An Innovative Approach for Predicting Prostate Cancer Gleason Grading: Machine Learning-based Fusion of Multimodal Ultrasound, Clinical and Laboratory Indicators 

An Innovative Approach for Predicting Prostate Cancer Gleason Grading: Machine Learning-based Fusion of Multimodal Ultrasound, Clinical and Laboratory Indicators | 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 An Innovative Approach for Predicting Prostate Cancer Gleason Grading: Machine Learning-based Fusion of Multimodal Ultrasound, Clinical and Laboratory Indicators Wenlong Xie, GuangZhen Wu, XiaoChen Qi, Lin Zhong, LiYing Guo, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6998221/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background: Prostate cancer is a common malignancy among elderly males with a growing incidence. While prostate biopsy remains the gold standard for diagnosis, this invasive procedure is poorly tolerated by some patients. The Gleason grade group (GGG) plays a critical role in predicting metastatic risk, guiding treatment selection, and is strongly associated with survival outcomes. Consequently, noninvasive prediction of prostate cancer Gleason grading has emerged as a research priority. This study aimed to develop a noninvasive predictive model integrating multimodal ultrasound data and clinical laboratory biomarkers to preoperatively determine GGGs in prostate cancer patients. Methods: This single-center prospective study enrolled 329 prostate cancer patients meeting predefined inclusion criteria. All participants underwent prostate biopsy with subsequent Gleason grading and were categorized into three groups: low-grade (Gleason score ≤6), intermediate-grade (Gleason score 7), and high-grade (Gleason score ≥8). Thirty-seven predictive parameters were collected, including clinical laboratory biomarkers, systemic inflammatory markers (e.g., neutrophil-to-lymphocyte ratio), and multimodal ultrasound data: Grayscale sonographic characteristics, contrast-enhanced ultrasound (CEUS) parameters, elastography parameters, and radiofrequency signal data. Following feature selection, five clinically significant predictors were identified. Multiple machine learning algorithms were implemented for predictive modeling, and model performance was quantified using accuracy, recall, and F1-score. Results: Six machine learning-based predictive models were developed and evaluated. The Decision Tree model achieved an accuracy of 0.818, recall of 0.818, and F1-score of 0.816. The Random Forest classifier demonstrated an accuracy of 0.820, recall of 0.820, and F1-score of 0.820. The K-Nearest Neighbors algorithm yielded an accuracy of 0.788, recall of 0.788, and F1-score of 0.801. The Gradient Boosting Decision Tree (GBDT) model exhibited superior predictive capability with an accuracy of 0.848, recall of 0.848, and F1-score of 0.849. The XGBoost algorithm had an accuracy of 0.818, recall of 0.789, and F1-score of 0.796, while the Naive Bayes classifier attained an accuracy of 0.773, recall of 0.773, and F1-score of 0.779. Comparative analysis revealed that the GBDT model demonstrated optimal performance among the evaluated algorithms, suggesting its potential clinical significance in predicting Gleason grades. Conclusion : Ultrasonography, being noninvasive, radiation-free, and cost-effective, demonstrates high clinical feasibility for implementation in routine practice, particularly in primary healthcare settings. The predictive model established through multimodal ultrasound parameters effectively predicts the Gleason grade of prostate cancer. prostate cancer Gleason grading ultrasonography predictive model machine learning inflammatory biomarkers Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Introduction Prostate cancer remains the second most prevalent malignancy and the fifth leading cause of cancer-related mortality among men worldwide. The Global Cancer Statistics report estimates 1.4 million new cases and 375,000 deaths attributed to prostate cancer annually ( 1 ). Contemporary epidemiological studies attribute the escalating incidence to accelerated population aging and widespread implementation of prostate-specific antigen (PSA) screening protocols ( 2 ). Despite therapeutic advancements, the 5-year survival rate for metastatic prostate cancer persists below 30% ( 3 ), underscoring the critical need for early lesion detection and precise risk stratification. Accurate histological grading enables clinicians to formulate personalized treatment strategies, thereby mitigating overtreatment risks while optimizing oncological outcomes ( 4 ). The Gleason grading system serves as the cornerstone histopathological framework for evaluating prostate cancer aggressiveness, stratifying tumor biological behavior through meticulous analysis of glandular architectural patterns in hematoxylin and eosin (H&E)-stained specimens. Originally conceptualized by Dr. Donald Gleason in 1966, this classification has undergone two seminal revisions under the auspices of the International Society of Urological Pathology in 2014 and 2019 ( 5 , 6 ). These updates have culminated in the contemporary five-tier Grade Group system, ranging from Grade Group 1 (well-differentiated) to Grade Group 5 (poorly differentiated). Extensive clinical evidence supports the prognostic value of the Gleason grade groups to metastatic risk (hazard ratio = 3.21, 95%CI: 2.74–3.76), guiding therapeutic pathway selection, and influencing overall survival outcomes (p < 0.001) ( 7 ). While prostate biopsy remains the gold standard for prostate cancer diagnosis, emerging evidence suggests discrepancies between biopsy-derived Gleason grades and those identified in postoperative histopathological specimens, potentially leading to misclassification of disease aggressiveness and suboptimal therapeutic decisions ( 8 ). To improve the accuracy of Gleason grade assessment, novel noninvasive methodologies have been increasingly explored alongside advancements in diagnostic technologies. A multicenter study conducted in 2022 identified circulating miR-141-3p as a promising biomarker for predicting Gleason grade progression during active surveillance, demonstrating 82% sensitivity in serial plasma analyses ( 9 ). Loeb et al. further advanced serum-based risk stratification by establishing that combined prostate-specific antigen density (> 0.15 ng/mL/cm³) and Prostate Health Index (≥ 35) achieved 77% specificity in discriminating high-grade carcinomas (Gleason score ≥ 7), outperforming conventional PSA ( 10 ). Multiparametric magnetic resonance imaging (mpMRI), which integrates T2-weighted imaging, diffusion-weighted imaging (DWI), and dynamic contrast-enhanced sequences, has standardized lesion characterization. Using the PI-RADS v2.1 criteria (Prostate Imaging Reporting and Data System version 2.1), mpMRI achieves a negative predictive value exceeding 90% for excluding high-grade tumors (Gleason score ≥ 4 + 3) ( 11 ). Pallwein et al. demonstrated that combining hemodynamic parameters of contra-enhanced ultrasound with quantitative elasticity measurements can significantly improve the sensitivity of prostate cancer detection while predicting lesions with higher Gleason grades ( 12 ). Correas et al. utilized ultrasound elastography to quantify tissue stiffness differences, demonstrating 85% sensitivity in distinguishing low-grade (Gleason 6) from high-grade (Gleason ≥ 7) tumors ( 13 ). Since most current prostate biopsies are performed under ultrasound guidance, this study focuses on combining ultrasound multimodal analysis with clinical and laboratory indicators to predict prostate cancer Gleason grade groups. This approach provides two key advantages, including enabling noninvasive prediction of tumor aggressiveness and offering real-time guidance during biopsy procedures to help clinicians identify the primary tumor location. By directing biopsy needles to areas most likely containing the highest Gleason grade lesions, our approach effectively minimizes discrepancies between preoperative biopsy results and final Gleason grade determinations from radical prostatectomy specimens. This study employs a novel multimodal ultrasound protocol comprising: ( 1 ) dual-plane transrectal probe for ultrasonographic image characterization, ( 2 ) Elastography for tissue stiffness quantification, ( 3 ) contrast-enhanced ultrasound (CEUS) for vascular pattern analysis, and ( 4 ) ultrasound radiofrequency (RF) signal processing of raw data from diagnostic scanners. The RF signal analysis specifically utilizes unprocessed backscatter signals (frequency range: 2–10 MHz) to objectively quantify intrinsic tissue properties through spectral parameter extraction, circumventing potential artifacts introduced by conventional post-processing algorithms ( 14 ). To date, no academic research has explored the use of multimodal ultrasound in combination with clinical indicators to predict prostate cancer Gleason grading. This study aimed to establish a noninvasive predictive model for prostate cancer Gleason grading, providing more comprehensive diagnostic evidence for clinical practice. Methods Data Collection This single-center prospective study was conducted at the First Hospital affiliated to Dalian Medical University and enrolled 329 prostate cancer patients meeting predefined inclusion criteria between February 2024 and May 2025. This study was approved by the Ethics Committee of the The First Affiliated Hospital of Dalian Medical University. All participants provided written informed consent prior to enrollment and data collection. All procedures were conducted in accordance with the ethical standards of the institutional research committee and the Declaration of Helsinki. All participants underwent prostate biopsies followed by Gleason grading and were divided into three groups: low-grade (n = 147, Gleason score ≤ 6), intermediate-grade (n = 106, Gleason score 7), and high-grade (n = 76, Gleason score ≥ 8). Patient characteristics were divided into two categories: clinical and laboratory data and multimodal ultrasonic data. Multimodal ultrasound data were independently collected under double-blind conditions by two physicians with more than 10 years of experience in ultrasound diagnostic, and the measurements were averaged before being included in this study. Sample Inclusion Criteria The inclusion criteria were as follows: ( 1 ) the patient was suspected of prostate cancer and agreed to undergo prostate biopsy at our hospital, and was diagnosed with prostate cancer; ( 2 ) ultrasound imaging showed clearly defined nodules, and the patient had no known allergic reactions (e.g., protein allergies) to ultrasound contrast agent; ( 3 ) patients agreed to undergo multimodal ultrasound examination and other relevant contents and provided written informed consent; ( 4 ) no history of other tumors; ( 5 ) initial diagnostic examination for prostate cancer without endocrine therapy or radiotherapy. Sample Exclusion Criteria The exclusion criteria included the following: ( 1 ) patients with coagulation disorders; ( 2 ) patients unable to cooperate with prostate biopsy for other reasons, such as severe cardiac dysfunction, anal surgery, and mental illness; ( 3 ) incomplete clinical data; ( 4 ) history of prostate surgery ( 5 ) images were unclear or the internal nodules of the prostate could not be accurately located; ( 6 ) unable to collect the complete data required for the study. Clinical and Laboratory Data The collected clinical and laboratory parameters included age, body mass index (BMI), total prostate-specific antigen (tPSA), free PSA (fPSA), fPSA/tPSA ratio, neutrophil count, lymphocyte count, eosinophil count, monocyte count, platelet count, neutrophil-to-lymphocyte ratio (NLR), eosinophil-to-lymphocyte ratio (ELR), lymphocyte-to-monocyte ratio (LMR), and platelet-to-lymphocyte ratio (PLR). Multimodal Ultrasonic Data Transrectal Two-dimensional Ultrasound Data This study used two-dimensional transrectal ultrasound measurements. Patient were positioned in the left lateral position with both knees flexed to the chest. After the surface of the probe was evenly coated with sterile coupling gel, an isolation sleeve was placed on the probe before placing the probe at the level of the prostate. The gain was 60–70 dB, and the depth was adjusted to fully display the outline of the prostate (usually 8–12 cm) to ensure that clear and standard transverse and sagittal images could be obtained (Fig. 1 A). The following two-dimensional ultrasound parameters were collected: ( 1 ) prostate volume, ( 2 ) regularity of prostate shape, ( 3 ) uniformity of the internal echogenicity of the prostate, ( 4 ) integrity of the prostate capsule, ( 5 ) tumor shape, ( 6 ) tumor boundary, ( 7 ) maximum diameter of the tumor, and ( 8 ) tumor echogenicity. Elastography Data Strain elastography mode (sampling frame depth 40 ± 5 mm) was performed using the same inspection mode. Operators followed a real-time pressure feedback indicator to maintain stable pressure during the procedure. The parameters recorded were: elastic score and elastic ratio. The modified 5-point method was used for elasticity scoring as follows ( 15 ): one point indicated the whole lesion was green (strain rate ≥ 80%); Two points represented a mixture of blue and green mixed, with blue accounting for 75% or with rear sound shadow attenuation (Fig. 1 B). The elastic ratio was measured by manually selecting the regions of interest (ROI), and the average strain ratio of the target lesion to the normal gland was measured (Fig. 1 C). Contrast Ultrasound Data Contrast material was administered using SonoVue (Bracco Imaging), a sulfur hexafluoride microbubble formulation, through an antecubital vein by bolus injection of a 19-g trocar (dose, 2.4 mL), followed by flushing of 5 mL of normal saline (0.9% NaCl). The system was switched to CEUS mode with a mechanical index of 0.10 ± 0.02. The time-intensity curve of the lesion area was drawn using the Q-Lab software, and the corresponding parameters were recorded (Fig. 2 ). Parameters included: ( 1 ) Time to peak (Ttopk), ( 2 ) Area under the curve (Area), ( 3 ) Contrast agent onset to peak gradient (Grad), ( 4 ) Contrast agent arrival time (ATm), and ( 5 ) Peak intensity (PI). Radio-frequency Signal Data The system was configured in second harmonic mode for RF signal acquisition, with the transmitter frequency set at 5 MHz and the receiver frequency adjusted to 10 MHz. After collection, the RF signal (sampling rate 40 MHz) was exported through the original data port of the ultrasound system. Post-processing was performed using MATLAB (version R2019a.V96; MathWorks Inc.), which automatically generated eight quantitative parameters (Fig. 3 ). The parameters