Accurate localization of prostate cancer on magnetic resonance imaging (MRI) remains challenging due to the subtle appearance of cancer, resulting in missed clinically significant cancers, large inter-reader variability and large numbers of confounders that require biopsy confirmation. Vision foundation models have shown promise, but direct transfer to prostate MRI is challenging due to the substantial domain gap between natural images and prostate MRI and the subtle appearance of prostate cancer. We therefore developed prostate vision contrastive networks (ProViCNet), a weakly supervised model utilizing patch-level contrastive learning on MRI to support MRI-based screening, biopsy targeting and focal treatment planning. ProViCNet was trained and validated (using 4401 patients across six cohorts) as a prostate cancer detection model on MRI. Training labels were biopsy-confirmed radiologist annotations, while the evaluation labels included both biopsy and surgery-confirmed lesions. ProViCNet demonstrated consistent detection and segmentation performance across multiple internal and external validation cohorts, with area under the receiver operating characteristic curve (AUROC) values ranging from 0.875 to 0.966, and outperforming radiologists in the MRI expert reader study (0.907 versus 0.805, p < 0.01). We also integrated ProViCNet with serum PSA to develop a virtual screening test, which preserved high sensitivity for detecting clinically significant cancers while more than doubling specificity from 15% to 38% (p < 0.001) among men with PSA ≥ 4 ng/mL, thereby potentially reducing unnecessary biopsies. These findings highlight ProViCNet's potential for enhancing the accuracy of prostate cancer diagnosis and reducing unnecessary biopsies.

BACKGROUND AND OBJECTIVE:To assess whether conventional brightness-mode (B-mode) transrectal ultrasound images of the prostate reveal clinically significant cancers with the help of artificial intelligence methods. METHODS:This study included 2986 men who underwent biopsies at two institutions. We trained the PROstate Cancer detection on B-mode transrectal UltraSound images NETwork (ProCUSNet) to determine whether ultrasound can reliably detect cancer. Specifically, ProCUSNet is based on the well-established nnUNet frameworks, and seeks to detect and outline clinically significant cancer on three-dimensional (3D) examinations reconstructed from 2D screen captures. We compared ProCUSNet against (1) reference labels (n = 515 patients), (2) eight readers that interpreted B-mode ultrasound (n = 20-80 patients), and (3) radiologists interpreting magnetic resonance imaging (MRI) for clinical care (n = 110 radical prostatectomy patients). KEY FINDINGS AND LIMITATIONS:ProCUSNet found 82% clinically significant cancer cases with a lesion boundary error of up to 2.67 mm and detected 42% more lesions than ultrasound readers (sensitivity: 0.86 vs 0.44, p < 0.05, Wilcoxon test, Bonferroni correction). Furthermore, ProCUSNet has similar performance to radiologists interpreting MRI when accounting for registration errors (sensitivity: 0.79 vs 0.78, p > 0.05, Wilcoxon test, Bonferroni correction), while having the same targeting utility as a supplement to systematic biopsies. CONCLUSIONS AND CLINICAL IMPLICATIONS:ProCUSNet can localize clinically significant cancer on screen capture B-mode ultrasound, a task that is particularly challenging for clinicians reading these examinations. As a supplement to systematic biopsies, ProCUSNet appears comparable with MRI, suggesting its utility for targeting suspicious lesions during the biopsy and possibly for screening using ultrasound alone, in the absence of MRI.

