. 2024. Borofsky S, George AK, Gaur S, et al. What Are We Missing? False-Negative Cancers at Multiparametric MR Imaging of the Prostate. Radiology. 2018;286(1):186–95. Zhang J, Xu L, Zhang G, et al. Effects of dynamic contrast enhancement on transition zone prostate cancer in Prostate Imaging Reporting and Data System Version 2.1. Radiol Oncol. 2023;57(1):42–50. Bai X, Jiang Y, Zhang X, et al. The Value of Prostate-Specific Antigen-Related Indexes and Imaging Screening in the Diagnosis of Prostate Cancer. Cancer Manag Res. 2020;12:6821–6. Salib A, Halpern E, Eisenbrey J, et al. The evolving role of contrast-enhanced ultrasound in urology: a review. World J Urol. 2023;41(3):673–8. Yang G, Ruan L. Imaging findings of prostate tuberculosis by transrectal contrast-enhanced ultrasound and comparison with 2D ultrasound and pathology. Br J Radiol. 2022;95(1129):20210713. Jung EM, Wertheimer T, Putz FJ, et al. Contrast enhanced ultrasound (CEUS) with parametric imaging and time intensity curve analysis (TIC) for evaluation of the success of prostate arterial embolization (PAE) in cases of prostate hyperplasia. Clin Hemorheol Microcirc. 2020;76(2):143–53. Wink M, Frauscher F, Cosgrove D, et al. Contrast-enhanced ultrasound and prostate cancer; a multicentre European research coordination project. Eur Urol. 2008;54(5):982–92. Liu Y, Lu D, Xu G, et al. Diagnostic accuracy of qualitative and quantitative magnetic resonance imaging-guided contrast-enhanced ultrasound (MRI-guided CEUS) for the detection of prostate cancer: a prospective and multicenter study. Radiol Med. 2024;129(4):585–97. Lin Y, Belue MJ, Yilmaz EC, et al. Deep Learning-Based T2-Weighted MR Image Quality Assessment and Its Impact on Prostate Cancer Detection Rates. J Magn Reson Imaging. 2024;59(6):2215–23. Wildeboer RR, Mannaerts CK, van Sloun R, et al. Automated multiparametric localization of prostate cancer based on B-mode, shear-wave elastography, and contrast-enhanced ultrasound radiomics. Eur Radiol. 2020;30(2):806–15. Liang L, Zhi X, Sun Y, et al. A Nomogram Based on a Multiparametric Ultrasound Radiomics Model for Discrimination Between Malignant and Benign Prostate Lesions. Front Oncol. 2021;11:610785. Lorusso V, Kabre B, Pignot G, et al. External validation of the computerized analysis of TRUS of the prostate with the ANNA/C-TRUS system: a potential role of artificial intelligence for improving prostate cancer detection. World J Urol. 2023;41(3):619–25. Sun Y, Fang J, Shi Y, et al. Machine learning based on radiomics features combing B-mode transrectal ultrasound and contrast-enhanced ultrasound to improve peripheral zone prostate cancer detection. Abdom Radiol (NY). 2024;49(1):141–50. Turkbey B, Rosenkrantz AB, Haider MA, et al. Eur Urol. 2019;76(3):340–51. Prostate Imaging Reporting and Data System Version 2.1: 2019 Update of Prostate Imaging Reporting and Data System Version 2. Huang Y, Qian H, Zheng Y, Song H, Liu X. A radiomics model based on transrectal ultrasound for predicting prostate cancer. Med Ultrason. 2024;26(2):138–46. Maxeiner A, Fischer T, Schwabe J, et al. Contrast-Enhanced Ultrasound (CEUS) and Quantitative Perfusion Analysis in Patients with Suspicion for Prostate Cancer. Ultraschall Med. 2019;40(3):340–8. Wang Y, Feng G, Wang J et al. Contrast-Enhanced Ultrasound-Magnetic Resonance Imaging Radiomics Based Model for Predicting the Biochemical Recurrence of Prostate Cancer: A Feasibility Study. Comput Math Methods Med. 2022. 2022: 8090529. Li C, Deng M, Zhong X, et al. Multi-view radiomics and deep learning modeling for prostate cancer detection based on multi-parametric MRI. Front Oncol. 2023;13:1198899. Ou W, Lei J, Li M, et al. Ultrasound-based radiomics score for pre-biopsy prediction of prostate cancer to reduce unnecessary biopsies. Prostate. 2023;83(1):109–18. Additional Declarations No competing interests reported. Supplementary Files supplement.docx Cite Share Download PDF Status: Under Review Version 1 posted Reviewers invited by journal 05 Dec, 2025 Editor invited by journal 10 Nov, 2025 Editor assigned by journal 29 Sep, 2025 Submission checks completed at journal 29 Sep, 2025 First submitted to journal 29 Sep, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7653994","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":555885712,"identity":"3bcd7390-e1ed-4a10-98bd-aa0ca6238148","order_by":0,"name":"Senlin Bao","email":"","orcid":"","institution":"Ultrasound Department of Inner Mongolia Autonomous Region People's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Senlin","middleName":"","lastName":"Bao","suffix":""},{"id":555885713,"identity":"ecb8e4f9-cead-41ae-aa81-c71e23a5b2a3","order_by":1,"name":"Qian Liu","email":"","orcid":"","institution":"Ultrasound Department of Inner Mongolia Autonomous Region People's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Qian","middleName":"","lastName":"Liu","suffix":""},{"id":555885714,"identity":"03733b7f-934b-4a0c-a886-6dd4bcb27693","order_by":2,"name":"Ran Sun","email":"","orcid":"","institution":"Ultrasound Department of Inner Mongolia Autonomous Region People's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Ran","middleName":"","lastName":"Sun","suffix":""},{"id":555885715,"identity":"cd0ef750-74eb-43ea-8ff9-16e6e6ade387","order_by":3,"name":"Yue Sun","email":"","orcid":"","institution":"Ultrasound Department of Inner Mongolia Autonomous Region People's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yue","middleName":"","lastName":"Sun","suffix":""},{"id":555885716,"identity":"8aa816f3-cc3f-4ef1-bc74-a5084f3d789c","order_by":4,"name":"Li Yang","email":"","orcid":"","institution":"Tongliao People ＇ s Hospital","correspondingAuthor":false,"prefix":"","firstName":"Li","middleName":"","lastName":"Yang","suffix":""},{"id":555885717,"identity":"548b5e71-c906-4431-9cca-94a686bd7789","order_by":5,"name":"Danyan Liang","email":"","orcid":"","institution":"Ultrasound Department of Inner Mongolia Autonomous Region People's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Danyan","middleName":"","lastName":"Liang","suffix":""},{"id":555885718,"identity":"29c40991-c01c-48ff-ad77-b42a1cff5f9b","order_by":6,"name":"Xiujuan Ren","email":"","orcid":"","institution":"Ultrasound Department of Inner Mongolia Autonomous Region People's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Xiujuan","middleName":"","lastName":"Ren","suffix":""},{"id":555885719,"identity":"4127fc56-30bd-451a-a0f2-def27cc6a0ff","order_by":7,"name":"Wen Li","email":"","orcid":"","institution":"Jinan Thirdly People's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Wen","middleName":"","lastName":"Li","suffix":""},{"id":555885720,"identity":"aa6dd56b-821d-4923-b29e-c7e55eaacc8e","order_by":8,"name":"Heyang Liu","email":"","orcid":"","institution":"Ultrasound Department of Inner Mongolia Autonomous Region People's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Heyang","middleName":"","lastName":"Liu","suffix":""},{"id":555885721,"identity":"5e9ed93c-88c7-47e5-a6e9-24929890a893","order_by":9,"name":"Hua Hong","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA60lEQVRIiWNgGAWjYBACfvbmgw8/GNgkANmMDxIqaghrkew5lmwsUZEG0sJs8ODMMcJaDGb4mEnwnDkE0sIm+bCFmQgtEkAk2XYgj1+6/VpFYgMbA397dwJeLebSDQkGhW13iiXnnCm7kbhDhkHizNkNeLVYzjlwIEGy7Vnihhs5aTcSz7ABLc3Fr8XgRmLDAd62w2AtBYltzMRoSWZs4DkD0pJ+jIEoLcBAZmYGBnLizBk5zBIJZ47xEPQLP3v/95/AqEzsl0h/+PFHRY0cf3svfi1IgMcATBKrHATYH5CiehSMglEwCkYQAADLUVPE6iHndgAAAABJRU5ErkJggg==","orcid":"","institution":"Ultrasound Department of Inner Mongolia Autonomous Region People's Hospital","correspondingAuthor":true,"prefix":"","firstName":"Hua","middleName":"","lastName":"Hong","suffix":""}],"badges":[],"createdAt":"2025-09-19 03:23:21","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7653994/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7653994/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":97722092,"identity":"d08abe9e-f89d-4428-8b3d-3dcc2538dfc2","added_by":"auto","created_at":"2025-12-08 15:39:07","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7603928,"visible":true,"origin":"","legend":"","description":"","filename":"BMC.docx","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/63c4d96bca59f087171aaf39.docx"},{"id":97895253,"identity":"96ab1665-b500-4a96-afc1-19e81c9e2b52","added_by":"auto","created_at":"2025-12-10 15:33:53","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":10873,"visible":true,"origin":"","legend":"","description":"","filename":"8be63934087f4ea2ab7759d7aef0f223.json","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/8590a1ea9627f52a0d5bb733.json"},{"id":97722098,"identity":"681669c9-ffce-4527-8be0-f45b2bc3d924","added_by":"auto","created_at":"2025-12-08 15:39:08","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":434266,"visible":true,"origin":"","legend":"","description":"","filename":"supplement.docx","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/8dd8dd39f7658412b134adf0.docx"},{"id":97722103,"identity":"524dd6ae-b87f-4c6f-84f6-9b4f580fc390","added_by":"auto","created_at":"2025-12-08 15:39:08","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":125582,"visible":true,"origin":"","legend":"","description":"","filename":"8be63934087f4ea2ab7759d7aef0f2231enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/68c97670450327166f1332ae.xml"},{"id":97722091,"identity":"11c0dbbb-bc40-4d97-91c4-33bb596c28d2","added_by":"auto","created_at":"2025-12-08 15:39:07","extension":"jpeg","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":93467,"visible":true,"origin":"","legend":"","description":"","filename":"groupimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/f665df1f028925f4ba9fccae.jpeg"},{"id":97722075,"identity":"38f80707-a043-44bc-8f8f-b5bb98d814f2","added_by":"auto","created_at":"2025-12-08 15:39:06","extension":"jpeg","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":22679,"visible":true,"origin":"","legend":"","description":"","filename":"groupimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/4f92cc6b0ebea72f4b13d84b.jpeg"},{"id":97722113,"identity":"6dd77c31-6263-471c-bc52-a6cac5151c17","added_by":"auto","created_at":"2025-12-08 15:39:09","extension":"jpeg","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":26978,"visible":true,"origin":"","legend":"","description":"","filename":"groupimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/77c3bb9aec1dd1454161c8b5.jpeg"},{"id":97722095,"identity":"8c9104f0-f7ab-465a-b01f-81ff6f99e03c","added_by":"auto","created_at":"2025-12-08 15:39:08","extension":"jpeg","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15406,"visible":true,"origin":"","legend":"","description":"","filename":"groupimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/3909a12814fa32e59b74cfb9.jpeg"},{"id":97722078,"identity":"6a7a0f40-7934-4a37-81cd-b70e9f946a0b","added_by":"auto","created_at":"2025-12-08 15:39:06","extension":"jpeg","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":30754,"visible":true,"origin":"","legend":"","description":"","filename":"groupimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/ac6cc2523cd716dfb3de2d48.jpeg"},{"id":97722085,"identity":"961acaaa-6708-4bd2-aa01-1ac6e2d0e8c8","added_by":"auto","created_at":"2025-12-08 15:39:07","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":30170,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinegroupimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/390dfec083df68bfc153f51f.png"},{"id":97722104,"identity":"0e4c3856-9cae-4387-8df0-306684240ccc","added_by":"auto","created_at":"2025-12-08 