of ultrasound RF signal included: ( 1 ) Fractal dimension (Higuchi), which quantifies the complexity and irregularity of signal or tissue structure; ( 2 ) Slope, which is used to analyze the attenuation characteristics of tissue to ultrasound; ( 3 ) Intercept, which reflects the overall strength of the echo signal; ( 4 ) Mid-band parameter, combined with the slope and intercept, represents the average attenuation slope in the frequency band. Additionally, the spectrum of the RF signal was split into four equal frequency bands: Low frequency bands (S1), Medium-low frequency bands (S2), Medium-high frequency bands (S3), and High frequency bands (S4). These parameters represent the attenuation coefficients in different frequency bands and quantify tissue properties from different perspectives, such as structural complexity, attenuation properties, scattering intensity, and frequency band specificity, providing multidimensional information for noninvasive diagnosis. Data Analysis Data were analyzed using Python version 3.12.4, with multiple Python libraries applied for data processing, model construction, and evaluation. Scikit-learn provided foundational support for machine learning algorithms. Pandas and NumPy were employed for data manipulation and analysis. Key computational models included: DecisionTreeClassifier (Decision Tree), RandomForestClassifier (Random Forest), GradientBoostingClassifier (Gradient Boosting Decision Trees, GBDT), KNeighborsClassifier (K-Nearest Neighbors, KNN), XGBoost (Extreme Gradient Boosting), and naive_bayes (Naïve Bayes). The metrics module of scikit-learn were used to calculate accuracy, recall, and F1 values to evaluate model performance. Results were visualized using the matplotlib library. Results Development of a Prediction Model During the feature selection phase, non-parametric tests were first applied to assess the relationship between the dependent variable (prostate cancer Gleason grading) and predictive parameters. Results were summarized in a three-line table (Table 1). The following parameters exhibited significant associations with Gleason grading (P < 0.05): TPSA, FPSA, Intercept, S1, TtoPK, Grad, PI, elastic_score, elastic_ratio, Mid_band, S4, and Atm. Both the random forest Gini impurity algorithm and FastSHAP algorithm were employed to calculate feature importance scores and contribution values. The relative importance rankings of these features are illustrated in Figures 4 and 5. Based on the non-parametric test results, metric importance evaluation, and SHAP value contribution analysis, we ultimately selected five predictive indicators (Grad, TtoPK, S1, Intercept, and PI) as characteristic features for assessing prostate cancer Gleason grade. Utilizing these indicators, we developed six machine learning prediction models, including Decision Tree, Random Forest, KNN, XGBoost, GBDT, and Naïve Bayes classifiers for outcome prediction. The data was divided into the training set and the test set in a 4:1 ratio during the construction of the prediction model. Model Predictive Performance The decision tree model visualization is shown below Figure 6. Decision Tree, Random Forest, KNN, XGBoost, GBDT, and Naïve Bayes model visualization are shown below Figure 7. To evaluate model performance, the accuracy rate, recall rate and F1 value of each model were calculated (Table 2 and Figure 8). Comprehensive comparative analysis of model performance metrics revealed that the GBDT model achieved close to 85% accuracy, with superior F1-score and recall rates compared to other models, establishing it as the optimal predictive model for Gleason grade classification in prostate cancer. Discussion Overview of Current Research Prostate cancer has a high incidence rate and significantly impacts the health of middle-aged and elderly males. Owing to its histological characteristics, the tumor frequently harbors coexisting cancer foci with varying degrees of differentiation ( 16 ). The differentiation status of these lesions, which is currently mainly distinguished by Gleason grading, directly determines clinical treatment strategies ( 17 ). Therefore, accurate assessment of the Gleason grade is critical in prostate cancer management. Most studies have demonstrated discrepancies between pathological findings from prostate biopsy and subsequent radical prostatectomy specimens, primarily manifested as lower Gleason grade group assignments in biopsy results compared to final surgical pathology ( 18 ). Such underestimation can mislead clinical judgement of disease severity and lead to inappropriate therapeutic decision-making‌ (19). Compared to other imaging modalities, ultrasound examination offers several advantages, including speed of assessment, absence of ionizing radiation, and affordability, rendering it an indispensable modality for prostate cancer screening. Furthermore, transrectal ultrasound remains the primary method for guiding prostate biopsies. Consequently, identifying sonographic characteristics associated with high-Gleason-score lesions within the prostate holds dual clinical significance, including enabling noninvasive prediction of Gleason grades and improving biopsy accuracy by allowing precise targeting of high-grade lesions. This approach enhances diagnostic concordance between biopsy results and final postoperative histopathology. Historically, imaging research on prostate cancer has predominantly focused on MRI for both diagnosis and Gleason grade determination. The advent of mpMRI and the continuous refinement of the PI-RADS v2.1 have significantly enhanced diagnostic precision for prostate cancer ( 20 , 21 ). Recent studies indicate that mpMRI demonstrates a high negative predictive value (NPV > 90%) for excluding clinically significant high-grade carcinomas (Gleason score ≥ 4 + 3) ( 22 ). In their study, Pecoraro M et al. found that poorly differentiated tumors are associated with restricted diffusion, establishing a statistically significant correlation between reduced apparent diffusion coefficient (ADC) values and higher Gleason grades( 23 ). A meta-analysis by Ponsiglione et al., encompassing 17 studies with a pooled cohort of 4,236 patients, demonstrated that radiomics-based models derived from prostate mpMRI achieved a pooled sensitivity and specificity of 82% and 79%, respectively (area under the curve [AUC] = 0.83) for detecting extraprostatic extension (EPE). This represents a 12% improvement in specificity compared to conventional MRI assessment ( 24 ). Despite demonstrating excellent sensitivity and specificity in prostate cancer diagnosis, MRI has several limitations: ( 1 ) high equipment costs restrict its accessibility in resource-limited settings ( 25 ); ( 2 ) moderate interobserver agreement for PI-RADS scoring (κ = 0.50–0.65), indicating persistent variability influenced by radiologists' experience ( 26 ); ( 3 ) contraindications such as renal insufficiency, claustrophobia, and ferromagnetic implants; ( 4 ) prolonged scan durations and substantial per-examination costs ( 27 ). Ultrasonography demonstrates distinct clinical advantages that address these limitations. First, it eliminates the need for contrast agent administration, thereby avoiding the risk of nephrogenic systemic fibrosis risk associated with gadolinium-based agents ( 28 ). Second, the procedure is operationally simpler, making it suitable for resource-limited medical settings. Third, ultrasound allows for rapid, real-time screening and dynamic monitoring, as well as precise guidance for interventions. The ongoing technological evolution of ultrasound systems and diversification of diagnostic methodologies have established ultrasonography as an indispensable modality in the diagnostic workflow of prostatic diseases. Advantages of Ultrasound Examination In recent years, the diagnostic methods of prostate cancer by ultrasound technology have been constantly updated. For instance, Dias et al. demonstrated that high-frequency micro-ultrasound can achieve real-time detection of EPE with 87% sensitivity and seminal vesicle invasion with 79% specificity, while maintaining a negative predictive value (NPV) of 93% for Gleason grade group upgrading ( 29 ). Liu et al. developed a biparametric ultrasound (BUS) scoring system by integrating grayscale ultrasound, doppler blood flow imaging, and CEUS features for risk stratification of clinically significant prostate cancer. Their findings revealed diagnostic performance comparable to mpMRI in both training and validation cohorts. Notably, in the low-PSA subgroup (PSA < 10 ng/mL), BUS demonstrated significantly superior specificity over MRI (65% vs. 39%, respectively) ( 30 ). Cheng et al. developed a Gleason score prediction model for prostate cancer by extracting 103 radiomics features from transrectal ultrasound images of 838 patients. Using LASSO regression for feature selection, they identified 10 major radiological features. The random forest algorithm achieved an AUC of 0.770 in the testing cohort, with peak predictive performance for high-grade tumors (Gleason score ≥ 9; AUC = 0.847). These results underscore the potential of ultrasound-based radiomics in noninvasive histopathological grading ( 31 ). While most ultrasound-based studies rely on single-modality imaging, our study comprehensively incorporates multimodal ultrasonographic metrics, including primary grayscale ultrasound characteristics, CEUS perfusion parameters, elastography findings, and RF signal-derived quantitative data—an approach not previously reported in existing literature. Furthermore, we integrated patients' baseline clinical characteristics with routine laboratory parameters. These readily accessible parameters are noninvasive and radiation-free, without imposing additional financial burdens, and demonstrate high reproducibility and patient acceptability. This methodology demonstrates potential for widespread clinical adoption, particularly in resource-limited settings and primary care institutions. Innovative Aspects of the Study This study innovatively incorporated RF signals and elasticity ratio as novel parameters, a methodological advancement not previously documented in existing literature. Ultrasonic RF signals represent unprocessed raw high-frequency electrical signals directly acquired and converted by piezoelectric transducers in ultrasound probes. These signals preserve comprehensive amplitude, phase, and frequency information, establishing a robust data foundation for high-precision tissue characterization ( 32 , 33 ). By analyzing acquired RF signals using mathematical modeling techniques such as spectral analysis and time-frequency transformation, the physical properties of tissues can be directly acquired with reduced empirical dependence, thereby yielding more objective and credible results ( 34 ). The elasticity ratio, defined as the ratio of hardness between normal prostate tissue and prostate cancer lesions, mitigates confounding factors such as respiratory motion and probe compression compared to conventional elastography techniques, thereby providing a more accurate representation of the lesion's factual hardness within the prostate ( 35 ).‌ Systemic inflammatory indices, including the NLR and PLR, indirectly reflect the interplay between inflammatory responses and immunosuppression within the tumor microenvironment by quantifying the homeostasis of immune cell subsets in peripheral blood (36). In recent years, systemic inflammatory indices have emerged as pivotal biomarkers reflecting the tumor microenvironment and host immune status, demonstrating significant value in prognostic evaluation and malignancy prediction of cancers. Building upon recent research advances, this study pioneers the incorporation of multiple systemic inflammatory indices combined with multimodal ultrasound parameters for predicting prostate cancer Gleason grading. Analysis of Results As demonstrated in the SHAP analysis, the predictors elastic ratio and Grad exhibited strong correlations. This observation may be attributed to their close conformity to tumor biological characteristics and histopathological alterations. An increase in tumor malignancy grade is accompanied by multiple pathological processes, including increased tumor cell proliferation and density, stromal fibrosis with collagen deposition, alterations in extracellular matrix composition, and inflammatory responses accompanied by immune infiltration. Collectively, these processes contribute to enhanced tissue stiffness, a phenomenon consistent with the principle that higher Gleason grades correlate with increased tissue rigidity in prostate cancer. These findings align with those reported by Barr et al. ( 37 ). Furthermore, the elastic ratio adopted is inherently resistant to confounding factors such as operator-dependent compression force, pelvic floor muscle contraction, and respiratory excursion magnitude, which frequently distort absolute elasticity measurements in prior studies ( 38 ). Compared with previous studies, the index of elastic ratio represent a more accurate measurement of the hardness than only measuring the elastic value of the lesion. Therefore, we believe that the above content is the fundamental reason for the high correlation of elastic ratio. Grad reflects the rate at which the contrast agent enters the lesion (arterial phase) until reaching peak enhancement intensity (peak phase), quantified as the slope of enhancement intensity per unit time. A higher Grad value indicates faster contrast agent microbubble filling within the lesion per unit time, suggesting elevated local blood perfusion rates ( 39 ). This phenomenon may be associated with tumor-related hemodynamic factors, such as arteriovenous shunting caused by invasive tumor growth, which reduces vascular resistance and facilitates rapid perfusion. The tumor demonstrates a hypervascular state characterized by abundant but architecturally disorganized neovascularization, resulting in pathologically enhanced blood supply perfusion ( 40 ).Consequently, higher Gleason grades are associated with greater tumor invasiveness concomitant with more pronounced disruption of glandular architecture, increased neovascular density, and elevated frequencies of arteriovenous shunts, collectively contributing to augmented Grad values. However, the model's algorithmic reliance on highly correlated parameters such as Grad and elastic ratio may introduce diagnostic bias through inadequate consideration of other confounding indicators. Limitations and Future Directions Although the prediction model constructed in this study, based on multimodal ultrasound imaging, clinical, and laboratory features, demonstrated certain clinical value in Gleason grade assessment, the following limitations should be acknowledged. First, this study included 329 patients with a relatively small sample size, and all data were derived from a single medical institution, which may introduce selection bias. Second, there is a lack of sufficient validation cohorts, particularly external validation. Finally, the selection of ROI during the analysis of RF signal images remains subjective and operator-dependent. Despite these limitations, this study holds potential for further exploration. Future research could focus on several key areas. First, developing a comprehensive predictive system to integrate cross-domain multimodal data, such as mpMRI, PET-CT, and liquid biopsy biomarkers. Second, expanding the sample size using multicenter prospective cohort studies. Third, implementing artificial intelligence technologies such as deep learning algorithms to automate the identification of multimodal ultrasound imaging data, thereby reducing operator dependency. Finally, future experimental designs should aim to improve differentiation between histologically Gleason Grade Group 2 (3 + 4) and Grade Group 3 (4 + 3) subtypes. Conclusions In conclusion, multimodal ultrasound reflects the characteristics of prostate cancer lesions from different perspectives, Contrast-enhanced ultrasound patterns provide a more thorough analysis of tumor vascularity, while analysis of radiofrequency signals objectively extracts information directly from the raw fundamental data. Especially by combining various modalities and using the method of machine learning analysis, the optimal prediction indicators can be found more accurately, thereby establishing a prediction model. Our study demonstrates that machine learning-based models can effectively stratify malignancy degree of prostate cancer, thereby guiding subsequent patient management. Declarations Ethics approval and consent to participate This study complied with the 1964 Helsinki Declaration and its lateramendments ethical standards, This study has been granted ethical approval by the Ethics Committee of First Affiliated Hospital of Dalian Medical University.