Magnetic Resonance Imaging (MRI) is increasingly being used to detect prostate cancer, yet its interpretation can be challenging due to subtle differences between benign and cancerous tissue. Recently, Denoising Diffusion Probabilistic Models (DDPMs) have shown great utility for medical image segmentation, modeling the process as noise removal in standard Gaussian distributions. In this study, we further enhance DDPMs by introducing the knowledge that the occurrence of cancer varies across the prostate (e.g., ∼70% of prostate cancers occur in the peripheral zone). We quantify such heterogeneity with a registration pipeline to calculate voxel-level cancer distribution mean and variances. Our proposed approach, ProstAtlasDiff, relies on DDPMs that use the cancer atlas to model noise removal and segment cancer on MRI. We trained and evaluated the performance of ProstAtlasDiff in detecting clinically significant cancer in a multi-institution multi-scanner dataset, and compared it with alternative models. In a lesion-level evaluation, ProstAtlasDiff achieved statistically significantly higher accuracy (0.91 vs. 0.85, p<0.001), specificity (0.91 vs. 0.84, p<0.001), positive predictive value (PPV, 0.50 vs. 0.35, p<0.001), compared to alternative models. ProstAtlasDiff also offers more accurate cancer outlines, achieving a higher Dice Coefficient (0.33 vs. 0.31, p<0.01). Furthermore, we evaluated ProstAtlasDiff in an independent cohort of 91 patients who underwent radical prostatectomy to compare its performance to that of radiologists, relative to whole-mount histopathology ground truth. ProstAtlasDiff detected 16% (15 lesions out of 93) more clinically significant cancers compared to radiologists (sensitivity: 0.90 vs. 0.75, p<0.01), and was comparable in terms of ROC-AUC, PR-AUC, PPV, accuracy, and Dice coefficient (p≥0.05). Furthermore, we evaluated ProstAtlasDiff in a second independent cohort of 537 subjects and observed that ProsAtlasDiff outperformed alternative approaches. These results suggest that ProstAltasDiff has the potential to assist in localizing cancer for biopsy guidance and treatment planning.

Pre-biopsy magnetic resonance imaging (MRI) is increasingly used to target suspicious prostate lesions. This has led to artificial intelligence (AI) applications improving MRI-based detection of clinically significant prostate cancer (CsPCa). However, MRI-detected lesions must still be mapped to transrectal ultrasound (TRUS) images during biopsy, which results in missing CsPCa. This study systematically evaluates a multimodal AI framework integrating MRI and TRUS image sequences to enhance CsPCa identification. The study included 3110 patients from three cohorts across two institutions who underwent prostate biopsy. The proposed framework, based on the 3D UNet architecture, was evaluated on 1700 test cases, comparing performance to unimodal AI models that use either MRI or TRUS alone. Additionally, the proposed model was compared to radiologists in a cohort of 110 patients. The multimodal AI approach achieved superior sensitivity (80 (42 to radiologists, the multimodal model showed higher specificity (88 and Lesion Dice (38 demonstrate the potential of multimodal AI to improve CsPCa lesion targeting during biopsy and treatment planning, surpassing current unimodal models and radiologists; ultimately improving outcomes for prostate cancer patients.

Electrocardiogram (ECG) analysis is foundational for cardiovascular disease diagnosis, yet the performance of deep learning models is often constrained by limited access to annotated data. Self-supervised contrastive learning has emerged as a powerful approach for learning robust ECG representations from unlabeled signals. However, most existing methods generate only pairwise augmented views and fail to leverage the rich temporal structure of ECG recordings. In this work, we present a poly-window contrastive learning framework. We extract multiple temporal windows from each ECG instance to construct positive pairs and maximize their agreement via statistics. Inspired by the principle of slow feature analysis, our approach explicitly encourages the model to learn temporally invariant and physiologically meaningful features that persist across time. We validate our approach through extensive experiments and ablation studies on the PTB-XL dataset. Our results demonstrate that poly-window contrastive learning consistently outperforms conventional two-view methods in multi-label superclass classification, achieving higher AUROC (0.891 vs. 0.888) and F1 scores (0.680 vs. 0.679) while requiring up to four times fewer pre-training epochs (32 vs. 128) and 14.8% in total wall clock pre-training time reduction. Despite processing multiple windows per sample, we achieve a significant reduction in the number of training epochs and total computation time, making our method practical for training foundational models. Through extensive ablations, we identify optimal design choices and demonstrate robustness across various hyperparameters. These findings establish poly-window contrastive learning as a highly efficient and scalable paradigm for automated ECG analysis and provide a promising general framework for self-supervised representation learning in biomedical time-series data.