15:39:08","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9650,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinegroupimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/a83f103ae06d34173911b612.png"},{"id":97722079,"identity":"5cae83f8-068a-4ceb-91ca-f021e66e3333","added_by":"auto","created_at":"2025-12-08 15:39:06","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8865,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinegroupimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/ec4f598e14f7fd833a83f5ff.png"},{"id":97895199,"identity":"8809cb65-d4b3-4910-96ce-c2cda3d033d8","added_by":"auto","created_at":"2025-12-10 15:33:45","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14288,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinegroupimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/804c962c7c5600df8e1a042c.png"},{"id":97722089,"identity":"0c2aa6ef-ca7e-4ba9-9e3f-c7a60107c192","added_by":"auto","created_at":"2025-12-08 15:39:07","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9418,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinegroupimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/93fc5c1dd5f904cfa5170a16.png"},{"id":97895342,"identity":"52c7b6e6-ba05-49e2-a3f2-71bd40aebdf0","added_by":"auto","created_at":"2025-12-10 15:34:01","extension":"xml","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":122767,"visible":true,"origin":"","legend":"","description":"","filename":"8be63934087f4ea2ab7759d7aef0f2231structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/df5f8155e86ec203cbe21daa.xml"},{"id":97895509,"identity":"6ad3251b-f6e7-4d45-8c4d-82411ffe8389","added_by":"auto","created_at":"2025-12-10 15:34:22","extension":"html","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":132675,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/4dcb9864d0c50b8e37b26822.html"},{"id":97894904,"identity":"02c48cdb-befd-4f3e-9f13-0e311526e720","added_by":"auto","created_at":"2025-12-10 15:33:11","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":622120,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u0026nbsp;\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/955b74614e578b849574350c.png"},{"id":97722083,"identity":"bf80ca05-04dc-4028-add6-cba801ff8893","added_by":"auto","created_at":"2025-12-08 15:39:07","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":2414703,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u0026nbsp;\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/2e96b1792ef7be617390e9a0.png"},{"id":97722088,"identity":"5cfe1c31-cdeb-478f-92a0-14fce65d9a7e","added_by":"auto","created_at":"2025-12-08 15:39:07","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":930490,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u0026nbsp;\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/8e70b59b12fdf1385b4e9b41.png"},{"id":97722069,"identity":"a8501609-038b-430d-96fe-e7575c6f05e9","added_by":"auto","created_at":"2025-12-08 15:39:04","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":152937,"visible":true,"origin":"","legend":"\u003cp\u003eTraining and testing group ROC curves for various classifiers. A.LR B.SVM C.KNN D.MLP\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/61f9d57b1c54a0cc5259882d.png"},{"id":97722093,"identity":"ad366e97-4ab8-4789-b99a-63fac0e5fdc8","added_by":"auto","created_at":"2025-12-08 15:39:07","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":178464,"visible":true,"origin":"","legend":"\u003cp\u003eROC curves of the radiomics model, clinical model, and radiomics-clinical nomogram in the training cohort (A) and testing cohort (B). ROC = receiver operating characteristic.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/2d15b8b6c31deb62274658f3.png"},{"id":97722096,"identity":"3ecc2cf4-3c49-4a7d-aa8d-0542e627c904","added_by":"auto","created_at":"2025-12-08 15:39:08","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":140475,"visible":true,"origin":"","legend":"\u003cp\u003eThe DCA of radiomics-clinical model in test cohort. DCA = decision curve analysis\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/749f118193f84ad8991c77ef.png"},{"id":97722105,"identity":"d9d02b42-45d9-42bb-ad2d-3060f0cade84","added_by":"auto","created_at":"2025-12-08 15:39:08","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":93884,"visible":true,"origin":"","legend":"\u003cp\u003eRadiomics-clinical nomogrampredicts benign and malignant prostate nodules. Clinic_Sig = Clinical Signature; Rad_Sig = Radiomics Signature.\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/03c539599b450a833452fff3.png"},{"id":97902606,"identity":"a3fdffbe-9771-4280-9dad-b3d138d00b00","added_by":"auto","created_at":"2025-12-10 15:53:04","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":6155390,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/fad9f49c-4f0c-45ab-9943-28fde55205ca.pdf"},{"id":97722110,"identity":"875bbc8e-704a-4972-9c0f-3eb085202264","added_by":"auto","created_at":"2025-12-08 15:39:09","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":434266,"visible":true,"origin":"","legend":"","description":"","filename":"supplement.docx","url":"https://assets-eu.researchsquare.com/files/rs-7653994/v1/68623afb742fe777ad5cc7d0.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Development and validation of a radiomics deep learning signature from MRI-guided transrectal ultrasound combined with ultrasound imaging parameters for prostate cancer prediction","fulltext":[{"header":"Introduction","content":"\u003cp\u003eProstate cancer (PCa) is the foremost genitourinary cancer worldwide and the second most prevalent malignancy in the male population\u003csup\u003e[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]\u003c/sup\u003e. By 2024, it is estimated that the United States will have seen a total of 299,010 newly diagnosed cases of prostate cancer, with an anticipated 35,250 deaths attributable to the disease\u003csup\u003e[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]\u003c/sup\u003e. Therefore, accurate diagnosis of PCa is crucial for enhancing survival rates and prognosis in patients. Diagnosing prostate cancer is challenging due to the similarity of symptoms with other prostatitis-related conditions.\u003c/p\u003e\u003cp\u003eConventional approaches for the detection of prostate cancer principally encompass serum prostate-specific antigen (PSA) testing, digital rectal examination (DRE), and transrectal ultrasound (TRUS) guided needle biopsies. However, research has indicated that these methods exhibit low sensitivity and specificity\u003csup\u003e[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]\u003c/sup\u003e, and may also lead to complications such as infection, bleeding, and erectile dysfunction\u003csup\u003e[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]\u003c/sup\u003e. Many patients undergo unnecessary biopsies due to benign prostatic hyperplasia. Furthermore, there are numerous limitations associated with ultrasound-guided prostate puncture, and the primary drawbacks of transrectal or perineal prostate system puncture biopsy include false negatives, missed diagnosis of high-risk prostate cancer, and overdiagnosis\u003csup\u003e[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]\u003c/sup\u003e. Therefore, a non-invasive and accurate diagnosis of prostate cancer holds significant importance.\u003c/p\u003e\u003cp\u003eMRI provides better anatomical resolution and higher accuracy in the diagnosis of PCa, but the system is not without its limitations with regard to the detection of angiogenesis in PCa, and may also misdiagnose or miss some csPCa with a proportion of 58%\u003csup\u003e[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]\u003c/sup\u003e. Consequently, the MRI diagnosis of PCa remains a clinical challenge that necessitates further investigation. TRUS can accurately assess the size, shape, structure, and abnormalities of the prostate, providing real-time imaging that aids in timely diagnosis. This is of significance for the early detection of prostate cancer and the evaluation of the extent of lesions. Therefore, TRUS has significant value in the diagnosis of prostate diseases, particularly in its outstanding performance in early screening and localization of prostate cancer\u003csup\u003e[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/sup\u003e. As an emerging ultrasound technology, blood flow patterns can be dynamically visualized and even quantitatively analyzed by contrast agents \u003csup\u003e[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]\u003c/sup\u003e. Even at an early stage, CEUS can detect elevated blood flow resulting from new blood vessel formation associated to PCa\u003csup\u003e[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]\u003c/sup\u003e. In comparison with mp-MRI, CEUS is characterised by certain disadvantages, including the challenge of localising suspicious lesions due to suboptimal temporal and spatial resolutions\u003csup\u003e[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]\u003c/sup\u003e. However, studies have shown that CEUS when used in conjunction with MRI can enhance the diagnostic accuracy of suspicious lesiones\u003csup\u003e[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eAdditionally, the intersection of radiomics and deep learning represents a burgeoning interdisciplinary domain that integrates medical imaging with computer science to extract extensive quantitative data from medical images, establishing connections between these features and the onset, progression, and prognosis of diseases, which has great potential for clinical diagnosis and treatment. It is acknowledged that preceding studies have progressively highlighted the utilisation of radiomics or deep learning techniques in the analysis of CT or MRI examinations\u003csup\u003e[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]\u003c/sup\u003e, but few studies have combined radiomics and deep learning methods with CEUS parameters, especially under the guidance of MRI \u003csup\u003e[\u003cspan additionalcitationids=\"CR17 CR18\" citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eTherefore, we postulated that the integration of CEUS parameters and ultrasound radiomics might result in a more precise diagnosis of prostate cancer. The objective of the present study is to employ artificial and deep learning feature extraction approaches to extract multi-dimensional features from MP-MRI-guided B-mode ultrasound images of prostate nodules for constructing a radiomics model, and to establish a novel prostate cancer prediction model combined with contrast-enhanced ultrasound parameters and clinical risk factors for clinical application.