(NO. PJ-KS-KY-2025-184). Consent for publication Not applicable. Data availability The datasets generated and analyzed in this study are available from the corresponding author on reasonable request. Conflict of Interest The authors declare no competing interests. Funding Not applicable. Author Contributions WX: Preparation, creation and/or presentation of the published work, specifically writing the initial draft (including substantive translation); XQ and GW: Development or design of methodology; creation of models; LZ: Conducting a research and investigation process, specifically performing the experiments, or data/evidence collection; LG and MT: Application of statistical, mathematical, computational, or other formal techniques to analyze or synthesize study data; YC: Ideas; formulation or evolution of overarching research goals and aims. Acknowledgments Not application. References Sung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin. 2021;71(3):209-249. Siegel RL, Miller KD, Wagle NS, et al. Cancer Statistics, 2023. CA Cancer J Clin. 2023;73(1):17-48. Sekhoacha M, Riet K, Motloung P, Gumenku L, Adegoke A, Mashele S. Prostate Cancer Review: Genetics, Diagnosis, Treatment Options, and Alternative Approaches. Molecules. 2022;27(17):5730. Vickers AJ, Eastham JA, Scardino PT, Lilja H. The Memorial Sloan Kettering Cancer Center Recommendations for Prostate Cancer Screening. Urology. 2016;91:12-18. Epstein JI, Egevad L, Amin MB, et al. The 2014 International Society of Urological Pathology (ISUP) Consensus Conference on Gleason Grading of Prostatic Carcinoma: Definition of Grading Patterns and Proposal for a New Grading System. Am J Surg Pathol. 2016;40(2):244-252. Iczkowski KA, van Leenders GJLH, van der Kwast TH. The 2019 International Society of Urological Pathology (ISUP) Consensus Conference on Grading of Prostatic Carcinoma. Am J Surg Pathol. 2021;45(7):1007. Offermann A, Hohensteiner S, Kuempers C, et al. Prognostic Value of the New Prostate Cancer International Society of Urological Pathology Grade Groups. Front Med (Lausanne). 2017;4:157. Zhou L, Xu LL, Zheng LL, et al. Predictors of Gleason Grading Group Upgrading in Low-Risk Prostate Cancer Patients From Transperineal Biopsy After Radical Prostatectomy. Acad Radiol. 2024;31(7):2838-2847. Cheng HH, Mitchell PS, Kroh EM, et al. Circulating microRNA profiling identifies a subset of metastatic prostate cancer patients with evidence of cancer-associated hypoxia. PLoS One. 2013;8(7):e69239. Mathieu R, Castelli C, Fardoun T, et al. Cost analysis of prostate cancer detection including the prostate health index (phi). World J Urol. 2019;37(3):481-487. Fütterer JJ, Tempany C. Prostate MRI and image quality: The radiologist's perspective. Eur J Radiol. 2023;165:110930. Pallwein L, Mitterberger M, Gradl J, et al. Value of contrast-enhanced ultrasound and elastography in imaging of prostate cancer. Curr Opin Urol. 2007;17(1):39-47. Dong Y, Wang P, Geng H, Liu Y, Wang E. Ultrasound and advanced imaging techniques in prostate cancer diagnosis: A comparative study of mpMRI, TRUS, and PET/CT. J Xray Sci Technol. 2025;33(2):436-447. Ai H, Huang Y, Tai DI, Tsui PH, Zhou Z. Ultrasonic Assessment of Liver Fibrosis Using One-Dimensional Convolutional Neural Networks Based on Frequency Spectra of Radiofrequency Signals with Deep Learning Segmentation of Liver Regions in B-Mode Images: A Feasibility Study. Sensors (Basel). 2024;24(17):5513. Zhang Y, Meng Z, Xu Y, et al. Real-Time Elastography in the diagnosis of prostate cancer: a systematic review. Med Ultrason. 2019; 21(3):327-335. Yu A, Yamany T, Mojtahed A, et al. Combination MRI-targeted and systematic prostate biopsy may overestimate gleason grade on final surgical pathology and impact risk stratification. Urol Oncol. 2022;40(2):59.e1-59.e5. Singh V, Sharma K, Singh M, et al. Discrepancies in Gleason score between needle core biopsy and radical prostatectomy specimens with correlation between clinical and pathological staging. Urologia. 2024;91(3):518-524. Sanda MG, Cadeddu JA, Kirkby E, et al. Clinically Localized Prostate Cancer: AUA/ASTRO/SUO Guideline. Part I: Risk Stratification, Shared Decision Making, and Care Options. J Urol. 2018;199(3):683-690. Cornford P, van den Bergh RCN, Briers E, et al. EAU-EANM-ESTRO-ESUR-SIOG Guidelines on Prostate Cancer. Part II-2020 Update: Treatment of Relapsing and Metastatic Prostate Cancer. Eur Urol. 2021;79(2):263-282. Oerther B, Nedelcu A, Engel H, et al. Update on PI-RADS Version 2.1 Diagnostic Performance Benchmarks for Prostate MRI: Systematic Review and Meta-Analysis [published correction appears in Radiology. 2024]. Radiology. 2024;312(2):e233337. Singh D, Kumar V, Das CJ, Singh A, Mehndiratta A. Machine learning-based analysis of a semi-automated PI-RADS v2.1 scoring for prostate cancer. Front Oncol. 2022; 12:961985. Turkbey B, Rosenkrantz AB, Haider MA, et al. Prostate Imaging Reporting and Data System Version 2.1: 2019 Update of Prostate Imaging Reporting and Data System Version 2. Eur Urol. 2019;76(3):340-351. Pecoraro M, Dehghanpour A, Das JP, Woo S, Panebianco V. Evaluation of Prostate Cancer Recurrence with MR Imaging and Prostate Imaging for Recurrence Reporting Scoring System. Radiol Clin North Am. 2024;62(1):135-159. Ponsiglione A, Gambardella M, Stanzione A, et al. Radiomics for the identification of extraprostatic extension with prostate MRI: a systematic review and meta-analysis. Eur Radiol. 2024;34(6):3981-3991. Ursprung S, Herrmann J, Nikolaou K, et al. Die multiparametrische MRT der Prostata: Anforderungen und Grundlagen der Befundung [Multiparametric MRI of the prostate: requirements and principles regarding diagnostic reporting]. Urologie. 2023;62(5):449-458. Mottet N, van den Bergh RCN, Briers E, et al. EAU-EANM-ESTRO-ESUR-SIOG Guidelines on Prostate Cancer-2020 Update. Part 1: Screening, Diagnosis, and Local Treatment with Curative Intent. Eur Urol. 2021;79(2):243-262. Getaneh AM, Heijnsdijk EA, de Koning HJ. Cost-effectiveness of multiparametric magnetic resonance imaging and MRI-guided biopsy in a population-based prostate cancer screening setting using a micro-simulation model. Cancer Med. 2021;10(12):4046-4053. Kanda T, Ishii K, Kawaguchi H, Kitajima K, Takenaka D. High signal intensity in the dentate nucleus and globus pallidus on unenhanced T1-weighted MR images: relationship with increasing cumulative dose of a gadolinium-based contrast material. Radiology. 2014;270(3):834-841. Dias AB, Ghai S. Prostate Cancer Diagnosis with Micro-ultrasound: What We Know now and New Horizons. Radiol Clin North Am. 2024; 62(1):189-197. Liu X, Zhou H, Xu X, et al. A scoring diagnostic system based on biparametric ultrasound features for prostate cancer risk assessment. Quant Imaging Med Surg. 2023;13(6):3703-3715. Cheng T, Li H. Prediction of Gleason score in prostate cancer patients based on radiomic features of transrectal ultrasound images. Br J Radiol. 2024;97(1154):415-421. Zhou L, Zheng Y, Yao J, Chen L, Xu D. Association between papillary thyroid carcinoma and cervical lymph node metastasis based on ultrasonic radio frequency signals. Cancer Med. 2023;12(13):14305-14316. Xiao T, Shen W, Wang Q, Wu G, Yu J, Cui L. The detection of prostate cancer based on ultrasound RF signal. Front Oncol.2022; 12:946965. Wang Q, Jia X, Luo T, Yu J, Xia S. Deep learning algorithm using bispectrum analysis energy feature maps based on ultrasound radiofrequency signals to detect breast cancer. Front Oncol. 2023;13:1272427. Mutala TM, Mwango GN, Aywak A, Cioni D, Neri E. Determining the elastography strain ratio cut off value for differentiating benign from malignant breast lesions: systematic review and meta-analysis. Cancer Imaging. 2022;22(1):12. Wang Z, Liu H, Zhu Q, Chen J, Zhao J, Zeng H. Analysis of the immune-inflammatory indices for patients with metastatic hormone-sensitive and castration-resistant prostate cancer. BMC Cancer. 2024;24(1):817. Barr RG, Cosgrove D, Brock M, et al. WFUMB Guidelines and Recommendations on the Clinical Use of Ultrasound Elastography: Part 5. Prostate. Ultrasound Med Biol. 2017;43(1):27-48. Abedi M, Sahebi L, Eslami B, et al. Using a combination of superb microvascular imaging and other auxiliary ultrasound techniques to increase the accuracy of gray-scale ultrasound for breast masses. BMC Cancer. 2024;24(1):224. Sidhu PS, Cantisani V, Dietrich CF, et al. The EFSUMB Guidelines and Recommendations for the Clinical Practice of Contrast-Enhanced Ultrasound (CEUS) in Non-Hepatic Applications: Update 2017 (Long Version) [Die EFSUMB-Leitlinien und Empfehlungen für den klinischen Einsatz des kontrastverstärkten Ultraschalls (CEUS) bei nicht-hepatischen Anwendungen: Update 2017 (Langversion)]. Ultraschall Med. 2018;39(2):e2-e44. Wang B, Yang D, Zhang X, et al. The diagnostic value of contrast-enhanced ultrasonography in breast ductal abnormalities. Cancer Imaging. 2023; 23(1):25. Tables ‌Table 1. Association between all predictive parameters and Gleason Grade (low, medium, high) H(high) L(low) M(medium) P-value TPSA 302.46±571.91 174.62±508.47 48.58±76.22 0.000 FPSA 18.43±17.77 9.25±13.89 5.09±6.69 0.000 FPSA/TPSA 0.12±0.07 0.13±0.06 0.14±0.07 0.305 NLR 2.71±1.28 2.71±1.39 2.73±1.15 0.731 ELR 0.09±0.08 0.09±0.07 0.09±0.07 0.863 LMR 3.79±1.56 3.85±1.91 3.65±1.43 0.862 PLR 136.92±65.30 124.60±55.39 130.85±52.26 0.448 neutrophilic_granulocyte 4.13±1.51 4.08±1.28 4.18±1.06 0.624 lymphocyte 1.67±0.51 1.68±0.56 1.69±0.57 0.957 eosinophil 0.14±0.11 0.14±0.10 0.14±0.10 0.884 monocyte 0.49±0.20 0.52±0.27 0.50±0.18 0.778 thrombocyte 206.12±62.75 189.31±57.66 201.37±54.53 0.055 age 71.38±8.29 70.46±8.10 70.10±8.75 0.552 BMI 23.75±2.64 23.58±3.08 23.75±2.93 0.542 maximum_diameter 18.79±9.03 17.70±7.72 17.52±7.08 0.794 integrity_of_envelope 1.13±0.34 1.12±0.33 1.11±0.32 0.932 tumor_echo 1.16±0.21 1.22±0.34 1.18±0.26 0.879 prostate_volume 53.45±53.18 52.44±19.74 21.29±18.05 0.930 prostate_shape 1.13±1.16 1.15±0.34 0.36±0.36 0.883 inside_echo 1.28±1.16 1.23±0.45 0.36±0.42 0.093 tumor_shape 1.12±1.14 1.13±0.33 0.34±0.34 0.933 tumor_edge 1.17±1.11 1.12±0.38 0.31±0.33 0.413 elastic_score 3.70±3.08 3.43±0.98 1.01±0.97 0.000 elastic_ratio 8.53±5.80 6.44±1.25 2.18±0.73 0.000 Higuchi 2.67±2.66 2.65±0.13 0.12±0.14 0.635 Slope 0.68±0.70 0.69±0.15 0.15±0.16 0.855 Intercept 0.85±0.69 0.74±0.05 0.17±0.07 0.000 Mid_band 0.69±0.60 0.65±0.07 0.10±0.09 0.000 S1 0.21±0.16 0.18±0.01 0.04±0.01 0.000 S2 0.12±0.12 0.13±0.02 0.02±0.03 0.187 S3 0.09±0.09 0.09±0.04 0.04±0.04 0.649 S4 0.02±0.05 0.04±0.01 0.02±0.01 0.000 TtoPK 14.25±19.30 18.25±1.68 3.83±1.11 0.000 Area 598.41±584.74 581.18±77.59 95.21±62.54 0.102 Grad 1.82±1.26 1.54±0.10 0.41±1.49 0.000 Atm 14.39±14.47 15.14±1.49 1.57±1.88 0.002 PI 25.70±22.74 24.28±1.09 2.33±1.02 0.000 Table 2. Predictive performance of all models Model Accuracy Recall F1 DecisionTree 0.818 0.818 0.816 RandomForest 0.820 0.820 0.820 KNeighbors 0.788 0.788 0.801 GBDT 0.848 0.848 0.849 Xgboost 0.818 0.789 0.796 Naive_bayes 0.773 0.773 0.779 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6998221","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":495246413,"identity":"b20e7fe7-7520-41c3-a2e7-cf30070fa7f9","order_by":0,"name":"Wenlong Xie","email":"","orcid":"","institution":"First Affiliated Hospital of Dalian Medical University","correspondingAuthor":false,"prefix":"","firstName":"Wenlong","middleName":"","lastName":"Xie","suffix":""},{"id":495246419,"identity":"8053586d-4f6e-4aff-89ac-c9c5bdc2edbf","order_by":1,"name":"GuangZhen Wu","email":"","orcid":"","institution":"First Affiliated Hospital of Dalian Medical University","correspondingAuthor":false,"prefix":"","firstName":"GuangZhen","middleName":"","lastName":"Wu","suffix":""},{"id":495246421,"identity":"e5e2b999-e9f1-4b87-a3e1-21dbaca46572","order_by":2,"name":"XiaoChen Qi","email":"","orcid":"","institution":"First Affiliated Hospital of Dalian Medical University","correspondingAuthor":false,"prefix":"","firstName":"XiaoChen","middleName":"","lastName":"Qi","suffix":""},{"id":495246423,"identity":"20b0e1ff-67d0-4563-8ea5-21bda947ba4a","order_by":3,"name":"Lin Zhong","email":"","orcid":"","institution":"First Affiliated Hospital of Dalian Medical University","correspondingAuthor":false,"prefix":"","firstName":"Lin","middleName":"","lastName":"Zhong","suffix":""},{"id":495246429,"identity":"4e6e2e67-d1e5-4434-b750-29dc14449b5a","order_by":4,"name":"LiYing Guo","email":"","orcid":"","institution":"First Affiliated Hospital of Dalian Medical University","correspondingAuthor":false,"prefix":"","firstName":"LiYing","middleName":"","lastName":"Guo","suffix":""},{"id":495246431,"identity":"054a5baa-c70c-437c-bffa-20c5a98dc7f0","order_by":5,"name":"MengYing Tong","email":"","orcid":"","institution":"First Affiliated Hospital of Dalian Medical University","correspondingAuthor":false,"prefix":"","firstName":"MengYing","middleName":"","lastName":"Tong","suffix":""},{"id":495246432,"identity":"c3974b90-7f79-4ae3-878d-9b3bc018fc7a","order_by":6,"name":"Ying Che","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAzklEQVRIiWNgGAWjYHACAwaGAgkeNvbmAwc+VBCtxcBGho/nWOLBGWeI15JmIyeRY3yYt4UI9brtzdukeQwO87BJ5Hw4wNvAIM8vdgC/FrMzx4qNwVp43m44ILmDwXDm7AQCWm7kGD7OAWlhz91wwPAMQ4LBbUJa7r8xOAzWwpDz4EBiGzFabvCAbEnjYePIYThwkCgtZ9KKjf8Y2AD9cszgYMMZCSL8cvzwNskZFRL28u3Njz//qbCR55cmoAUdSJCmfBSMglEwCkYBdgAA8M1GEjuhUKEAAAAASUVORK5CYII=","orcid":"","institution":"First Affiliated Hospital of Dalian Medical University","correspondingAuthor":true,"prefix":"","firstName":"Ying","middleName":"","lastName":"Che","suffix":""}],"badges":[],"createdAt":"2025-06-28 13:53:17","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6998221/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6998221/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":88492462,"identity":"4ec58e2e-6abb-4456-aa7a-018b0c32426b","added_by":"auto","created_at":"2025-08-07 04:32:40","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":512294,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003e(A) \u003c/strong\u003eTransrectal ultrasound (TRUS) B-mode grayscale image of the prostate.The white arrow demarcates a hypoechoic lesion within the right prostatic lobe, demonstrating irregular morphology and ill-defined margins. The white triangle denotes the urinary bladder. \u003cstrong\u003e(B) \u003c/strong\u003eElastographic score image of the prostate. Within the region of interest (ROI), red denotes soft tissue, green indicates moderate stiffness, and blue corresponds to hard tissue. The lesion area marked by the white arrow exhibits predominantly blue coloration, demonstrating high stiffness consistent with an elasticity score of 5.