OBJECTIVES:To improve sensitivity and inter-reader consistency of prostate cancer localisation on micro-ultrasonography (MUS) by developing a deep learning model for automatic cancer segmentation, and to compare model performance with that of expert urologists. PATIENTS AND METHODS:We performed an institutional review board-approved prospective collection of MUS images from patients undergoing magnetic resonance imaging (MRI)-ultrasonography fusion guided biopsy at a single institution. Patients underwent 14-core systematic biopsy and additional targeted sampling of suspicious MRI lesions. Biopsy pathology and MRI information were cross-referenced to annotate the locations of International Society of Urological Pathology Grade Group (GG) ≥2 clinically significant cancer on MUS images. We trained a no-new U-Net model - the Prostate Micro-Ultrasound Network (ProMUS-NET) - to localise GG ≥2 cancer on these image stacks in a fivefold cross-validation. Performance was compared vs that of six expert urologists in a matched sub-cohort. RESULTS:The artificial intelligence (AI) model achieved an area under the receiver-operating characteristic curve of 0.92 and detected more cancers than urologists (lesion-level sensitivity 73% vs 58%; patient-level sensitivity 77% vs 66%). AI lesion-level sensitivity for peripheral zone lesions was 86.2%. CONCLUSIONS:Our AI model identified prostate cancer lesions on MUS with high sensitivity and specificity. Further work is ongoing to improve margin overlap, to reduce false positives, and to perform external validation. AI-assisted prostate cancer detection on MUS has great potential to improve biopsy diagnosis by urologists.

Prostate cancer is a major cause of cancer-related deaths in men, where early detection greatly improves survival rates. Although MRI-TRUS fusion biopsy offers superior accuracy by combining MRI's detailed visualization with TRUS's real-time guidance, it is a complex and time-intensive procedure that relies heavily on manual annotations, leading to potential errors. To address these challenges, we propose a fully automatic MRI-TRUS fusion-based segmentation method that identifies prostate tumors directly in TRUS images without requiring manual annotations. Unlike traditional multimodal fusion approaches that rely on naive data concatenation, our method integrates a registration-segmentation framework to align and leverage spatial information between MRI and TRUS modalities. This alignment enhances segmentation accuracy and reduces reliance on manual effort. Our approach was validated on a dataset of 1,747 patients from Stanford Hospital, achieving an average Dice coefficient of 0.212, outperforming TRUS-only (0.117) and naive MRI-TRUS fusion (0.132) methods, with significant improvements (p $<$ 0.01). This framework demonstrates the potential for reducing the complexity of prostate cancer diagnosis and provides a flexible architecture applicable to other multimodal medical imaging tasks.

Accurate prostate cancer diagnosis remains challenging. Even when using MRI, radiologists exhibit low specificity and significant inter-observer variability, leading to potential delays or inaccuracies in identifying clinically significant cancers. This leads to numerous unnecessary biopsies and risks of missing clinically significant cancers. Here we present prostate vision contrastive network (ProViCNet), prostate organ-specific vision foundation models for Magnetic Resonance Imaging (MRI) and Trans-Rectal Ultrasound imaging (TRUS) for comprehensive cancer detection. ProViCNet was trained and validated using 4,401 patients across six institutions, as a prostate cancer detection model on radiology images relying on patch-level contrastive learning guided by biopsy confirmed radiologist annotations. ProViCNet demonstrated consistent performance across multiple internal and external validation cohorts with area under the receiver operating curve values ranging from 0.875 to 0.966, significantly outperforming radiologists in the reader study (0.907 versus 0.805, p<0.001) for mpMRI, while achieving 0.670 to 0.740 for TRUS. We also integrated ProViCNet with standard PSA to develop a virtual screening test, and we showed that we can maintain the high sensitivity for detecting clinically significant cancers while more than doubling specificity from 15 unnecessary biopsies. These findings highlight that ProViCNet's potential for enhancing prostate cancer diagnosis accuracy and reduce unnecessary biopsies, thereby optimizing diagnostic pathways.