\u003c/p\u003e"},{"header":"Materials and methods","content":"\n\u003ch3\u003e1.Patients and data acquisition\u003c/h3\u003e\n\u003cp\u003e The present bicentric retrospective study was approved by the ethics committees of the relevant institutional reviewing boards at all collaborating hospitals, and the requirement for written consent was waived. A retrospective data collection was conducted on 504 patients with histopathologically-validated suspected prostate cancer at two hospitals between November 2019 and September 2023. In the end, 443 patients with prostate disease formed the study group. The training and internal testing cohorts of Hospital 1 (Inner Mongolia People's Hospital) comprised 270 and 68 patients, correspondingly. The external validation cohort of Hospital 2 (Tongliao People's Hospital) comprised 105 patients. The criteria that were utilised to determine the inclusion and exclusion of studies in this analysis are illustrated in the Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eThe following data is to be collated on every patient:Initial clinical and ultrasound image data, such as Age, prostate size [left and right diameter(LRD), upper and lower diameter (ULD), anteroposterior diameter(AD)], prostate volume(PV), total PSA (tPSA), free PSA (fPSA), peak systolic and diastolic blood flow(PS/PD), contrast agent entry time and full filling time(ET/FT), related angiographic parameters were retrieved from the medical documentation. The configuration of the study design and the pipeline are demonstrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\n\u003ch3\u003e2.Multiparametric magnetic resonance imaging (mp‑MRI) acquisition and interpretation\u003c/h3\u003e\n\u003cp\u003eMp-MRI was carried out using a GE Signa HDxt 3.0 T magnet. The sequences included T2-weighted fast spin-echo imaging (T2WI), diffusion weighted MR Imaging (DWI) and dynamic contrast-enhanced MRI (DCE-MRI), with a high b value of 1500 s/mm 2 and a slice thickness of 3.0 mm. The analysis of MR images was conducted by two radiologists, who utilised the Prostate Imaging Reporting and Data System version 2.1 (PI-RADS v2.1) for this purpose\u003csup\u003e[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]\u003c/sup\u003e. If there were discrepancies, the PI-RADS score was finally determined by another senior radiologist. Images in Digital Imaging and Communications in Medicine (DICOM) format of mpMRI were uploaded to Volumetric navigation system in Logiq GE 9 R6 (GE HealthCare, USA). The motion and direction of the probe were transmitted to the fusion system by triangulation registration between a magnetic field generator and an electromagnetic tracking sensor on an IC5-9-D end-scan probe (2.4\u0026ndash;9.6 MHz) equipped with a dedicated puncture frame. Targeted lesions were identified with reference to MRI. In patients exhibiting PI-RADS 2\u0026ndash;5 scores, the lesion that attained the highest PI-RADS rating (or possessed the most substantial diameter) was designated as the CEUS target lesion. The motion of the ultrasound image presented by the probe was synchronised with the MRI image to ensure accurate acquisition of the two-dimensional ultrasound image.\u003c/p\u003e\n\u003ch3\u003e3.Ultrasound imaging acquisition and image segmentation\u003c/h3\u003e\n\u003cp\u003eThe size and blood flow of the prostate were observed and recorded by transrectal ultrasound. CEUS was performed on the largest cross-sectional section of the targeted lesion. A total of 2.4 ml of SonoVue (Bracco, Italy) was injected into the median cubital vein via a bolus injection, followed promptly by 5 ml of standard saline. Lesion development was observed continuously on fixed sections for 180s, and dynamic Dicom images were retained. One or two cores were taken from each target, and then the fusion system was withdrawn into the sagittal plane for a standardized biopsy protocol (basal, middle, or apical glands; Left or right, 12 points), using a biopsy gun and an 18G biopsy needle, with a penetration depth of 15 mm or 20 mm. The sample was placed in the specimen bottle.\u003c/p\u003e\u003cp\u003eTIC curves were analyzed by calculating the video footage of CEUS with the imaging analysis software of the machine. Three ROIs are placed at the location where the contrast agent first enters, and the Wash-in equation [F(t)\u0026thinsp;=\u0026thinsp;A(1 - exp(-kt))\u0026thinsp;+\u0026thinsp;B] is used to fit the curve to obtain relevant contrast parameters: Peak intensity average gradient (Grad), arrival time (ATM), time to peak(TtoP), Area under the curve (Area), fitting curve equation coefficient (A, B, k), fitting curve equation error (MSE). The average of the TIC parameters obtained from three measurements is included in the clinical information.\u003c/p\u003e\u003cp\u003eThe images were imported into ITK-SNAP(version 4.0.2), an open-source software program, and the ROIs were manually segmented in B-mode ultrasound images compared with mp-MRI images. The completed outline image was saved as a mask file in nii format. After 2 months, 50 patients were randomly selected, and the region of interest was redelineated by 2 radiologists. The intraclass correlation coefficient (ICC) has been utilised as a metric to assess the consistency of tumour delineation.\u003c/p\u003e\n\u003ch3\u003e4.Radiomics feature extraction and deep learning features extraction\u003c/h3\u003e\n\u003cp\u003eIn this study, the \"One-key AI v3.2.2\" platform (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.medai.icu/\u003c/span\u003e\u003cspan address=\"http://www.medai.icu/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e, which is based on Pytorch 1.8.0) was used to extract 1593 radiomics features and 2048 deep learning features from B-mode ultrasound images. The regions of interest were then cropped with the largest areas that contained the entire tumour. The grey-scale images were processed to ensure background information was standardised and minimise noise disruption prior to imaging. Employing this methodology, the input image was reduced to a dimension of 224 \u0026times; 224 pixels through linear interpolation, whilst normalising the pixel intensity mean and variance to 0 and 1, thus ensuring consistency in data representation.\u003c/p\u003e\u003cp\u003eRadiomics Extraction: The hand crafted features(HCR) can be classified into three categories: geometric features, intensity features, and texture features. These are used to depict the two-dimensional shape characteristics of tumors, the first-order statistical distribution of intensity within tumor voxels, and intensity patterns, i.e., the first-order and higher-order spatial distributions of intensity. Here, texture features are extracted using the gray-level co-occurrence matrix (GLCM), gray-level run length matrix (GLRLM), gray-level size zone matrix (GLSZM), and neighborhood gray-level difference matrix (NGTDM) methods. A full overview of the HCR features extracted in this project is available in the Pyradiomics manual (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://pyradiomics.readthedocs.io\u003c/span\u003e\u003cspan address=\"http://pyradiomics.readthedocs.io\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe ROI with the maximum possible sagittal area was chosen for precise trimming before the DTL features were extracted. In this study, ResNet50 was used as the CNN model to extract deep learning features. ResNet50 was pre-trained on the ILSVRC-2012 database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.image-net.org/challenges/LSVRC/2012/\u003c/span\u003e\u003cspan address=\"https://www.image-net.org/challenges/LSVRC/2012/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). The slice with the largest tumour area was selected to depict every patient. Once the deep learning model had been trained, the features from the average pooled layer were taken as the deep learning features.\u003c/p\u003e\n\u003ch3\u003e5.Feature selection and model construction\u003c/h3\u003e\n\u003cp\u003eWe performed statistical screening of radiomic features using the Mann-Whitney U test. Only features with a significance level of p\u0026thinsp;\u0026lt;\u0026thinsp;0.05 were retained. Subsequently, Spearman\u0026rsquo;s rank correlation analysis was applied to evaluate inter-feature relationships among those with high reproducibility. To mitigate multicollinearity, feature pairs with a correlation coefficient exceeding 0.9 were identified, and a greedy recursive elimination approach was adopted to iteratively remove the most redundant feature from each highly correlated pair. This process preserved feature diversity while minimizing redundancy. Ultimately, 23 features were selected for further analysis.A least absolute shrinkage and selection operator (LASSO) regression model was then implemented on the training cohort to construct a predictive signature. LASSO incorporates L1 regularization, which drives the coefficients of less relevant features toward zero under the constraint of a tuning parameter λ. The optimal value of λ was determined via 10-fold cross-validation using the minimum error criterion. Features with non-zero coefficients were retained and linearly combined, weighted by their respective coefficients, to compute a radiomics score for each patient. All LASSO modeling procedures were conducted using the scikit-learn package in Python.\u003c/p\u003e\u003cp\u003eFollowing feature selection via LASSO, the retained features were used to train multiple machine learning classifiers\u0026mdash;including logistic regression (LR), support vector machine (SVM), random forest (RF), and extreme gradient boosting (XGBoost)\u0026mdash;for predictive model development. The final radiomic signature was derived through 5-fold cross-validation.To integrate the radiomic signature with clinical risk factors, a radiomics nomogram was developed on the validation cohort using multivariable logistic regression. This combined tool visually represents the contribution of both imaging and clinical variables. Calibration curves were plotted to evaluate the agreement between nomogram-predicted probabilities and actual observed outcomes.\u003c/p\u003e\u003cp\u003eTo handle the high-dimensional nature of deep learning-derived features (originally 2048 dimensions), we employed principal component analysis (PCA) for dimensionality reduction. By projecting the feature space onto 32 principal components, we improved model generalizability and reduced the risk of overfitting. In this work, we fused handcrafted radiomic features with deep learning-based features to leverage complementary information and enhance predictive performance. An early fusion strategy was adopted to merge feature subsets from both handcrafted and deep learning sources into a unified representation. Feature selection was then performed on this combined set following a procedure consistent with that used for handcrafted features alone.\u003c/p\u003e\n\u003ch3\u003e6.The building of the clinical model and Radiomics-Clinical model\u003c/h3\u003e\n\u003cp\u003eThe development process of clinical biomarkers shares a striking resemblance with that of radioactive markers. Our model incorporates a comprehensive array of risk factors, encompassing both clinical features\u0026mdash;such as age, tPSA, and fPSA\u0026mdash;and imaging parameters, including LRD, ULD, AD, PV, PS, PD, ET, FT, and associated contrast-enhanced ultrasound metrics. Initially, relevant features for model construction were identified using foundational statistical analyses (p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Subsequently, we utilized an identical suite of machine learning models for the construction of the radiomic model. In order to maintain a level playing field for comparisons, we rigorously applied a 5-fold cross-validation strategy and evaluated our model on an independent test cohort.\u003c/p\u003e\u003cp\u003eA radiomic-clinical integrated model was developed by combining the radiomic signature with clinical predictors. The diagnostic performance of this combined model was assessed in the test cohort. ROC curves were generated to evaluate its predictive accuracy. Calibration curves were plotted to visualize the agreement between predicted probabilities and actual outcomes, and the Hosmer-Lemeshow test was applied to statistically assess model fit. Decision curve analysis (DCA) was performed to quantify the clinical utility of the model across different threshold probabilities.