\u003cstrong\u003e (C) \u003c/strong\u003eElasticity ratio analysis of prostatic stiffness.\u003cstrong\u003e \u003c/strong\u003eA yellow dashed line ROI was placed within the lesion area to measure the stiffness value (E1). Another ROI of identical depth was positioned in adjacent normal prostatic tissue to quantify baseline stiffness (E2). The elasticity ratio (E1/E2) was calculated as 5.3/0.6 = 8.83, indicating significantly elevated stiffness in the lesion compared to normal tissue.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6998221/v1/254d50c2d78d7bb4f879d069.png"},{"id":88490906,"identity":"14557418-2943-43a0-a1ff-1648cdf0b982","added_by":"auto","created_at":"2025-08-07 04:16:40","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":399075,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003e(A) \u003c/strong\u003eContrast-enhanced ultrasound (CEUS) image. The lesion area, delineated by a yellow dashed line, demonstrates marked hyperenhancement with earlier arterial phase enhancement compared to the surrounding prostatic parenchyma. \u003cstrong\u003e(B) \u003c/strong\u003eTime-intensity curve (TIC) under the condition of CEUS. After selecting ROIs, quantitative parameters were automatically generated. The yellow curve represents the hemodynamic characteristics of the prostatic lesion, while the green curve corresponds to adjacent normal prostatic parenchyma.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6998221/v1/bb7328bb2a0f7d98240de14a.png"},{"id":88490907,"identity":"393505a1-e335-4c12-afc5-67960598b24b","added_by":"auto","created_at":"2025-08-07 04:16:40","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":327434,"visible":true,"origin":"","legend":"\u003cp\u003eAnalytical results of exported raw radiofrequency (RF) signals processed via MATLAB software. After demarcating the ROI with a red rectangle in the upper-left panel, the remaining three panels were automatically generated. These images quantify tissue characteristics through multidimensional perspectives derived from raw signal parameters, including structural complexity, attenuation properties, scattering intensity, spectral specificity, amplitude, and phase. This approach provides comprehensive multidimensional Image characteristics within the ROI, a capability not inherent in conventional B-mode grayscale ultrasonography. MATLAB then automatically performs the Fourier transform and normalization of the RF signal. Using the Higuchi algorithm, the Higuchi fractal dimension was calculated. The normalized power spectrum of the RF signals within the ROI (blue curve in the lower-right panel) was analyzed via linear regression fitting (red line in the lower-right panel), yielding three key parameters: slope, intercept, and mid-band frequency. Parameters S1–S4 denote the mean integrated values derived from quartile segmentation of the normalized frequency spectrum, corresponding to four equal-bandwidth partitions across the analyzed spectral range.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6998221/v1/36612416dda6712599ee499d.png"},{"id":88491788,"identity":"61381c6c-448c-4c71-bef2-d4be03cb8a5a","added_by":"auto","created_at":"2025-08-07 04:24:40","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":108038,"visible":true,"origin":"","legend":"\u003cp\u003eHorizontal bar chart of feature importance scores in prostate cancer prediction. The x-axis represents the importance score, with higher values indicating greater impact of the feature on the predictive model. As demonstrated, Grad ranked first in importance with a score of 0.199, highlighting its substantial contribution to the model's prediction of prostate cancer-related outcomes. In contrast, features such as tumor_edge exhibited scores approaching zero, reflecting minimal clinical relevance in this context.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-6998221/v1/177d7a5991fcf816a9c70360.png"},{"id":88491787,"identity":"6479b45e-a585-4f72-a5b7-bb42b1c0f155","added_by":"auto","created_at":"2025-08-07 04:24:40","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":70444,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP value distributions across different categories using the FastSHAP algorithm. The plot depicts the contribution levels of various features to different grades of prostate cancer.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-6998221/v1/07340f0212b5137646607424.png"},{"id":88491793,"identity":"9f85a210-fae2-423a-90d4-5dc74baadc6d","added_by":"auto","created_at":"2025-08-07 04:24:41","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":77513,"visible":true,"origin":"","legend":"\u003cp\u003eVisualization of decision tree model. The model is structured as a hierarchical tree comprising nodes and branches. The uppermost node, termed the root node, initiates a sequence of conditional splits that generate child nodes through progressive branching until reaching terminal leaf nodes. The leaf nodes ultimately provide the final classification outcomes\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-6998221/v1/741f13c3660c9f6fe0a2a146.png"},{"id":88492464,"identity":"88506d17-1726-413a-9865-265080ae94b5","added_by":"auto","created_at":"2025-08-07 04:32:41","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":67852,"visible":true,"origin":"","legend":"\u003cp\u003eConfusion matrix used to evaluate the performance of a prediction model.The rows represent the true label, while the columns represent the predicted label. The classification accuracy of the model in various categories can be more intuitively understood through this matrix.\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-6998221/v1/b60dfdfef253dba9a771a777.png"},{"id":88491790,"identity":"93919061-d28a-456e-bcc7-4d9bc070891e","added_by":"auto","created_at":"2025-08-07 04:24:40","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":61603,"visible":true,"origin":"","legend":"\u003cp\u003eParallel coordinates plot comparing the overall performance of all predictive models.\u003cstrong\u003e \u003c/strong\u003eGBDT model occupied the uppermost trajectory in the parallel coordinates plot, achieving the highest composite score across all models.\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-6998221/v1/9230056d0270c35af62a81ae.png"},{"id":91148628,"identity":"ad447d9d-d6ef-4f3e-a28e-953ac1fd9651","added_by":"auto","created_at":"2025-09-12 06:45:10","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2448502,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6998221/v1/9827a159-0244-4a5b-bf61-3bbbd2a21865.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"An Innovative Approach for Predicting Prostate Cancer Gleason Grading: Machine Learning-based Fusion of Multimodal Ultrasound, Clinical and Laboratory Indicators ","fulltext":[{"header":"Introduction","content":"\u003cp\u003eProstate cancer remains the second most prevalent malignancy and the fifth leading cause of cancer-related mortality among men worldwide. The Global Cancer Statistics report estimates 1.4\u0026nbsp;million new cases and 375,000 deaths attributed to prostate cancer annually (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). Contemporary epidemiological studies attribute the escalating incidence to accelerated population aging and widespread implementation of prostate-specific antigen (PSA) screening protocols (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). Despite therapeutic advancements, the 5-year survival rate for metastatic prostate cancer persists below 30% (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e), underscoring the critical need for early lesion detection and precise risk stratification. Accurate histological grading enables clinicians to formulate personalized treatment strategies, thereby mitigating overtreatment risks while optimizing oncological outcomes (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe Gleason grading system serves as the cornerstone histopathological framework for evaluating prostate cancer aggressiveness, stratifying tumor biological behavior through meticulous analysis of glandular architectural patterns in hematoxylin and eosin (H\u0026amp;E)-stained specimens. Originally conceptualized by Dr. Donald Gleason in 1966, this classification has undergone two seminal revisions under the auspices of the International Society of Urological Pathology in 2014 and 2019 (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). These updates have culminated in the contemporary five-tier Grade Group system, ranging from Grade Group 1 (well-differentiated) to Grade Group 5 (poorly differentiated). Extensive clinical evidence supports the prognostic value of the Gleason grade groups to metastatic risk (hazard ratio\u0026thinsp;=\u0026thinsp;3.21, 95%CI: 2.74\u0026ndash;3.76), guiding therapeutic pathway selection, and influencing overall survival outcomes (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eWhile prostate biopsy remains the gold standard for prostate cancer diagnosis, emerging evidence suggests discrepancies between biopsy-derived Gleason grades and those identified in postoperative histopathological specimens, potentially leading to misclassification of disease aggressiveness and suboptimal therapeutic decisions (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e). To improve the accuracy of Gleason grade assessment, novel noninvasive methodologies have been increasingly explored alongside advancements in diagnostic technologies. A multicenter study conducted in 2022 identified circulating miR-141-3p as a promising biomarker for predicting Gleason grade progression during active surveillance, demonstrating 82% sensitivity in serial plasma analyses (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e). Loeb et al. further advanced serum-based risk stratification by establishing that combined prostate-specific antigen density (\u0026gt;\u0026thinsp;0.15 ng/mL/cm\u0026sup3;) and Prostate Health Index (\u0026ge;\u0026thinsp;35) achieved 77% specificity in discriminating high-grade carcinomas (Gleason score\u0026thinsp;\u0026ge;\u0026thinsp;7), outperforming conventional PSA (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eMultiparametric magnetic resonance imaging (mpMRI), which integrates T2-weighted imaging, diffusion-weighted imaging (DWI), and dynamic contrast-enhanced sequences, has standardized lesion characterization. Using the PI-RADS v2.1 criteria (Prostate Imaging Reporting and Data System version 2.1), mpMRI achieves a negative predictive value exceeding 90% for excluding high-grade tumors (Gleason score\u0026thinsp;\u0026ge;\u0026thinsp;4\u0026thinsp;+\u0026thinsp;3) (\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e). Pallwein et al. demonstrated that combining hemodynamic parameters of contra-enhanced ultrasound with quantitative elasticity measurements can significantly improve the sensitivity of prostate cancer detection while predicting lesions with higher Gleason grades (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e). Correas et al. utilized ultrasound elastography to quantify tissue stiffness differences, demonstrating 85% sensitivity in distinguishing low-grade (Gleason 6) from high-grade (Gleason\u0026thinsp;\u0026ge;\u0026thinsp;7) tumors (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eSince most current prostate biopsies are performed under ultrasound guidance, this study focuses on combining ultrasound multimodal analysis with clinical and laboratory indicators to predict prostate cancer Gleason grade groups. This approach provides two key advantages, including enabling noninvasive prediction of tumor aggressiveness and offering real-time guidance during biopsy procedures to help clinicians identify the primary tumor location. By directing biopsy needles to areas most likely containing the highest Gleason grade lesions, our approach effectively minimizes discrepancies between preoperative biopsy results and final Gleason grade determinations from radical prostatectomy specimens.\u003c/p\u003e\u003cp\u003eThis study employs a novel multimodal ultrasound protocol comprising: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) dual-plane transrectal probe for ultrasonographic image characterization, (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) Elastography for tissue stiffness quantification, (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) contrast-enhanced ultrasound (CEUS) for vascular pattern analysis, and (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) ultrasound radiofrequency (RF) signal processing of raw data from diagnostic scanners. The RF signal analysis specifically utilizes unprocessed backscatter signals (frequency range: 2\u0026ndash;10 MHz) to objectively quantify intrinsic tissue properties through spectral parameter extraction, circumventing potential artifacts introduced by conventional post-processing algorithms (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eTo date, no academic research has explored the use of multimodal ultrasound in combination with clinical indicators to predict prostate cancer Gleason grading. This study aimed to establish a noninvasive predictive model for prostate cancer Gleason grading, providing more comprehensive diagnostic evidence for clinical practice.