You have accessJournal of UrologyProstate Cancer: Detection & Screening V (PD50)1 May 2024PD50-01 AI VS. UROLOGISTS: A COMPARATIVE ANALYSIS FOR PROSTATE CANCER DETECTION ON TRANSRECTAL B-MODE ULTRASOUND Sulaiman Vesal, Indrani Bhattacharya, Hassan Jahanandish, Moonhyung Choi, Steve Ran Zhou, Zachary Kornberg, Elijah Richard Sommer, Richard E. Fan, Mirabela Rusu, and Geoffrey A. Sonn Sulaiman VesalSulaiman Vesal , Indrani BhattacharyaIndrani Bhattacharya , Hassan JahanandishHassan Jahanandish , Moonhyung ChoiMoonhyung Choi , Steve Ran ZhouSteve Ran Zhou , Zachary KornbergZachary Kornberg , Elijah Richard SommerElijah Richard Sommer , Richard E. FanRichard E. Fan , Mirabela RusuMirabela Rusu , and Geoffrey A. SonnGeoffrey A. Sonn View All Author Informationhttps://doi.org/10.1097/01.JU.0001008620.35181.96.01AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: Transrectal ultrasound-guided (TRUS) is widely used to guide prostate biopsy, but its effectiveness is limited due to low sensitivity and signal-to-noise ratio in B-mode TRUS images. While MRI fusion improves prostate biopsy accuracy, only ∼25% of biopsies nationwide are performed following MRI. Instead, most biopsies are still performed using TRUS alone. Better prostate cancer (PCa) detection on B-mode TRUS images could greatly improve biopsy targeting. METHODS: In this study, we developed ProsDectNet, an AI model to detect and localize clinically significant prostate cancer in B-mode TRUS images. ProsDectNet consists of a lesion detection backbone and a false positive reduction component to enhance model performance. We trained and validated ProsDectNet using a cohort of 289 patients (N=110 with a Gleason score ≥7, and N=179 without cancer) who underwent MRI-TRUS fusion targeted biopsy and tested our approach on an independent group of 41 (N=20 with a Gleason score ≥7, and N=21 without cancer) patients. For comparison, we asked three urologist-readers and one radiologist-reader (blinded to MRI or clinical data) to review all test set TRUS images and annotate any areas of suspected cancer with unlimited time. Successful lesion identification was defined as any overlap between AI predictions and ground-truth labels. We computed lesion-level and patient-level sensitivity, specificity, and positive predictive value (PPV) to compare ProsDectNet performance vs. readers. RESULTS: ProsDectNet demonstrated a patient-level sensitivity and specificity of 74.0% and 67.0%, and lesion-level sensitivity and specificity of 66.0% and 90.0%, respectively, surpassing the average urologist's results (50.0% sensitivity, 70.0% specificity, 43.0% lesion-level sensitivity, and 91.0% specificity). Additionally, its patient-level PPV of 67.0% outperformed the average urologists' PPV of 62.0%. CONCLUSIONS: Our study developed a high-performing AI model to identify cancer on TRUS images. The model outperformed urologists in detecting PCa. This emphasizes its potential for improving biopsy guidance with B-mode ultrasound, especially in scenarios where MRI was unavailable. Ongoing research aims to validate and enhance our approach through expanded datasets and real-world clinical evaluations. Download PPT Source of Funding: Departments of Radiology and Urology, Stanford University, National Cancer Institute of the National Institutes of Health (R37CA260346 to M.R), and the generous philanthropic support of our patients (G.S.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health © 2024 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 211Issue 5SMay 2024Page: e1056 Advertisement Copyright & Permissions© 2024 by American Urological Association Education and Research, Inc.Metrics Author Information Sulaiman Vesal More articles by this author Indrani Bhattacharya More articles by this author Hassan Jahanandish More articles by this author Moonhyung Choi More articles by this author Steve Ran Zhou More articles by this author Zachary Kornberg More articles by this author Elijah Richard Sommer More articles by this author Richard E. Fan More articles by this author Mirabela Rusu More articles by this author Geoffrey A. Sonn More articles by this author Expand All Advertisement PDF downloadLoading ...

Background and objective:Micro-ultrasound (MUS) uses a high-frequency transducer with superior resolution to conventional ultrasound, which may differentiate prostate cancer from normal tissue and thereby allow targeted biopsy. Preliminary evidence has shown comparable sensitivity to magnetic resonance imaging (MRI), but consistency between users has yet to be described. Our objective was to assess agreement of MUS interpretation across multiple readers. Methods:After institutional review board approval, we prospectively collected MUS images for 57 patients referred for prostate biopsy after multiparametric MRI from 2022 to 2023. MUS images were interpreted by six urologists at four institutions with varying experience (range 2-6 yr). Readers were blinded to MRI results and clinical data. The primary outcome was reader agreement on the locations of suspicious lesions, measured in terms of Light's κ and positive percent agreement (PPA). Reader sensitivity for identification of grade group (GG) ≥2 prostate cancer was a secondary outcome. Key findings and limitations:Analysis revealed a κ value of 0.30 (95% confidence interval [CI] 0.21-0.39). PPA was 33% (95% CI 25-42%). The mean patient-level sensitivity for GG ≥2 cancer was 0.66 ± 0.05 overall and 0.87 ± 0.09 when cases with anterior lesions were excluded. Readers were 12 times more likely to detect higher-grade cancers (GG ≥3), with higher levels of agreement for this subgroup (κ 0.41, PPA 45%). Key limitations include the inability to prospectively biopsy reader-delineated targets and the inability of readers to perform live transducer maneuvers. Conclusions and clinical implications:Inter-reader agreement on the location of suspicious lesions on MUS is lower than rates previously reported for MRI. MUS sensitivity for cancer in the anterior gland is lacking. Patient summary:The ability to find cancer on imaging scans can vary between doctors. We found that there was frequent disagreement on the location of prostate cancer when doctors were using a new high-resolution scan method called micro-ultrasound. This suggests that the performance of micro-ultrasound is not yet consistent enough to replace MRI (magnetic resonance imaging) for diagnosis of prostate cancer.