\u003c/p\u003e\n\u003ch3\u003e7.Statistical analysis\u003c/h3\u003e\n\u003cp\u003eAll statistical analyses were conducted using SPSS (version 23.0; IBM Corp., Armonk, NY, USA) and Python (version 3.7). Continuous variables were presented as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation (SD), while categorical variables were summarized as frequency and percentage. Group comparisons for continuous variables were performed using the Student\u0026rsquo;s t-test or the Mann\u0026ndash;Whitney U test, as appropriate, and the chi-square test was used for categorical variables. A 95% confidence interval (CI) was estimated through bootstrapping with 2000 iterations. Differences in sensitivity and specificity between the deep learning-based radiomics (DLR) model and radiologists were assessed using McNemar\u0026rsquo;s test. A two-sided p-value less than 0.05 was considered statistically significant.\u003c/p\u003e"},{"header":"Results","content":"\n\u003ch3\u003e1.Patient characteristics\u003c/h3\u003e\n\u003cp\u003eA total of 338 participants were enrolled based on the predefined inclusion and exclusion criteria, which comprised 188 patients diagnosed with prostate cancer and 150 controls without the disease. These participants were then randomly allocated into a training cohort containing 270 individuals and a validation cohort consisting of 68 subjects.In the training group, there were statistically significant differences between the benign group and the malignant group in age, PV, tPSA, fPSA, LRD, PD, RI and CEUS parameter AREA (P\u0026thinsp;\u0026lt;\u0026thinsp;0.05). There was no significant difference in the distribution of the other 7 parameters of CEUS (A, B, k, MSE, TtoPK, AREA, Grad, ATM) (P\u0026thinsp;\u0026gt;\u0026thinsp;0.05). Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e shows the clinical characteristics of patients in the training cohort and testing cohort.\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\u003eBaseline Patient Characteristics in Training cohort and Test cohort.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"10\"\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\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e\u003cp\u003eTraining cohort(n\u0026thinsp;=\u0026thinsp;270)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colspan=\"3\" nameend=\"c7\" namest=\"c5\"\u003e\u003cp\u003eTesting cohort(n\u0026thinsp;=\u0026thinsp;68)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colspan=\"3\" nameend=\"c10\" namest=\"c8\"\u003e\u003cp\u003eValidation cohort(n\u0026thinsp;=\u0026thinsp;108)\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\u003eBenign group\u003c/p\u003e\u003cp\u003e(n\u0026thinsp;=\u0026thinsp;120)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eMalignant group\u003c/p\u003e\u003cp\u003e(n\u0026thinsp;=\u0026thinsp;150)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eBenign group(n\u0026thinsp;=\u0026thinsp;30)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eMalignant group(n\u0026thinsp;=\u0026thinsp;38)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eBenign group(n\u0026thinsp;=\u0026thinsp;46)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003eMalignant group(n\u0026thinsp;=\u0026thinsp;62)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003eP\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003etPSA(ng/m)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e15.34\u0026thinsp;\u0026plusmn;\u0026thinsp;13.92\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e71.88\u0026thinsp;\u0026plusmn;\u0026thinsp;260.42\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e19.71\u0026thinsp;\u0026plusmn;\u0026thinsp;19.16\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e69.21\u0026thinsp;\u0026plusmn;\u0026thinsp;71.54\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e19.27\u0026thinsp;\u0026plusmn;\u0026thinsp;19.40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e48.10\u0026thinsp;\u0026plusmn;\u0026thinsp;36.68\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003efPSA(ng/ml)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2.31\u0026thinsp;\u0026plusmn;\u0026thinsp;4.76\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e14.24\u0026thinsp;\u0026plusmn;\u0026thinsp;25.50\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e3.22\u0026thinsp;\u0026plusmn;\u0026thinsp;3.57\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e16.74\u0026thinsp;\u0026plusmn;\u0026thinsp;18.36\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e2.89\u0026thinsp;\u0026plusmn;\u0026thinsp;3.50\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e12.68\u0026thinsp;\u0026plusmn;\u0026thinsp;16.70\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge(years)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e65.89\u0026thinsp;\u0026plusmn;\u0026thinsp;6.34\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e71.17\u0026thinsp;\u0026plusmn;\u0026thinsp;6.12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e68.67\u0026thinsp;\u0026plusmn;\u0026thinsp;7.47\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e70.87\u0026thinsp;\u0026plusmn;\u0026thinsp;9.34\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.418\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e65.49\u0026thinsp;\u0026plusmn;\u0026thinsp;7.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e70.08\u0026thinsp;\u0026plusmn;\u0026thinsp;6.90\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLRD(mm)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e53.48\u0026thinsp;\u0026plusmn;\u0026thinsp;6.25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e51.73\u0026thinsp;\u0026plusmn;\u0026thinsp;6.32\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e0.009\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e55.27\u0026thinsp;\u0026plusmn;\u0026thinsp;5.83\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e54.29\u0026thinsp;\u0026plusmn;\u0026thinsp;8.17\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.347\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e54.13\u0026thinsp;\u0026plusmn;\u0026thinsp;5.72\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e52.30\u0026thinsp;\u0026plusmn;\u0026thinsp;7.50\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e\u003cb\u003e0.034\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eULD(mm)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e42.51\u0026thinsp;\u0026plusmn;\u0026thinsp;8.84\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e41.32\u0026thinsp;\u0026plusmn;\u0026thinsp;9.28\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.088\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e46.20\u0026thinsp;\u0026plusmn;\u0026thinsp;8.43\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e44.00\u0026thinsp;\u0026plusmn;\u0026thinsp;11.85\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.223\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e42.91\u0026thinsp;\u0026plusmn;\u0026thinsp;7.58\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e41.30\u0026thinsp;\u0026plusmn;\u0026thinsp;10.53\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e0.103\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAD(mm)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e44.66\u0026thinsp;\u0026plusmn;\u0026thinsp;9.50\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e44.13\u0026thinsp;\u0026plusmn;\u0026thinsp;8.49\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.659\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e46.67\u0026thinsp;\u0026plusmn;\u0026thinsp;11.91\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e47.63\u0026thinsp;\u0026plusmn;\u0026thinsp;11.73\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.877\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e45.94\u0026thinsp;\u0026plusmn;\u0026thinsp;9.55\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e44.02\u0026thinsp;\u0026plusmn;\u0026thinsp;8.79\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e0.281\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePV(cm3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e56.34\u0026thinsp;\u0026plusmn;\u0026thinsp;27.89\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e52.36\u0026thinsp;\u0026plusmn;\u0026thinsp;27.64\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.179\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e65.60\u0026thinsp;\u0026plusmn;\u0026thinsp;28.98\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e66.02\u0026thinsp;\u0026plusmn;\u0026thinsp;42.36\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.429\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e58.62\u0026thinsp;\u0026plusmn;\u0026thinsp;25.09\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e54.26\u0026thinsp;\u0026plusmn;\u0026thinsp;35.82\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e0.104\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePS(cm/s)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e26.38\u0026thinsp;\u0026plusmn;\u0026thinsp;13.09\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e27.57\u0026thinsp;\u0026plusmn;\u0026thinsp;12.67\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.288\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e34.90\u0026thinsp;\u0026plusmn;\u0026thinsp;17.84\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e33.05\u0026thinsp;\u0026plusmn;\u0026thinsp;19.80\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.268\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e25.13\u0026thinsp;\u0026plusmn;\u0026thinsp;12.95\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e28.21\u0026thinsp;\u0026plusmn;\u0026thinsp;14.94\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e0.264\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePD(cm/s)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e7.62\u0026thinsp;\u0026plusmn;\u0026thinsp;3.