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eData Collection\u003c/h2\u003e\u003cp\u003e This single-center prospective study was conducted at the First Hospital affiliated to Dalian Medical University and enrolled 329 prostate cancer patients meeting predefined inclusion criteria between February 2024 and May 2025. This study was approved by the Ethics Committee of the The First Affiliated Hospital of Dalian Medical University. All participants provided written informed consent prior to enrollment and data collection. All procedures were conducted in accordance with the ethical standards of the institutional research committee and the Declaration of Helsinki. All participants underwent prostate biopsies followed by Gleason grading and were divided into three groups: low-grade (n\u0026thinsp;=\u0026thinsp;147, Gleason score\u0026thinsp;\u0026le;\u0026thinsp;6), intermediate-grade (n\u0026thinsp;=\u0026thinsp;106, Gleason score 7), and high-grade (n\u0026thinsp;=\u0026thinsp;76, Gleason score\u0026thinsp;\u0026ge;\u0026thinsp;8). Patient characteristics were divided into two categories: clinical and laboratory data and multimodal ultrasonic data. Multimodal ultrasound data were independently collected under double-blind conditions by two physicians with more than 10 years of experience in ultrasound diagnostic, and the measurements were averaged before being included in this study.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eSample Inclusion Criteria\u003c/h3\u003e\n\u003cp\u003eThe inclusion criteria were as follows: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) the patient was suspected of prostate cancer and agreed to undergo prostate biopsy at our hospital, and was diagnosed with prostate cancer; (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) ultrasound imaging showed clearly defined nodules, and the patient had no known allergic reactions (e.g., protein allergies) to ultrasound contrast agent; (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) patients agreed to undergo multimodal ultrasound examination and other relevant contents and provided written informed consent; (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) no history of other tumors; (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e) initial diagnostic examination for prostate cancer without endocrine therapy or radiotherapy.\u003c/p\u003e\n\u003ch3\u003eSample Exclusion Criteria\u003c/h3\u003e\n\u003cp\u003eThe exclusion criteria included the following: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) patients with coagulation disorders; (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) patients unable to cooperate with prostate biopsy for other reasons, such as severe cardiac dysfunction, anal surgery, and mental illness; (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) incomplete clinical data; (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) history of prostate surgery (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e) images were unclear or the internal nodules of the prostate could not be accurately located; (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e) unable to collect the complete data required for the study.\u003c/p\u003e\n\u003ch3\u003eClinical and Laboratory Data\u003c/h3\u003e\n\u003cp\u003eThe collected clinical and laboratory parameters included age, body mass index (BMI), total prostate-specific antigen (tPSA), free PSA (fPSA), fPSA/tPSA ratio, neutrophil count, lymphocyte count, eosinophil count, monocyte count, platelet count, neutrophil-to-lymphocyte ratio (NLR), eosinophil-to-lymphocyte ratio (ELR), lymphocyte-to-monocyte ratio (LMR), and platelet-to-lymphocyte ratio (PLR).\u003c/p\u003e\n\u003ch3\u003eMultimodal Ultrasonic Data\u003c/h3\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003eTransrectal Two-dimensional Ultrasound Data\u003c/h2\u003e\u003cp\u003eThis study used two-dimensional transrectal ultrasound measurements. Patient were positioned in the left lateral position with both knees flexed to the chest. After the surface of the probe was evenly coated with sterile coupling gel, an isolation sleeve was placed on the probe before placing the probe at the level of the prostate. The gain was 60\u0026ndash;70 dB, and the depth was adjusted to fully display the outline of the prostate (usually 8\u0026ndash;12 cm) to ensure that clear and standard transverse and sagittal images could be obtained (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA). The following two-dimensional ultrasound parameters were collected: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) prostate volume, (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) regularity of prostate shape, (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) uniformity of the internal echogenicity of the prostate, (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) integrity of the prostate capsule, (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e) tumor shape, (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e) tumor boundary, (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e) maximum diameter of the tumor, and (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e) tumor echogenicity.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eElastography Data\u003c/h3\u003e\n\u003cp\u003eStrain elastography mode (sampling frame depth 40\u0026thinsp;\u0026plusmn;\u0026thinsp;5 mm) was performed using the same inspection mode. Operators followed a real-time pressure feedback indicator to maintain stable pressure during the procedure. The parameters recorded were: elastic score and elastic ratio. The modified 5-point method was used for elasticity scoring as follows (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e): one point indicated the whole lesion was green (strain rate\u0026thinsp;\u0026ge;\u0026thinsp;80%); Two points represented a mixture of blue and green mixed, with blue accounting for \u0026lt;\u0026thinsp;25%; Three points indicated blue accounting for 25\u0026ndash;50%; Four points indicated blue accounting for 50\u0026ndash;75%; and five points indicated blue accounting for \u0026gt;\u0026thinsp;75% or with rear sound shadow attenuation (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB). The elastic ratio was measured by manually selecting the regions of interest (ROI), and the average strain ratio of the target lesion to the normal gland was measured (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eC).\u003c/p\u003e\n\u003ch3\u003eContrast Ultrasound Data\u003c/h3\u003e\n\u003cp\u003e Contrast material was administered using SonoVue (Bracco Imaging), a sulfur hexafluoride microbubble formulation, through an antecubital vein by bolus injection of a 19-g trocar (dose, 2.4 mL), followed by flushing of 5 mL of normal saline (0.9% NaCl). The system was switched to CEUS mode with a mechanical index of 0.10\u0026thinsp;\u0026plusmn;\u0026thinsp;0.02. The time-intensity curve of the lesion area was drawn using the Q-Lab software, and the corresponding parameters were recorded (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Parameters included: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) Time to peak (Ttopk), (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) Area under the curve (Area), (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) Contrast agent onset to peak gradient (Grad), (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) Contrast agent arrival time (ATm), and (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e) Peak intensity (PI).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003eRadio-frequency Signal Data\u003c/h2\u003e\u003cp\u003eThe system was configured in second harmonic mode for RF signal acquisition, with the transmitter frequency set at 5 MHz and the receiver frequency adjusted to 10 MHz. After collection, the RF signal (sampling rate 40 MHz) was exported through the original data port of the ultrasound system. Post-processing was performed using MATLAB (version R2019a.V96; MathWorks Inc.), which automatically generated eight quantitative parameters (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). The parameters of ultrasound RF signal included: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) Fractal dimension (Higuchi), which quantifies the complexity and irregularity of signal or tissue structure; (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) Slope, which is used to analyze the attenuation characteristics of tissue to ultrasound; (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) Intercept, which reflects the overall strength of the echo signal; (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) Mid-band parameter, combined with the slope and intercept, represents the average attenuation slope in the frequency band.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eAdditionally, the spectrum of the RF signal was split into four equal frequency bands: Low frequency bands (S1), Medium-low frequency bands (S2), Medium-high frequency bands (S3), and High frequency bands (S4). These parameters represent the attenuation coefficients in different frequency bands and quantify tissue properties from different perspectives, such as structural complexity, attenuation properties, scattering intensity, and frequency band specificity, providing multidimensional information for noninvasive diagnosis.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003eData Analysis\u003c/h2\u003e\u003cp\u003eData were analyzed using Python version 3.12.4, with multiple Python libraries applied for data processing, model construction, and evaluation. Scikit-learn provided foundational support for machine learning algorithms. Pandas and NumPy were employed for data manipulation and analysis. Key computational models included: DecisionTreeClassifier (Decision Tree), RandomForestClassifier (Random Forest), GradientBoostingClassifier (Gradient Boosting Decision Trees, GBDT), KNeighborsClassifier (K-Nearest Neighbors, KNN), XGBoost (Extreme Gradient Boosting), and naive_bayes (Na\u0026iuml;ve Bayes). The metrics module of scikit-learn were used to calculate accuracy, recall, and F1 values to evaluate model performance. Results were visualized using the matplotlib library.\u003c/p\u003e\u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eDevelopment of a Prediction Model\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDuring the feature selection phase, non-parametric tests were first applied to assess the relationship between the dependent variable (prostate cancer Gleason grading) and predictive parameters. Results were summarized in a three-line table (Table 1). The following parameters exhibited significant associations with Gleason grading (P \u0026lt; 0.05): TPSA, FPSA, Intercept, S1, TtoPK, Grad, PI, elastic_score, elastic_ratio, Mid_band, S4, and Atm. Both the random forest Gini impurity algorithm and FastSHAP algorithm were employed to calculate feature importance scores and contribution values. The relative importance rankings of these features are illustrated in Figures 4 and 5.\u003c/p\u003e\n\u003cp\u003eBased on the non-parametric test results, metric importance evaluation, and SHAP value contribution analysis, we ultimately selected five predictive indicators (Grad, TtoPK, S1, Intercept, and PI) as characteristic features for assessing prostate cancer Gleason grade. Utilizing these indicators, we developed six machine learning prediction models, including Decision Tree, Random Forest, KNN, XGBoost, GBDT, and Naïve Bayes classifiers for outcome prediction. The data was divided into the training set and the test set in a 4:1 ratio during the construction of the prediction model.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eModel Predictive Performance\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe decision tree model visualization is shown below Figure 6. Decision Tree, Random Forest, KNN, XGBoost, GBDT, and Naïve Bayes model visualization are shown below Figure 7. To evaluate model performance, the accuracy rate, recall rate and F1 value of each model were calculated (Table 2 and Figure 8). Comprehensive comparative analysis of model performance metrics revealed that the GBDT model achieved close to 85% accuracy, with superior F1-score and recall rates compared to other models, establishing it as the optimal predictive model for Gleason grade classification in prostate cancer.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003eOverview of Current Research\u003c/h2\u003e\u003cp\u003eProstate cancer has a high incidence rate and significantly impacts the health of middle-aged and elderly males. Owing to its histological characteristics, the tumor frequently harbors coexisting cancer foci with varying degrees of differentiation (\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e). The differentiation status of these lesions, which is currently mainly distinguished by Gleason grading, directly determines clinical treatment strategies (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e). Therefore, accurate assessment of the Gleason grade is critical in prostate cancer management.\u003c/p\u003e\u003cp\u003eMost studies have demonstrated discrepancies between pathological findings from prostate biopsy and subsequent radical prostatectomy specimens, primarily manifested as lower Gleason grade group assignments in biopsy results compared to final surgical pathology (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e). Such underestimation can mislead clinical judgement of disease severity and lead to inappropriate therapeutic decision-making\u0026zwnj; (19).\u003c/p\u003e\u003cp\u003eCompared to other imaging modalities, ultrasound examination offers several advantages, including speed of assessment, absence of ionizing radiation, and affordability, rendering it an indispensable modality for prostate cancer screening. Furthermore, transrectal ultrasound remains the primary method for guiding prostate biopsies. Consequently, identifying sonographic characteristics associated with high-Gleason-score lesions within the prostate holds dual clinical significance, including enabling noninvasive prediction of Gleason grades and improving biopsy accuracy by allowing precise targeting of high-grade lesions. This approach enhances diagnostic concordance between biopsy results and final postoperative histopathology.