The alignment of MRI and ultrasound images of the prostate is crucial in detecting prostate cancer during biopsies, directly affecting the accuracy of prostate cancer diagnosis. However, due to the low signal-to-noise ratio of ultrasound images and the varied imaging properties of the prostate between MRI and ultrasound, it's challenging to efficiently and accurately align MRI and ultrasound images of the prostate. This study aims to present an effective affine transformation method that can automatically register prostate MRI and ultrasound images. In real-world clinical practice, it may increase the effectiveness of prostate cancer biopsies and the accuracy of prostate cancer diagnosis.

Image registration can map the ground truth extent of prostate cancer from histopathology images onto MRI, facilitating the development of machine learning methods for early prostate cancer detection. Here, we present RAdiology PatHology Image Alignment (RAPHIA), an end-to-end pipeline for efficient and accurate registration of MRI and histopathology images. RAPHIA automates several time-consuming manual steps in existing approaches including prostate segmentation, estimation of the rotation angle and horizontal flipping in histopathology images, and estimation of MRI-histopathology slice correspondences. By utilizing deep learning registration networks, RAPHIA substantially reduces computational time. Furthermore, RAPHIA obviates the need for a multimodal image similarity metric by transferring histopathology image representations to MRI image representations and vice versa. With the assistance of RAPHIA, novice users achieved expert-level performance, and their mean error in estimating histopathology rotation angle was reduced by 51% (12 degrees vs 8 degrees), their mean accuracy of estimating histopathology flipping was increased by 5% (95.3% vs 100%), and their mean error in estimating MRI-histopathology slice correspondences was reduced by 45% (1.12 slices vs 0.62 slices). When compared to a recent conventional registration approach and a deep learning registration approach, RAPHIA achieved better mapping of histopathology cancer labels, with an improved mean Dice coefficient of cancer regions outlined on MRI and the deformed histopathology (0.44 vs 0.48 vs 0.50), and a reduced mean per-case processing time (51 vs 11 vs 4.5 min). The improved performance by RAPHIA allows efficient processing of large datasets for the development of machine-learning models for prostate cancer detection on MRI. Our code is publicly available at: https://github.com/pimed/RAPHIA.

Prostate cancer is the second-most lethal cancer in men. Since early diagnosis and treatment can drastically increase the 5-year survival rate of patients to >99%, magnetic resonance imaging (MRI) has been utilized due to its high sensitivity of 88%. However, due to lack of access to MRI, transrectal b-mode ultrasound (TRUS)-guided systematic prostate biopsy remains the standard of care for 93% of patients. While ubiquitous, TRUS-guided prostate biopsy suffers from the lack of lesion targeting, resulting in a sensitivity of 48%. To address this gap, we perform a preliminary study to assess the feasibility of localizing clinically significant cancer on b-mode ultrasound images of the prostate as input and propose a deep learning framework that learns to distinguish cancer at the pixel level. The proposed deep learning framework consists of a convolutional network with deep supervision at various scales and a clinical decision module that simultaneously learns to reduce false positive lesion predictions. We evaluated our deep learning framework using b-mode TRUS data with pathology confirmation from 330 patients, including 123 patients with pathology-confirmed cancer. Our results demonstrate the feasibility of using b-mode ultrasound images to localize prostate cancer lesions with a patient- level sensitivity and specificity of 68% and 91% respectively, compared to the reported clinical standard of 48% and 99%. The outcomes of this study show the promise of using a deep learning framework to localize prostate cancer lesions on the universally available b-mode ultrasound images; eventually improving the prostate biopsy procedures and enhancing the clinical outcomes for prostate cancer patients.