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7.16\u0026thinsp;\u0026plusmn;\u0026thinsp;3.69\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e0.037\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e9.50\u0026thinsp;\u0026plusmn;\u0026thinsp;5.15\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e8.34\u0026thinsp;\u0026plusmn;\u0026thinsp;4.98\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.299\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e8.00\u0026thinsp;\u0026plusmn;\u0026thinsp;3.93\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e7.08\u0026thinsp;\u0026plusmn;\u0026thinsp;3.89\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e0.209\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRI\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.69\u0026thinsp;\u0026plusmn;\u0026thinsp;0.09\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.72\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e0.003\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.72\u0026thinsp;\u0026plusmn;\u0026thinsp;0.09\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.72\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.990\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.67\u0026thinsp;\u0026plusmn;\u0026thinsp;0.08\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.72\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e\u003cb\u003e0.006\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eET(s)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e16.47\u0026thinsp;\u0026plusmn;\u0026thinsp;4.07\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e17.28\u0026thinsp;\u0026plusmn;\u0026thinsp;4.20\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.092\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e17.50\u0026thinsp;\u0026plusmn;\u0026thinsp;4.30\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e16.63\u0026thinsp;\u0026plusmn;\u0026thinsp;3.86\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.921\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e17.04\u0026thinsp;\u0026plusmn;\u0026thinsp;4.49\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e16.98\u0026thinsp;\u0026plusmn;\u0026thinsp;4.28\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e0.511\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFT(s)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e29.99\u0026thinsp;\u0026plusmn;\u0026thinsp;7.87\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e30.42\u0026thinsp;\u0026plusmn;\u0026thinsp;7.38\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.430\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e29.57\u0026thinsp;\u0026plusmn;\u0026thinsp;7.13\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e29.87\u0026thinsp;\u0026plusmn;\u0026thinsp;8.18\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.946\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e30.83\u0026thinsp;\u0026plusmn;\u0026thinsp;7.88\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e31.20\u0026thinsp;\u0026plusmn;\u0026thinsp;7.82\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e0.587\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e24.26\u0026thinsp;\u0026plusmn;\u0026thinsp;3.65\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e24.76\u0026thinsp;\u0026plusmn;\u0026thinsp;4.94\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.819\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e23.30\u0026thinsp;\u0026plusmn;\u0026thinsp;2.98\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e24.05\u0026thinsp;\u0026plusmn;\u0026thinsp;3.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.478\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e24.01\u0026thinsp;\u0026plusmn;\u0026thinsp;3.49\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e24.93\u0026thinsp;\u0026plusmn;\u0026thinsp;3.78\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e0.197\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eB\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-64.49\u0026thinsp;\u0026plusmn;\u0026thinsp;2.84\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e-64.75\u0026thinsp;\u0026plusmn;\u0026thinsp;3.24\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.441\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-63.81\u0026thinsp;\u0026plusmn;\u0026thinsp;2.28\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e-64.29\u0026thinsp;\u0026plusmn;\u0026thinsp;3.35\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.926\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e-64.06\u0026thinsp;\u0026plusmn;\u0026thinsp;3.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e-64.93\u0026thinsp;\u0026plusmn;\u0026thinsp;3.11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e0.086\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ek\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.29\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.27\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.176\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.28\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.28\u0026thinsp;\u0026plusmn;\u0026thinsp;0.11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e1.000\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.29\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.26\u0026thinsp;\u0026plusmn;\u0026thinsp;0.09\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e0.106\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMSE\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.16\u0026thinsp;\u0026plusmn;\u0026thinsp;0.81\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.12\u0026thinsp;\u0026plusmn;\u0026thinsp;0.66\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.890\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e1.05\u0026thinsp;\u0026plusmn;\u0026thinsp;0.48\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1.23\u0026thinsp;\u0026plusmn;\u0026thinsp;0.71\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.634\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.99\u0026thinsp;\u0026plusmn;\u0026thinsp;0.45\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e1.19\u0026thinsp;\u0026plusmn;\u0026thinsp;0.60\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e0.101\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTtoPK(s)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e11.35\u0026thinsp;\u0026plusmn;\u0026thinsp;4.43\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e10.75\u0026thinsp;\u0026plusmn;\u0026thinsp;3.16\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.420\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e11.99\u0026thinsp;\u0026plusmn;\u0026thinsp;4.04\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e10.32\u0026thinsp;\u0026plusmn;\u0026thinsp;2.99\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.157\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e11.26\u0026thinsp;\u0026plusmn;\u0026thinsp;4.37\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e10.66\u0026thinsp;\u0026plusmn;\u0026thinsp;2.88\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e0.838\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAREA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e199.95\u0026thinsp;\u0026plusmn;\u0026thinsp;80.92\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e175.24\u0026thinsp;\u0026plusmn;\u0026thinsp;61.65\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e0.008\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e211.49\u0026thinsp;\u0026plusmn;\u0026thinsp;100.68\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e163.12\u0026thinsp;\u0026plusmn;\u0026thinsp;48.64\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e\u003cb\u003e0.006\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e194.92\u0026thinsp;\u0026plusmn;\u0026thinsp;70.64\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e171.95\u0026thinsp;\u0026plusmn;\u0026thinsp;49.98\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e\u003cb\u003e0.045\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGrad\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2.15\u0026thinsp;\u0026plusmn;\u0026thinsp;0.68\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2.15\u0026thinsp;\u0026plusmn;\u0026thinsp;0.60\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.974\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e2.05\u0026thinsp;\u0026plusmn;\u0026thinsp;0.78\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e2.24\u0026thinsp;\u0026plusmn;\u0026thinsp;0.66\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.248\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e2.19\u0026thinsp;\u0026plusmn;\u0026thinsp;0.73\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e2.18\u0026thinsp;\u0026plusmn;\u0026thinsp;0.57\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e0.951\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eATM(s)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e15.65\u0026thinsp;\u0026plusmn;\u0026thinsp;4.29\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e16.45\u0026thinsp;\u0026plusmn;\u0026thinsp;3.78\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.064\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e16.92\u0026thinsp;\u0026plusmn;\u0026thinsp;4.78\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e15.98\u0026thinsp;\u0026plusmn;\u0026thinsp;3.19\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.595\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e15.95\u0026thinsp;\u0026plusmn;\u0026thinsp;4.77\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e16.03\u0026thinsp;\u0026plusmn;\u0026thinsp;3.34\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e0.490\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"10\"\u003eData are mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation (range) at the time of examination.\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"10\"\u003ePSA prostate-specific antigen ,tPSA total PSA, fPSA free PSA, LRD prostate left and right diameter, ULD prostate upper and lower diameter, AD prostate anteroposterior diameter, PV prostate volume ,PS/PD peak systolic and diastolic blood flow, ET/FT contrast agent entry time and full filling time, A/B/k fitting curve equation coefficient, MSE fitting curve equation error, TtoPK time to peak, AREA area under the curve, Grad peak intensity average gradient, ATM arrival time. SD standard deviation.\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\n\u003ch3\u003e2.Establishment and performance of the radiomics model\u003c/h3\u003e\n\u003cp\u003eThe original 3306 features incorporating both radiological and deep learning aspects were meticulously pruned to a concise set of 34 highly relevant features. This selection process yielded a refined dataset consisting of 29 radiological characteristics and 5 attributes derived from deep learning, which collectively informed the development of a novel multi-parameter Rad-score. The process of screening radiomics features and the detail of the features shows in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eRadiomics machine learning models were developed using all selected features, with the optimal model determined through comparison with LR, SVM, KNN, and MLP classifiers. Notably, the SVM algorithm yielded the highest AUC value within the training cohort, reaching 0.936 (95% confidence interval: 0.905\u0026ndash;0.966), with corresponding sensitivities and specificities of 0.787 and 0.950, respectively. In the validation cohort, the SVM model demonstrated an AUC of 0.823 (95% CI: 0.721\u0026ndash;0.925), with sensitivities and specificities of 0.684 and 0.833, respectively (refer to Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e for detailed statistics).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eDifferences between various machine learning models.