\u003c/p\u003e\u003cp\u003eHistorically, imaging research on prostate cancer has predominantly focused on MRI for both diagnosis and Gleason grade determination. The advent of mpMRI and the continuous refinement of the PI-RADS v2.1 have significantly enhanced diagnostic precision for prostate cancer (\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e). Recent studies indicate that mpMRI demonstrates a high negative predictive value (NPV\u0026thinsp;\u0026gt;\u0026thinsp;90%) for excluding clinically significant high-grade carcinomas (Gleason score\u0026thinsp;\u0026ge;\u0026thinsp;4\u0026thinsp;+\u0026thinsp;3) (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e). In their study, Pecoraro M et al. found that poorly differentiated tumors are associated with restricted diffusion, establishing a statistically significant correlation between reduced apparent diffusion coefficient (ADC) values and higher Gleason grades(\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e). A meta-analysis by Ponsiglione et al., encompassing 17 studies with a pooled cohort of 4,236 patients, demonstrated that radiomics-based models derived from prostate mpMRI achieved a pooled sensitivity and specificity of 82% and 79%, respectively (area under the curve [AUC]\u0026thinsp;=\u0026thinsp;0.83) for detecting extraprostatic extension (EPE). This represents a 12% improvement in specificity compared to conventional MRI assessment (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e). Despite demonstrating excellent sensitivity and specificity in prostate cancer diagnosis, MRI has several limitations: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) high equipment costs restrict its accessibility in resource-limited settings (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e); (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) moderate interobserver agreement for PI-RADS scoring (κ\u0026thinsp;=\u0026thinsp;0.50\u0026ndash;0.65), indicating persistent variability influenced by radiologists' experience (\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e); (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) contraindications such as renal insufficiency, claustrophobia, and ferromagnetic implants; (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) prolonged scan durations and substantial per-examination costs (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eUltrasonography demonstrates distinct clinical advantages that address these limitations. First, it eliminates the need for contrast agent administration, thereby avoiding the risk of nephrogenic systemic fibrosis risk associated with gadolinium-based agents (\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e). Second, the procedure is operationally simpler, making it suitable for resource-limited medical settings. Third, ultrasound allows for rapid, real-time screening and dynamic monitoring, as well as precise guidance for interventions. The ongoing technological evolution of ultrasound systems and diversification of diagnostic methodologies have established ultrasonography as an indispensable modality in the diagnostic workflow of prostatic diseases.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\u003ch2\u003eAdvantages of Ultrasound Examination\u003c/h2\u003e\u003cp\u003eIn recent years, the diagnostic methods of prostate cancer by ultrasound technology have been constantly updated. For instance, Dias et al. demonstrated that high-frequency micro-ultrasound can achieve real-time detection of EPE with 87% sensitivity and seminal vesicle invasion with 79% specificity, while maintaining a negative predictive value (NPV) of 93% for Gleason grade group upgrading (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eLiu et al. developed a biparametric ultrasound (BUS) scoring system by integrating grayscale ultrasound, doppler blood flow imaging, and CEUS features for risk stratification of clinically significant prostate cancer. Their findings revealed diagnostic performance comparable to mpMRI in both training and validation cohorts. Notably, in the low-PSA subgroup (PSA\u0026thinsp;\u0026lt;\u0026thinsp;10 ng/mL), BUS demonstrated significantly superior specificity over MRI (65% vs. 39%, respectively) (\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eCheng et al. developed a Gleason score prediction model for prostate cancer by extracting 103 radiomics features from transrectal ultrasound images of 838 patients. Using LASSO regression for feature selection, they identified 10 major radiological features. The random forest algorithm achieved an AUC of 0.770 in the testing cohort, with peak predictive performance for high-grade tumors (Gleason score\u0026thinsp;\u0026ge;\u0026thinsp;9; AUC\u0026thinsp;=\u0026thinsp;0.847). These results underscore the potential of ultrasound-based radiomics in noninvasive histopathological grading (\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eWhile most ultrasound-based studies rely on single-modality imaging, our study comprehensively incorporates multimodal ultrasonographic metrics, including primary grayscale ultrasound characteristics, CEUS perfusion parameters, elastography findings, and RF signal-derived quantitative data\u0026mdash;an approach not previously reported in existing literature. Furthermore, we integrated patients' baseline clinical characteristics with routine laboratory parameters. These readily accessible parameters are noninvasive and radiation-free, without imposing additional financial burdens, and demonstrate high reproducibility and patient acceptability. This methodology demonstrates potential for widespread clinical adoption, particularly in resource-limited settings and primary care institutions.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e\u003ch2\u003eInnovative Aspects of the Study\u003c/h2\u003e\u003cp\u003eThis study innovatively incorporated RF signals and elasticity ratio as novel parameters, a methodological advancement not previously documented in existing literature. Ultrasonic RF signals represent unprocessed raw high-frequency electrical signals directly acquired and converted by piezoelectric transducers in ultrasound probes. These signals preserve comprehensive amplitude, phase, and frequency information, establishing a robust data foundation for high-precision tissue characterization (\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e). By analyzing acquired RF signals using mathematical modeling techniques such as spectral analysis and time-frequency transformation, the physical properties of tissues can be directly acquired with reduced empirical dependence, thereby yielding more objective and credible results (\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e). The elasticity ratio, defined as the ratio of hardness between normal prostate tissue and prostate cancer lesions, mitigates confounding factors such as respiratory motion and probe compression compared to conventional elastography techniques, thereby providing a more accurate representation of the lesion's factual hardness within the prostate (\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e).\u0026zwnj; Systemic inflammatory indices, including the NLR and PLR, indirectly reflect the interplay between inflammatory responses and immunosuppression within the tumor microenvironment by quantifying the homeostasis of immune cell subsets in peripheral blood (36). In recent years, systemic inflammatory indices have emerged as pivotal biomarkers reflecting the tumor microenvironment and host immune status, demonstrating significant value in prognostic evaluation and malignancy prediction of cancers. Building upon recent research advances, this study pioneers the incorporation of multiple systemic inflammatory indices combined with multimodal ultrasound parameters for predicting prostate cancer Gleason grading.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec20\" class=\"Section2\"\u003e\u003ch2\u003eAnalysis of Results\u003c/h2\u003e\u003cp\u003eAs demonstrated in the SHAP analysis, the predictors elastic ratio and Grad exhibited strong correlations. This observation may be attributed to their close conformity to tumor biological characteristics and histopathological alterations. An increase in tumor malignancy grade is accompanied by multiple pathological processes, including increased tumor cell proliferation and density, stromal fibrosis with collagen deposition, alterations in extracellular matrix composition, and inflammatory responses accompanied by immune infiltration. Collectively, these processes contribute to enhanced tissue stiffness, a phenomenon consistent with the principle that higher Gleason grades correlate with increased tissue rigidity in prostate cancer. These findings align with those reported by Barr et al. (\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e). Furthermore, the elastic ratio adopted is inherently resistant to confounding factors such as operator-dependent compression force, pelvic floor muscle contraction, and respiratory excursion magnitude, which frequently distort absolute elasticity measurements in prior studies (\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e). Compared with previous studies, the index of elastic ratio represent a more accurate measurement of the hardness than only measuring the elastic value of the lesion. Therefore, we believe that the above content is the fundamental reason for the high correlation of elastic ratio.\u003c/p\u003e\u003cp\u003eGrad reflects the rate at which the contrast agent enters the lesion (arterial phase) until reaching peak enhancement intensity (peak phase), quantified as the slope of enhancement intensity per unit time. A higher Grad value indicates faster contrast agent microbubble filling within the lesion per unit time, suggesting elevated local blood perfusion rates (\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e). This phenomenon may be associated with tumor-related hemodynamic factors, such as arteriovenous shunting caused by invasive tumor growth, which reduces vascular resistance and facilitates rapid perfusion. The tumor demonstrates a hypervascular state characterized by abundant but architecturally disorganized neovascularization, resulting in pathologically enhanced blood supply perfusion (\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e).Consequently, higher Gleason grades are associated with greater tumor invasiveness concomitant with more pronounced disruption of glandular architecture, increased neovascular density, and elevated frequencies of arteriovenous shunts, collectively contributing to augmented Grad values. However, the model's algorithmic reliance on highly correlated parameters such as Grad and elastic ratio may introduce diagnostic bias through inadequate consideration of other confounding indicators.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec21\" class=\"Section2\"\u003e\u003ch2\u003eLimitations and Future Directions\u003c/h2\u003e\u003cp\u003eAlthough the prediction model constructed in this study, based on multimodal ultrasound imaging, clinical, and laboratory features, demonstrated certain clinical value in Gleason grade assessment, the following limitations should be acknowledged. First, this study included 329 patients with a relatively small sample size, and all data were derived from a single medical institution, which may introduce selection bias. Second, there is a lack of sufficient validation cohorts, particularly external validation. Finally, the selection of ROI during the analysis of RF signal images remains subjective and operator-dependent.\u003c/p\u003e\u003cp\u003eDespite these limitations, this study holds potential for further exploration. Future research could focus on several key areas. First, developing a comprehensive predictive system to integrate cross-domain multimodal data, such as mpMRI, PET-CT, and liquid biopsy biomarkers. Second, expanding the sample size using multicenter prospective cohort studies. Third, implementing artificial intelligence technologies such as deep learning algorithms to automate the identification of multimodal ultrasound imaging data, thereby reducing operator dependency. Finally, future experimental designs should aim to improve differentiation between histologically Gleason Grade Group 2 (3\u0026thinsp;+\u0026thinsp;4) and Grade Group 3 (4\u0026thinsp;+\u0026thinsp;3) subtypes.\u003c/p\u003e\u003c/div\u003e"},{"header":"Conclusions","content":"\u003cp\u003eIn conclusion, multimodal ultrasound reflects the characteristics of prostate cancer lesions from different perspectives, Contrast-enhanced ultrasound patterns provide a more thorough analysis of tumor vascularity, while analysis of radiofrequency signals objectively extracts information directly from the raw fundamental data. Especially by combining various modalities and using the method of machine learning analysis, the optimal prediction indicators can be found more accurately, thereby establishing a prediction model. Our study demonstrates that machine learning-based models can effectively stratify malignancy degree of prostate cancer, thereby guiding subsequent patient management.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study complied with the 1964 Helsinki Declaration and its lateramendments ethical standards, This study has been granted ethical approval by the Ethics Committee of First Affiliated Hospital of Dalian Medical University.(NO.\u0026nbsp;PJ-KS-KY-2025-184).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets generated and analyzed in this study are available from the corresponding author on reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflict of Interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWX:\u0026nbsp;Preparation, creation and/or presentation of the published work, specifically writing the initial draft (including substantive translation); XQ and GW: Development or design of methodology; creation of models; LZ: Conducting a research and investigation process, specifically performing the experiments, or data/evidence collection; LG and MT: Application of statistical, mathematical, computational, or other formal techniques to analyze or synthesize study data; YC:\u0026nbsp;Ideas; formulation or evolution of overarching research goals and aims.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot application.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eSung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin. 2021;71(3):209-249.