You have accessJournal of UrologySurgical Technology & Simulation: Artificial Intelligence II (PD27)1 May 2024PD27-03 A DEEP LEARNING MODEL FOR AUTOMATED PROSTATE CANCER DETECTION ON MICRO-ULTRASOUND Steve R. Zhou, Lichun Zhang, Moon Hyung Choi, Sulaiman Vesal, Richard E. Fan, Geoffrey Sonn, and Mirabela Rusu Steve R. ZhouSteve R. Zhou , Lichun ZhangLichun Zhang , Moon Hyung ChoiMoon Hyung Choi , Sulaiman VesalSulaiman Vesal , Richard E. FanRichard E. Fan , Geoffrey SonnGeoffrey Sonn , and Mirabela RusuMirabela Rusu View All Author Informationhttps://doi.org/10.1097/01.JU.0001008580.58088.27.03AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: The Prostate Risk Identification Using Micro-ultrasound (PRIMUS) protocol is validated to help urologists identify lesions for targeted biopsy (TB) without magnetic resonance imaging (MRI). Early studies have shown that micro-ultrasound (MUS)-guided TB has comparable sensitivity to MRI. However, as seen with PI-RADSv2.1, PRIMUS accuracy is likely user- and experience-dependent. The nnUNet is a well-established deep learning semantic segmentation model with successful applications to medical imaging. We therefore sought to automate PCa detection on MUS with an nnUNet 3D model. METHODS: We performed an IRB-approved prospective collection of MUS images from patients undergoing MRI/US-guided biopsy at a single institution. Images consisted of a single sagittal sweep through the prostate which constituted the input to our AI model. All patients received trans-perineal (TP) TB of PI-RADSv2.1 grade≥3 lesions as well as 14-core systematic biopsy (SB). Biopsy-confirmed MRI lesions in T2, DWI, and ADC sequences were co-registered to MUS to help annotate ground truth lesions. We used an nnUNet 3D model for supervised training and five-fold cross-validation for performance evaluation. Prostates were divided into 30 sectors for lesion-level analysis to report sensitivity and specificity. RESULTS: Our dataset included 41 patients with 44 biopsy-confirmed lesions. 7 (17%) patients had a negative biopsy, 34 (83%) patients had PCa, and 29 (71%) had grade group (GG)≥2 PCa. Model sensitivity was 0.77 and specificity was 0.85 for identifying GG≥2 PCa lesions on MUS. Overall accuracy was 0.85. Median Dice coefficient was 0.147 (IQR 0.04-0.34). The nnUnet detected most cancers, but the model also tended to annotate false positives due to imaging artifacts such as shadowing or calcifications (Figure 1). CONCLUSIONS: Our AI model identified PCa lesions on MUS with high sensitivity and specificity. Further work is ongoing to improve margin overlap as evidenced by our Dice coefficient and to perform external validation. Download PPT Source of Funding: None © 2024 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 211Issue 5SMay 2024Page: e551 Advertisement Copyright & Permissions© 2024 by American Urological Association Education and Research, Inc.Metrics Author Information Steve R. Zhou More articles by this author Lichun Zhang More articles by this author Moon Hyung Choi More articles by this author Sulaiman Vesal More articles by this author Richard E. Fan More articles by this author Geoffrey Sonn More articles by this author Mirabela Rusu More articles by this author Expand All Advertisement PDF downloadLoading ...