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"13\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c10\" colnum=\"10\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c11\" colnum=\"11\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c12\" colnum=\"12\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c13\" colnum=\"13\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModel name\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAccuracy\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAUC\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003e95% CI\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eSensitivity\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eSpecificity\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003ePPV\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e\u003cp\u003eNPV\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c9\"\u003e\u003cp\u003ePrecision\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c10\"\u003e\u003cp\u003eRecall\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c11\"\u003e\u003cp\u003eF1\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c12\"\u003e\u003cp\u003eThreshold\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c13\"\u003e\u003cp\u003eTask\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLR\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.793\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.868\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.825\u0026ndash;0.910\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.847\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.725\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.794\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.791\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.794\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.847\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e0.819\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e\u003cp\u003e0.454\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c13\"\u003e\u003cp\u003eTraining\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.872\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.921\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.844\u0026ndash;0.998\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.929\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.789\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.867\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.882\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.867\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.929\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e0.897\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e\u003cp\u003e0.551\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c13\"\u003e\u003cp\u003eTesting\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.735\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.765\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.648\u0026ndash;0.882\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.868\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.567\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.717\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.773\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.717\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.868\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e0.786\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e\u003cp\u003e0.400\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c13\"\u003e\u003cp\u003eValidation\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSVM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.859\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.936\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.905\u0026ndash;0.966\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.787\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.95\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.952\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.781\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.952\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.787\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e0.861\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e\u003cp\u003e0.669\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c13\"\u003e\u003cp\u003eTraining\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.745\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.803\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.678\u0026ndash;0.928\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.821\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.632\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.767\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.706\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.767\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.821\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e0.793\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e\u003cp\u003e0.5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c13\"\u003e\u003cp\u003eTesting\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.75\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.823\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.721\u0026ndash;0.925\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.684\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.833\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.839\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.676\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.839\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.684\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e0.754\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e\u003cp\u003e0.628\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c13\"\u003e\u003cp\u003eValidation\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eKNN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.681\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.822\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.775\u0026ndash;0.869\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.500\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.908\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.872\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.592\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.872\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.500\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e0.636\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e\u003cp\u003e0.600\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c13\"\u003e\u003cp\u003eTraining\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.574\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.742\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.602\u0026ndash;0.883\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.321\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.947\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.9\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.486\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.9\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.321\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e0.474\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e\u003cp\u003e0.6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c13\"\u003e\u003cp\u003eTesting\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.529\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.723\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.603\u0026ndash;0.844\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.184\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.967\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.875\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.483\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.875\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.184\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e0.304\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e\u003cp\u003e0.800\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c13\"\u003e\u003cp\u003eValidation\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMLP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.833\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.902\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.866\u0026ndash;0.938\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.853\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.808\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.848\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.815\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.848\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.853\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e0.85\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e\u003cp\u003e0.554\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c13\"\u003e\u003cp\u003eTraining\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.766\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.814\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.689\u0026ndash;0.939\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.786\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.737\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.815\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.7\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.815\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.786\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e0.8\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e\u003cp\u003e0.463\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c13\"\u003e\u003cp\u003eTesting\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.779\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.844\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.749\u0026ndash;0.939\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.763\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.8\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.829\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.727\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.829\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.763\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e0.795\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e\u003cp\u003e0.560\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c13\"\u003e\u003cp\u003eValidation\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"13\"\u003eAUC, area under the curve; PPV, positive predictive value; NPV, negative predictive value; F1\u0026thinsp;=\u0026thinsp;2 x (precision x recall)/(precision\u0026thinsp;+\u0026thinsp;recall); LR, logistic regression; SVM, support vector machine; KNN, K‑nearest neighbors; MLP, multi‑layer perceptron. AUC, area under the curve; PPV, positive predictive value; NPV, negative predictive value; F1\u0026thinsp;=\u0026thinsp;2 x (precision x recall)/(precision\u0026thinsp;+\u0026thinsp;recall);\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\n\u003ch3\u003e3.Establishment and performance of the clinical model and RadiomicsClinical model\u003c/h3\u003e\n\u003cp\u003eThe clinical model was constructed based on significant features, as indicated by P-values (\u0026lt;\u0026thinsp;0.05) from the training cohort, including clinical parameters tPSA, fPSA, Age, and imaging characteristics LRD, PD, RI, and ARE, which collectively define the clinical signatures.\u003c/p\u003e\u003cp\u003eThe refined clinical model yielded an AUC of 0.746 (95%CI 0.652\u0026ndash;0.840) with a sensitivity and specificity of 0.693 and 0.883, respectively, in the training dataset. Furthermore, the model demonstrated an AUC of 0.809 (95%CI 0.708\u0026ndash;0.910) in the validation cohort, with corresponding sensitivities and specificities of 0.605 and 0.900 (refer to Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e for detailed results).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003ePredictive Performance of Three Models in the Training Cohort and Validation Cohort.