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eSiegel RL, Miller KD, Wagle NS, et al. Cancer Statistics, 2023. CA Cancer J Clin. 2023;73(1):17-48.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eSekhoacha M, Riet K, Motloung P, Gumenku L, Adegoke A, Mashele S. Prostate Cancer Review: Genetics, Diagnosis, Treatment Options, and Alternative Approaches. Molecules. 2022;27(17):5730.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eVickers AJ, Eastham JA, Scardino PT, Lilja H. The Memorial Sloan Kettering Cancer Center Recommendations for Prostate Cancer Screening. Urology. 2016;91:12-18.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eEpstein JI, Egevad L, Amin MB, et al. The 2014 International Society of Urological Pathology (ISUP) Consensus Conference on Gleason Grading of Prostatic Carcinoma: Definition of Grading Patterns and Proposal for a New Grading System. Am J Surg Pathol. 2016;40(2):244-252.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eIczkowski KA, van Leenders GJLH, van der Kwast TH. The 2019 International Society of Urological Pathology (ISUP) Consensus Conference on Grading of Prostatic Carcinoma. Am J Surg Pathol. 2021;45(7):1007.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eOffermann A, Hohensteiner S, Kuempers C, et al. Prognostic Value of the New Prostate Cancer International Society of Urological Pathology Grade Groups. Front Med (Lausanne). 2017;4:157.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eZhou L, Xu LL, Zheng LL, et al. Predictors of Gleason Grading Group Upgrading in Low-Risk Prostate Cancer Patients From Transperineal Biopsy After Radical Prostatectomy. Acad Radiol. 2024;31(7):2838-2847.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eCheng HH, Mitchell PS, Kroh EM, et al. Circulating microRNA profiling identifies a subset of metastatic prostate cancer patients with evidence of cancer-associated hypoxia. PLoS One. 2013;8(7):e69239.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eMathieu R, Castelli C, Fardoun T, et al. Cost analysis of prostate cancer detection including the prostate health index (phi). World J Urol. 2019;37(3):481-487.\u003c/li\u003e\n \u003cli\u003eF\u0026uuml;tterer JJ, Tempany C. Prostate MRI and image quality: The radiologist\u0026apos;s perspective. Eur J Radiol. 2023;165:110930.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003ePallwein L, Mitterberger M, Gradl J, et al. Value of contrast-enhanced ultrasound and elastography in imaging of prostate cancer. Curr Opin Urol. 2007;17(1):39-47.\u003c/li\u003e\n \u003cli\u003eDong Y, Wang P, Geng H, Liu Y, Wang E. Ultrasound and advanced imaging techniques in prostate cancer diagnosis: A comparative study of mpMRI, TRUS, and PET/CT. J Xray Sci Technol. 2025;33(2):436-447.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eAi H, Huang Y, Tai DI, Tsui PH, Zhou Z. Ultrasonic Assessment of Liver Fibrosis Using One-Dimensional Convolutional Neural Networks Based on Frequency Spectra of Radiofrequency Signals with Deep Learning Segmentation of Liver Regions in B-Mode Images: A Feasibility Study. Sensors (Basel). 2024;24(17):5513.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eZhang Y, Meng Z, Xu Y, et al. Real-Time Elastography in the diagnosis of prostate cancer: a systematic review. Med Ultrason. 2019;\u0026nbsp;21(3):327-335.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eYu A, Yamany T, Mojtahed A, et al. Combination MRI-targeted and systematic prostate biopsy may overestimate gleason grade on final surgical pathology and impact risk stratification. Urol Oncol. 2022;40(2):59.e1-59.e5.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eSingh V, Sharma K, Singh M, et al. Discrepancies in Gleason score between needle core biopsy and radical prostatectomy specimens with correlation between clinical and pathological staging. Urologia. 2024;91(3):518-524.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eSanda MG, Cadeddu JA, Kirkby E, et al. Clinically Localized Prostate Cancer: AUA/ASTRO/SUO Guideline. Part I: Risk Stratification, Shared Decision Making, and Care Options. J Urol. 2018;199(3):683-690.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eCornford P, van den Bergh RCN, Briers E, et al. EAU-EANM-ESTRO-ESUR-SIOG Guidelines on Prostate Cancer. Part II-2020 Update: Treatment of Relapsing and Metastatic Prostate Cancer. Eur Urol. 2021;79(2):263-282.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eOerther B, Nedelcu A, Engel H, et al. Update on PI-RADS Version 2.1 Diagnostic Performance Benchmarks for Prostate MRI: Systematic Review and Meta-Analysis [published correction appears in Radiology. 2024]. Radiology. 2024;312(2):e233337.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eSingh D, Kumar V, Das CJ, Singh A, Mehndiratta A. Machine learning-based analysis of a semi-automated PI-RADS v2.1 scoring for prostate cancer. Front Oncol. 2022; 12:961985.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eTurkbey B, Rosenkrantz AB, Haider MA, et al. Prostate Imaging Reporting and Data System Version 2.1: 2019 Update of Prostate Imaging Reporting and Data System Version 2. Eur Urol. 2019;76(3):340-351.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003ePecoraro M, Dehghanpour A, Das JP, Woo S, Panebianco V. Evaluation of Prostate Cancer Recurrence with MR Imaging and Prostate Imaging for Recurrence Reporting Scoring System. Radiol Clin North Am. 2024;62(1):135-159.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003ePonsiglione A, Gambardella M, Stanzione A, et al. Radiomics for the identification of extraprostatic extension with prostate MRI: a systematic review and meta-analysis. Eur Radiol. 2024;34(6):3981-3991.\u003c/li\u003e\n \u003cli\u003eUrsprung S, Herrmann J, Nikolaou K, et al. Die multiparametrische MRT der Prostata: Anforderungen und Grundlagen der Befundung [Multiparametric MRI of the prostate: requirements and principles regarding diagnostic reporting]. Urologie. 2023;62(5):449-458.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eMottet N, van den Bergh RCN, Briers E, et al. EAU-EANM-ESTRO-ESUR-SIOG Guidelines on Prostate Cancer-2020 Update. Part 1: Screening, Diagnosis, and Local Treatment with Curative Intent. Eur Urol. 2021;79(2):243-262.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eGetaneh AM, Heijnsdijk EA, de Koning HJ. Cost-effectiveness of multiparametric magnetic resonance imaging and MRI-guided biopsy in a population-based prostate cancer screening setting using a micro-simulation model. Cancer Med. 2021;10(12):4046-4053.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eKanda T, Ishii K, Kawaguchi H, Kitajima K, Takenaka D. High signal intensity in the dentate nucleus and globus pallidus on unenhanced T1-weighted MR images: relationship with increasing cumulative dose of a gadolinium-based contrast material. Radiology. 2014;270(3):834-841.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eDias AB, Ghai S. Prostate Cancer Diagnosis with Micro-ultrasound: What We Know now and New Horizons. Radiol Clin North Am. 2024; 62(1):189-197.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eLiu X, Zhou H, Xu X, et al. A scoring diagnostic system based on biparametric ultrasound features for prostate cancer risk assessment. Quant Imaging Med Surg. 2023;13(6):3703-3715.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eCheng T, Li H. Prediction of Gleason score in prostate cancer patients based on radiomic features of transrectal ultrasound images. Br J Radiol. 2024;97(1154):415-421.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eZhou L, Zheng Y, Yao J, Chen L, Xu D. Association between papillary thyroid carcinoma and cervical lymph node metastasis based on ultrasonic radio frequency signals. Cancer Med. 2023;12(13):14305-14316.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eXiao T, Shen W, Wang Q, Wu G, Yu J, Cui L. The detection of prostate cancer based on ultrasound RF signal. Front Oncol.2022; 12:946965.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eWang Q, Jia X, Luo T, Yu J, Xia S. Deep learning algorithm using bispectrum analysis energy feature maps based on ultrasound radiofrequency signals to detect breast cancer. Front Oncol. 2023;13:1272427.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eMutala TM, Mwango GN, Aywak A, Cioni D, Neri E. Determining the elastography strain ratio cut off value for differentiating benign from malignant breast lesions: systematic review and meta-analysis. Cancer Imaging. 2022;22(1):12.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eWang Z, Liu H, Zhu Q, Chen J, Zhao J, Zeng H. Analysis of the immune-inflammatory indices for patients with metastatic hormone-sensitive and castration-resistant prostate cancer. BMC Cancer. 2024;24(1):817.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eBarr RG, Cosgrove D, Brock M, et al. WFUMB Guidelines and Recommendations on the Clinical Use of Ultrasound Elastography: Part 5. Prostate. Ultrasound Med Biol. 2017;43(1):27-48.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eAbedi M, Sahebi L, Eslami B, et al. Using a combination of superb microvascular imaging and other auxiliary ultrasound techniques to increase the accuracy of gray-scale ultrasound for breast masses. BMC Cancer. 2024;24(1):224.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eSidhu PS, Cantisani V, Dietrich CF, et al. The EFSUMB Guidelines and Recommendations for the Clinical Practice of Contrast-Enhanced Ultrasound (CEUS) in Non-Hepatic Applications: Update 2017 (Long Version) [Die EFSUMB-Leitlinien und Empfehlungen f\u0026uuml;r den klinischen Einsatz des kontrastverst\u0026auml;rkten Ultraschalls (CEUS) bei nicht-hepatischen Anwendungen: Update 2017 (Langversion)]. Ultraschall Med. 2018;39(2):e2-e44.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eWang B, Yang D, Zhang X, et al. The diagnostic value of contrast-enhanced ultrasonography in breast ductal abnormalities. Cancer Imaging. 2023; 23(1):25. \u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003e\u0026zwnj;Table 1. Association between all predictive parameters and Gleason Grade (low, medium, high)\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003eH(high)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003eL(low)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003eM(medium)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003eP-value\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eTPSA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e302.46\u0026plusmn;571.91\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e174.62\u0026plusmn;508.47\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e48.58\u0026plusmn;76.22\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.000\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eFPSA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e18.43\u0026plusmn;17.77\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e9.25\u0026plusmn;13.89\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e5.09\u0026plusmn;6.69\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.000\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eFPSA/TPSA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.12\u0026plusmn;0.07\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.13\u0026plusmn;0.06\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.14\u0026plusmn;0.07\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.305\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eNLR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e2.71\u0026plusmn;1.28\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e2.71\u0026plusmn;1.39\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e2.73\u0026plusmn;1.15\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.731\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eELR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.09\u0026plusmn;0.08\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.09\u0026plusmn;0.07\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.09\u0026plusmn;0.07\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.863\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eLMR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e3.79\u0026plusmn;1.56\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e3.85\u0026plusmn;1.91\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e3.65\u0026plusmn;1.43\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.862\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003ePLR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e136.92\u0026plusmn;65.30\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e124.60\u0026plusmn;55.39\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e130.85\u0026plusmn;52.26\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.448\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eneutrophilic_granulocyte\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e4.13\u0026plusmn;1.51\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e4.08\u0026plusmn;1.28\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e4.18\u0026plusmn;1.06\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.624\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003elymphocyte\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.67\u0026plusmn;0.51\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.68\u0026plusmn;0.56\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.69\u0026plusmn;0.57\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.957\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eeosinophil\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.14\u0026plusmn;0.11\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.14\u0026plusmn;0.10\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.14\u0026plusmn;0.10\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.884\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003emonocyte\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.49\u0026plusmn;0.20\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.52\u0026plusmn;0.27\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.50\u0026plusmn;0.18\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.778\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003ethrombocyte\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e206.12\u0026plusmn;62.75\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e189.31\u0026plusmn;57.66\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e201.37\u0026plusmn;54.53\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.055\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eage\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e71.38\u0026plusmn;8.29\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e70.46\u0026plusmn;8.10\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e70.10\u0026plusmn;8.75