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"10\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c10\" colnum=\"10\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSignature\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTraining Cohort (n\u0026thinsp;=\u0026thinsp;270)\u003c/p\u003e\u003cp\u003eAUC (95% CI)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eSensitivity*\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSpecificity*\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eTesting cohort(n\u0026thinsp;=\u0026thinsp;68)\u003c/p\u003e\u003cp\u003eAUC (95% CI)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eSensitivity*\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003eSpecificity*\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e\u003cp\u003eValidation Cohort (n\u0026thinsp;=\u0026thinsp;108)\u003c/p\u003e\u003cp\u003eAUC (95% CI)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c9\"\u003e\u003cp\u003eSensitivity*\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c10\"\u003e\u003cp\u003eSpecificity*\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eClinic Signature\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.856(0.812\u0026ndash;0.901)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.693\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.883\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.917(0.826\u0026ndash;1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.893\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.895\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.809 (0.708\u0026ndash;0.910)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.605\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.900\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad Signature\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.936(0.905\u0026ndash;0.966)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.787\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.951\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.803(0.678\u0026ndash;0.928)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.821\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.632\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.823 (0.721\u0026ndash;0.925)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.684\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.833\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRadiomics-clinical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.956(0.931\u0026ndash;0.981)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.893\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.925\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.951(0.881\u0026ndash;1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.893\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.947\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.889 (0.813\u0026ndash;0.966)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.737\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.933\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"10\"\u003e*Balanced sensitivity and specificity at the cutoff yielding the largest Youden index value.\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eThe radiomics-clinical model yielded an AUC of 0.956 (95% CI 0.931\u0026ndash;0.981) in the training cohort, with corresponding sensitivities and specificities of 0.893 and 0.925. In the validation cohort, the model demonstrated an AUC of 0.889 (95% CI 0.813\u0026ndash;0.966), with sensitivities and specificities of 0.737 and 0.933, respectively. (Refer to Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e for details.) Subsequently, a nomogram was developed based on the radiomics-clinical model. (Please see Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e7\u003c/span\u003e for the nomogram.) The Decision Curve Analysis (DCA) revealed that the model confers substantial benefits in the majority of scenarios. (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e6\u003c/span\u003e provides further insights into these benefits.)\u003c/p\u003e\u003cp\u003eDelong's examination of the comparison model elucidated a significant disparity in the AUC between the radiomic-clinical model and the sole clinical model within the testing cohort (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Conversely, no significant differentiation was observed between the radiomic-clinical model and the radiomic model (p\u0026thinsp;=\u0026thinsp;0.123).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eTRUS is the most commonly used method in clinical diagnosis of prostate cancer, which has the advantages of high imaging resolution, non-invasive, strong real-time imaging ability, high sensitivity and strong specificity. Recently, there have been study\u003csup\u003e[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]\u003c/sup\u003ethat simply establish radiomics models of transrectal ultrasound to predict prostate cancer, and good results have been achieved. Compared with conventional ultrasound, CEUS can perform real-time dynamic imaging, observe actual blood flow, and evaluate blood perfusion, thus realizing clear visualization of tissue and lesion microcirculation perfusion, and significantly improving the diagnostic efficiency of prostate cancer\u003csup\u003e[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]\u003c/sup\u003e. If the data extracted from static B-model images can be combined with CEUS parameters, it is believed that the diagnosis of prostate cancer will be further improved.\u003c/p\u003e\u003cp\u003eMining deep information from traditional imaging and improving diagnosis and treatment methods are the main directions to conquer prostate cancer. As a non-invasive new technology that can extract features with high throughput, imaging omics has emerged at the moment, showing great advantages in tumor phenotyping, treatment decision-making and prognosis analysis. At present, the imaging omics of prostate cancer mainly focuses on the diagnosis and differential diagnosis of CT, MRI and PET images, tumor staging, efficacy evaluation and prognosis prediction. Yong Wang et al. \u003csup\u003e[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]\u003c/sup\u003eused clinical data CEUS parameters MRI images to establish a model for predicting biochemical recurrence of prostate cancer (PCa) after surgery and to verify its clinical validity, which can help improve its prognosis. Chunyu Li et al. \u003csup\u003e[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]\u003c/sup\u003e established an imaging model based on multi-parameter MRI images to distinguish prostate cancer from prostate hyperplasia, thereby helping clinicians to make better clinical treatment decisions and reduce unnecessary prostate biopsies. In terms of ultrasound imaging of prostate cancer, Wei Ou MD et al. \u003csup\u003e[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]\u003c/sup\u003e developed and validated the imaging score to predict prostate cancer before biopsy in order to reduce unnecessary prostate cancer biopsies. However, they only conducted a single imaging study, without applying the blood perfusion characteristics and deep learning methods of prostate cancer.\u003c/p\u003e\u003cp\u003eIn a prior investigation into the diagnosis of prostate cancer, Ya Sun et al.\u003csup\u003e[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]\u003c/sup\u003e developed a machine learning model that integrated pyradiomic radiomic features with B-mode transrectal ultrasound and contrast-enhanced ultrasound. This innovative model aimed to enhance the detection of peripheral prostate cancer and supply empirical evaluation metrics. In our present study, our target mapping based on CEUS demonstrates a higher degree of accuracy when compared to the direct mapping on the poorly defined B-mode image. And we extended this approach by incorporating a deep learning methodology, performing feature fusion prior to model development, and integrating clinical risk factors to construct a hybrid model. Our findings corroborate that the fusion of deep learning methodologies can significantly enhance the diagnostic accuracy and clinical utility of differentiating PCa. In the present investigation, we developed three predictive models\u0026mdash;each grounded in radiomic model, clinical model, and a hybrid radiomic-clinical model, respectively. Notably, the integrated radiomic-clinical model exhibited superior predictive performance, demonstrating a commendable accuracy in prognostication for prostate cancer. Additionally, a nomogram was constructed to assist urologists and sonographers in clinical decision-making.\u003c/p\u003e\u003cp\u003eRadiomics-based B-medel TRUS image analysis has yielded commendable predictive outcomes, with the resulting model markedly outperforming both the clinical model and the standalone radiomics model. This suggests a promising role for radiomics in the prognostication of prostate cancer, and underscores the superiority of integrated models when compared to their monolithic counterparts.\u003c/p\u003e\u003cp\u003eCertainly, several constraints are inherent to our investigation. Initially, this study collected patients from only two centers, and enrollment of patients from more centers is needed to improve the generalizability of the model in clinical application. Secondly, the extant evidence necessitates corroboration via extensive, prospective, randomized controlled trials. Thirdly, manual lesion segmentation is time-consuming and easy to be affected by the experience of the image reader. In the future, it is necessary to explore automatic or semi-automatic methods to describe the lesions more accurately. Moreover, our inquiry is exclusively predicated on machine learning and deep learning methodologies, excluding more cutting-edge techniques such as deep transfer learning. Subsequent to this iteration of research, we intend to incorporate deep transfer learning approaches into our methodology.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThe radiomics-clinical machine learning model based on B-model combined with traditional radiomics and deep learning can be used for preoperative prediction of prostate cancer patients, providing a new and promising method for urological surgeons to select suitable prostate cancer patients for treatment.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eTRUS: Transrectal ultrasound; CEUS: Contrast-enhanced ultrasound; PCa: Prostate cancer; PSA: Prostate-specific antigen; DRE: Digital rectal examination; LRD: left and right diameter; ULD: upper and lower diameter; AD: anteroposterior diameter; \u0026nbsp;PV: prostate volume; tPSA total PSA; fPSA: free PSA; PS/PD: peak systolic and diastolic blood flow; ET/FT: contrast agent entry time and full filling time; mp‑MRI: Multiparametric magnetic resonance imaging; T2WI: 2-weighted fast spin-echo imaging; DWI: diffusion weighted MR Imaging; DCE-MRI: dynamic contrast-enhanced MRI; DICOM: Digital Imaging and Communications in Medicine; Grad: Peak intensity average gradient; ATM: arrival time; TtoP: time to peak; Area: Area under the curve; ICC: intraclass correlation coefficient; HCR: hand crafted features; GLCM: gray-level co-occurrence matrix; GLRLM: gray-level run length matrix; GLSZM: gray-level size zone matrix; NGTDM: neighborhood gray-level difference matrix; LASSO: least absolute shrinkage and selection operator; LR: logistic regression; SVM: support vector machine; RF: random forest; XGBoost: extreme gradient boosting; PCA: principal component analysis; DCA: Decision curve analysis\u003c/p\u003e\n"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study complies with the 1964 Helsinki Declaration and its subsequent amendments or equivalent ethical standards. It has been approved by the Ethics Committee of Inner Mongolia Autonomous Region People\u0026apos;s Hospital (Approval Number: SC-07/01KT2024022).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll data generated or analyzed during this study are included in this published\u0026nbsp;\u003c/p\u003e\n\u003cp\u003earticle.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeclaration of Competing Interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was supported by the Science and Technology Program of the Joint Fund of Scientific Research for the Public Hospitals of Inner Mongolia Academy of Medical Sciences(2023GLLH0020).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026rsquo; Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSB：conceived and designed the analysis, collected the data, performed the analysis, wrote the paper. QL: conceived and designed the analysis, collected the data, performed the analysis, wrote the paper. HH: conceived and designed the analysis. RS: collected the data. YS: collected the data. LY: collected the data, contributed data or analysis tools. DL: contributed data or analysis tools. XR: Contributed data or analysis tools. WL: performed the analysis. HL: performed the analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgment\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCertain experiments were conducted on the Onekey AI platform, for which we are grateful for the support and assistance provided by Onekey AI and its development team in this scientific research endeavor.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eBray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2024;74(3):229\u0026ndash;63.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSiegel RL, Giaquinto AN, Jemal A. Cancer statistics, 2024. CA Cancer J Clin. 2024;74(1):12\u0026ndash;49.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFalagario UG, Sanguedolce F, Dovey Z, et al. Prostate cancer biomarkers: a practical review based on different clinical scenarios. Crit Rev Clin Lab Sci. 2022;59(5):297\u0026ndash;308.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eIslam MA, Alam SM, Reza AM. Urosepsis and Bacteriuria in Patients Undergoing TRUS-Guided Prostate Biopsy. Mymensingh Med J. 2023;32(2):330\u0026ndash;7.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMorelli M, Sampogna G, Molteni S, et al. The impact of prostate biopsy on erectile and ejaculatory function: A prospective study. Arch Ital Urol Androl. 2022;94(4):420\u0026ndash;3.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJacewicz M, Rud E, Lauritzen P, Baco E. Non-infectious adverse events of transperineal prostate biopsies performed under local anaesthesia. BJU Int. 2024.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBorofsky S, George AK, Gaur S, et al. What Are We Missing? False-Negative Cancers at Multiparametric MR Imaging of the Prostate. Radiology. 2018;286(1):186\u0026ndash;95.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang J, Xu L, Zhang G, et al. Effects of dynamic contrast enhancement on transition zone prostate cancer in Prostate Imaging Reporting and Data System Version 2.1. Radiol Oncol. 2023;57(1):42\u0026ndash;50.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBai X, Jiang Y, Zhang X, et al. The Value of Prostate-Specific Antigen-Related Indexes and Imaging Screening in the Diagnosis of Prostate Cancer. Cancer Manag Res. 2020;12:6821\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSalib A, Halpern E, Eisenbrey J, et al. The evolving role of contrast-enhanced ultrasound in urology: a review. World J Urol. 2023;41(3):673\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYang G, Ruan L. Imaging findings of prostate tuberculosis by transrectal contrast-enhanced ultrasound and comparison with 2D ultrasound and pathology. Br J Radiol. 2022;95(1129):20210713.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJung EM, Wertheimer T, Putz FJ, et al. Contrast enhanced ultrasound (CEUS) with parametric imaging and time intensity curve analysis (TIC) for evaluation of the success of prostate arterial embolization (PAE) in cases of prostate hyperplasia. Clin Hemorheol Microcirc. 2020;76(2):143\u0026ndash;53.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWink M, Frauscher F, Cosgrove D, et al. Contrast-enhanced ultrasound and prostate cancer; a multicentre European research coordination project. Eur Urol. 2008;54(5):982\u0026ndash;92.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiu Y, Lu D, Xu G, et al. Diagnostic accuracy of qualitative and quantitative magnetic resonance imaging-guided contrast-enhanced ultrasound (MRI-guided CEUS) for the detection of prostate cancer: a prospective and multicenter study. Radiol Med. 2024;129(4):585\u0026ndash;97.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLin Y, Belue MJ, Yilmaz EC, et al. Deep Learning-Based T2-Weighted MR Image Quality Assessment and Its Impact on Prostate Cancer Detection Rates. J Magn Reson Imaging. 2024;59(6):2215\u0026ndash;23.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWildeboer RR, Mannaerts CK, van Sloun R, et al. Automated multiparametric localization of prostate cancer based on B-mode, shear-wave elastography, and contrast-enhanced ultrasound radiomics. Eur Radiol. 2020;30(2):806\u0026ndash;15.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiang L, Zhi X, Sun Y, et al. A Nomogram Based on a Multiparametric Ultrasound Radiomics Model for Discrimination Between Malignant and Benign Prostate Lesions. Front Oncol. 2021;11:610785.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLorusso V, Kabre B, Pignot G, et al. External validation of the computerized analysis of TRUS of the prostate with the ANNA/C-TRUS system: a potential role of artificial intelligence for improving prostate cancer detection. World J Urol. 2023;41(3):619\u0026ndash;25.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSun Y, Fang J, Shi Y, et al. Machine learning based on radiomics features combing B-mode transrectal ultrasound and contrast-enhanced ultrasound to improve peripheral zone prostate cancer detection. Abdom Radiol (NY). 2024;49(1):141\u0026ndash;50.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTurkbey B, Rosenkrantz AB, Haider MA, et al. Eur Urol. 2019;76(3):340\u0026ndash;51. Prostate Imaging Reporting and Data System Version 2.1: 2019 Update of Prostate Imaging Reporting and Data System Version 2.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHuang Y, Qian H, Zheng Y, Song H, Liu X. A radiomics model based on transrectal ultrasound for predicting prostate cancer. Med Ultrason. 2024;26(2):138\u0026ndash;46.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMaxeiner A, Fischer T, Schwabe J, et al. Contrast-Enhanced Ultrasound (CEUS) and Quantitative Perfusion Analysis in Patients with Suspicion for Prostate Cancer. Ultraschall Med. 2019;40(3):340\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang Y, Feng G, Wang J et al. Contrast-Enhanced Ultrasound-Magnetic Resonance Imaging Radiomics Based Model for Predicting the Biochemical Recurrence of Prostate Cancer: A Feasibility Study. Comput Math Methods Med. 2022. 2022: 8090529.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi C, Deng M, Zhong X, et al. Multi-view radiomics and deep learning modeling for prostate cancer detection based on multi-parametric MRI. Front Oncol. 2023;13:1198899.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eOu W, Lei J, Li M, et al. Ultrasound-based radiomics score for pre-biopsy prediction of prostate cancer to reduce unnecessary biopsies. Prostate. 2023;83(1):109\u0026ndash;18.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-imaging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmim","sideBox":"Learn more about [BMC Medical Imaging](http://bmcmedimaging.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmim/default.aspx","title":"BMC Medical Imaging","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Prostate cancer, radiomics, deep learning, MRI-TRUS fusion, contrast-enhanced ultrasound, machine learning, predictive model","lastPublishedDoi":"10.21203/rs.3.rs-7653994/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7653994/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eObjective\u003c/h2\u003e\u003cp\u003eThis study aimed to develop and validate an integrated radiomics deep learning signature combining MRI-guided transrectal ultrasound (TRUS) and contrast-enhanced ultrasound (CEUS) parameters for improved prediction of prostate cancer (PCa).\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003eThis bicentric retrospective study enrolled 443 patients with suspected PCa confirmed by histopathology. Each patient underwent multiparametric MRI and TRUS-guided CEUS. Radiomic and deep learning features were extracted from B-mode ultrasound images using an AI-based platform. Feature selection was performed using statistical and regression methods. Machine learning classifiers were developed and merged with clinical parameters into a combined model. Performance was evaluated via ROC analysis, calibration curves, and decision curve analysis.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eThe SVM-based radiomics model achieved an area under the curve (AUC) of 0.936 (95% CI: 0.905\u0026ndash;0.966) in the training cohort and 0.823 (95% CI: 0.721\u0026ndash;0.925) in the validation cohort. The clinical model alone yielded an AUC of 0.856 (95% CI: 0.812\u0026ndash;0.901) in the training cohort and 0.809 (95% CI: 0.708\u0026ndash;0.910) in the validation cohort. The integrated radiomics-clinical model demonstrated superior performance, with an AUC of 0.956 (95% CI: 0.931\u0026ndash;0.981) in the training cohort and 0.889 (95% CI: 0.813\u0026ndash;0.966) in the validation cohort. DCA confirmed the clinical utility of the combined model across a wide range of threshold probabilities.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e\u003cp\u003eThe integration of radiomics deep learning features with CEUS parameters and clinical risk factors significantly enhances the accuracy of PCa prediction. This non-invasive approach shows promise for supporting clinical decision-making and reducing unnecessary biopsies.\u003c/p\u003e","manuscriptTitle":"Development and validation of a radiomics deep learning signature from MRI-guided transrectal ultrasound combined with ultrasound imaging parameters for prostate cancer prediction","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-08 15:38:37","doi":"10.21203/rs.3.rs-7653994/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewersInvited","content":"","date":"2025-12-05T14:48:12+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-11-10T12:48:34+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-09-29T10:34:13+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-09-29T09:54:27+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Imaging","date":"2025-09-29T09:43:32+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-imaging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmim","sideBox":"Learn more about [BMC Medical Imaging](http://bmcmedimaging.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmim/default.aspx","title":"BMC Medical Imaging","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"750d1592-de50-4073-bf22-c000e8b14ed7","owner":[],"postedDate":"December 8th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-12-08T15:38:37+00:00","versionOfRecord":[],"versionCreatedAt":"2025-12-08 15:38:37","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7653994","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7653994","identity":"rs-7653994","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}