\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.552\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eBMI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e23.75\u0026plusmn;2.64\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e23.58\u0026plusmn;3.08\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e23.75\u0026plusmn;2.93\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.542\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003emaximum_diameter\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e18.79\u0026plusmn;9.03\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e17.70\u0026plusmn;7.72\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e17.52\u0026plusmn;7.08\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.794\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eintegrity_of_envelope\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.13\u0026plusmn;0.34\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.12\u0026plusmn;0.33\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.11\u0026plusmn;0.32\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.932\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003etumor_echo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.16\u0026plusmn;0.21\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.22\u0026plusmn;0.34\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.18\u0026plusmn;0.26\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.879\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eprostate_volume\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e53.45\u0026plusmn;53.18\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e52.44\u0026plusmn;19.74\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e21.29\u0026plusmn;18.05\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.930\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eprostate_shape\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.13\u0026plusmn;1.16\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.15\u0026plusmn;0.34\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.36\u0026plusmn;0.36\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.883\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003einside_echo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.28\u0026plusmn;1.16\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.23\u0026plusmn;0.45\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.36\u0026plusmn;0.42\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.093\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003etumor_shape\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.12\u0026plusmn;1.14\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.13\u0026plusmn;0.33\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.34\u0026plusmn;0.34\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.933\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003etumor_edge\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.17\u0026plusmn;1.11\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.12\u0026plusmn;0.38\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.31\u0026plusmn;0.33\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.413\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eelastic_score\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e3.70\u0026plusmn;3.08\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e3.43\u0026plusmn;0.98\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.01\u0026plusmn;0.97\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.000\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eelastic_ratio\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e8.53\u0026plusmn;5.80\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e6.44\u0026plusmn;1.25\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e2.18\u0026plusmn;0.73\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.000\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eHiguchi\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e2.67\u0026plusmn;2.66\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e2.65\u0026plusmn;0.13\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.12\u0026plusmn;0.14\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.635\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eSlope\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.68\u0026plusmn;0.70\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.69\u0026plusmn;0.15\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.15\u0026plusmn;0.16\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.855\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eIntercept\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.85\u0026plusmn;0.69\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.74\u0026plusmn;0.05\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.17\u0026plusmn;0.07\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.000\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eMid_band\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.69\u0026plusmn;0.60\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.65\u0026plusmn;0.07\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.10\u0026plusmn;0.09\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.000\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eS1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.21\u0026plusmn;0.16\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.18\u0026plusmn;0.01\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.04\u0026plusmn;0.01\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.000\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eS2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.12\u0026plusmn;0.12\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.13\u0026plusmn;0.02\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.02\u0026plusmn;0.03\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.187\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eS3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.09\u0026plusmn;0.09\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.09\u0026plusmn;0.04\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.04\u0026plusmn;0.04\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.649\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eS4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.02\u0026plusmn;0.05\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.04\u0026plusmn;0.01\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.02\u0026plusmn;0.01\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.000\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eTtoPK\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e14.25\u0026plusmn;19.30\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e18.25\u0026plusmn;1.68\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e3.83\u0026plusmn;1.11\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.000\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eArea\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e598.41\u0026plusmn;584.74\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e581.18\u0026plusmn;77.59\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e95.21\u0026plusmn;62.54\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.102\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eGrad\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.82\u0026plusmn;1.26\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.54\u0026plusmn;0.10\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.41\u0026plusmn;1.49\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.000\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003eAtm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e14.39\u0026plusmn;14.47\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e15.14\u0026plusmn;1.49\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e1.57\u0026plusmn;1.88\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.002\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 175px;\"\u003e\n \u003cp\u003ePI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e25.70\u0026plusmn;22.74\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e24.28\u0026plusmn;1.09\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e2.33\u0026plusmn;1.02\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0.000\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003cp\u003eTable 2. Predictive performance of all models\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" class=\"fr-table-selection-hover\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e\n \u003cp\u003eModel\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003eAccuracy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003eRecall\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 102px;\"\u003e\n \u003cp\u003eF1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e\n \u003cp\u003eDecisionTree\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e0.818\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e0.818\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 102px;\"\u003e\n \u003cp\u003e0.816\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e\n \u003cp\u003eRandomForest\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e0.820\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e0.820\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 102px;\"\u003e\n \u003cp\u003e0.820\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e\n \u003cp\u003eKNeighbors\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e0.788\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e0.788\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 102px;\"\u003e\n \u003cp\u003e0.801\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e\n \u003cp\u003eGBDT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e0.848\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e0.848\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 102px;\"\u003e\n \u003cp\u003e0.849\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e\n \u003cp\u003eXgboost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e0.818\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e0.789\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 102px;\"\u003e\n \u003cp\u003e0.796\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e\n \u003cp\u003eNaive_bayes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e0.773\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e0.773\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 102px;\"\u003e\n \u003cp\u003e0.779\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"prostate cancer, Gleason grading, ultrasonography, predictive model, machine learning, inflammatory biomarkers","lastPublishedDoi":"10.21203/rs.3.rs-6998221/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6998221/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground: \u003c/strong\u003eProstate cancer is a common malignancy among elderly males with a growing incidence. While prostate biopsy remains the gold standard for diagnosis, this invasive procedure is poorly tolerated by some patients. The Gleason grade group (GGG) plays a critical role in predicting metastatic risk, guiding treatment selection, and is strongly associated with survival outcomes. Consequently, noninvasive prediction of prostate cancer Gleason grading has emerged as a research priority. This study aimed to develop a noninvasive predictive model integrating multimodal ultrasound data and clinical laboratory biomarkers to preoperatively determine GGGs in prostate cancer patients.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods:\u003c/strong\u003e This single-center prospective study enrolled 329 prostate cancer patients meeting predefined inclusion criteria. All participants underwent prostate biopsy with subsequent Gleason grading and were categorized into three groups: low-grade (Gleason score ≤6), intermediate-grade (Gleason score 7), and high-grade (Gleason score ≥8). Thirty-seven predictive parameters were collected, including clinical laboratory biomarkers, systemic inflammatory markers (e.g., neutrophil-to-lymphocyte ratio), and multimodal ultrasound data: Grayscale sonographic characteristics, contrast-enhanced ultrasound (CEUS) parameters, elastography parameters, and radiofrequency signal data. Following feature selection, five clinically significant predictors were identified. Multiple machine learning algorithms were implemented for predictive modeling, and model performance was quantified using accuracy, recall, and F1-score.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults: \u003c/strong\u003eSix machine learning-based predictive models were developed and evaluated. The Decision Tree model achieved an accuracy of 0.818, recall of 0.818, and F1-score of 0.816. The Random Forest classifier demonstrated an accuracy of 0.820, recall of 0.820, and F1-score of 0.820. The K-Nearest Neighbors algorithm yielded an accuracy of 0.788, recall of 0.788, and F1-score of 0.801. The Gradient Boosting Decision Tree (GBDT) model exhibited superior predictive capability with an accuracy of 0.848, recall of 0.848, and F1-score of 0.849. The XGBoost algorithm had an accuracy of 0.818, recall of 0.789, and F1-score of 0.796, while the Naive Bayes classifier attained an accuracy of 0.773, recall of 0.773, and F1-score of 0.779. Comparative analysis revealed that the GBDT model demonstrated optimal performance among the evaluated algorithms, suggesting its potential clinical significance in predicting Gleason grades.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion\u003c/strong\u003e: Ultrasonography, being noninvasive, radiation-free, and cost-effective, demonstrates high clinical feasibility for implementation in routine practice, particularly in primary healthcare settings. The predictive model established through multimodal ultrasound parameters effectively predicts the Gleason grade of prostate cancer.\u003c/p\u003e","manuscriptTitle":"An Innovative Approach for Predicting Prostate Cancer Gleason Grading: Machine Learning-based Fusion of Multimodal Ultrasound, Clinical and Laboratory Indicators ","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-07 04:16:35","doi":"10.21203/rs.3.rs-6998221/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"02a9fe3b-bdf7-40b0-b545-ad189ed74088","owner":[],"postedDate":"August 7th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-09-22T07:23:28+00:00","versionOfRecord":[],"versionCreatedAt":"2025-08-07 04:16:35","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6998221","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6998221","identity":"rs-6998221","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}