You have accessJournal of UrologyProstate Cancer: Detection & Screening I (MP19)1 May 2024MP19-17 INTEGRATING MR AND ULTRASOUND IMAGES FOR AI-BASED PROSTATE CANCER DETECTION IN TRANSRECTAL ULTRASOUND IMAGES: A COMPARATIVE ASSESSMENT WITH CLINICIANS Hassan Jahanandish, Sulaiman Vesal, Indrani Bhattacharya, Zachary Kornberg, Steve Ran Zhou, Elijah Richard Sommer, Moon Hyung Choi, Richard E. Fan, Mirabela Rusu, and Geoffrey A. Sonn Hassan JahanandishHassan Jahanandish , Sulaiman VesalSulaiman Vesal , Indrani BhattacharyaIndrani Bhattacharya , Zachary KornbergZachary Kornberg , Steve Ran ZhouSteve Ran Zhou , Elijah Richard SommerElijah Richard Sommer , Moon Hyung ChoiMoon Hyung Choi , Richard E. FanRichard E. Fan , Mirabela RusuMirabela Rusu , and Geoffrey A. SonnGeoffrey A. Sonn View All Author Informationhttps://doi.org/10.1097/01.JU.0001008716.22569.77.17AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: While MRI-guided biopsies have greatly improved prostate cancer detection, most biopsies are still performed using b-mode transrectal ultrasound (TRUS) imaging alone. TRUS biopsy detects just 48% of prostate cancer. Recent studies have focused on machine learning (ML) approaches for detecting prostate cancer in MR images. This limits the deployment of these models to the minority of men who receive a pre-biopsy prostate MRI. To make ML accessible to all men undergoing biopsy, we developed a novel deep learning framework that utilizes biomarkers from both MRI and TRUS images during training to detect prostate cancer foci in TRUS images alone, eliminating the need for MR images in deployment. Additionally, we evaluated our ML model performance against four urologists. METHODS: Our framework comprises a multimodal deep neural network that learns from both MRI and TRUS images, alongside a unimodal network that exclusively uses TRUS images as input both at training time and deployment. The multimodal network was first pre-trained to identify clinically significant prostate cancer (grade group ≥2). Then, this pre-trained model was used to guide the training of the unimodal TRUS-only network. We trained and tested our framework on a dataset of 102 patients (82 training and 20 test cases), with whole-mount pathology as the ground truth. Further, a baseline TRUS-only model was trained using the same dataset with no guidance from the multimodal model. Four urologists, with an average of 5 (±4.8) years of experience reading TRUS prostate images, reviewed the test cohort's TRUS images, manually annotating suspicious lesions without time restrictions. RESULTS: The multimodal-guided TRUS-only model achieved a sensitivity and specificity of 80% and 70%, respectively. This significantly outperformed the unguided baseline model, which achieved a performance of 54% and 48%. Furthermore, expert clinicians achieved a lower sensitivity of 35%, with a higher specificity of 92% compared to our ML approach. CONCLUSIONS: Our results demonstrate the effectiveness of our approach in integrating MRI and TRUS images for prostate cancer detection in TRUS images. The higher sensitivity compared to expert clinicians shows promise for enhancing prostate cancer biopsy diagnosis. Download PPT Source of Funding: Departments of Radiology and Urology, Stanford University, National Cancer Institute of the National Institutes of Health (R37CA260346 to M.R.), and the generous philanthropic support of our patients (G.S.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health © 2024 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 211Issue 5SMay 2024Page: e317 Advertisement Copyright & Permissions© 2024 by American Urological Association Education and Research, Inc.Metrics Author Information Hassan Jahanandish More articles by this author Sulaiman Vesal More articles by this author Indrani Bhattacharya More articles by this author Zachary Kornberg More articles by this author Steve Ran Zhou More articles by this author Elijah Richard Sommer More articles by this author Moon Hyung Choi More articles by this author Richard E. Fan More articles by this author Mirabela Rusu More articles by this author Geoffrey A. Sonn More articles by this author Expand All Advertisement PDF downloadLoading ...

Prostate cancer is a leading cause of cancer-related deaths among men. The recent development of high frequency, micro-ultrasound imaging offers improved resolution compared to conventional ultrasound and potentially a better ability to differentiate clinically significant cancer from normal tissue. However, the features of prostate cancer remain subtle, with ambiguous borders with normal tissue and large variations in appearance, making it challenging for both machine learning and humans to localize it on micro-ultrasound images. We propose a novel Mask Enhanced Deeply-supervised Micro-US network, termed MedMusNet, to automatically and more accurately segment prostate cancer to be used as potential targets for biopsy procedures. MedMusNet leverages predicted masks of prostate cancer to enforce the learned features layer-wisely within the network, reducing the influence of noise and improving overall consistency across frames. MedMusNet successfully detected 76 Dice Similarity Coefficient of 0.365, significantly outperforming the baseline Swin-M2F in specificity and accuracy (Wilcoxon test, Bonferroni correction, p-value<0.05). While the lesion-level and patient-level analyses showed improved performance compared to human experts and different baseline, the improvements did not reach statistical significance, likely on account of the small cohort. We have presented a novel approach to automatically detect and segment clinically significant prostate cancer on B-mode micro-ultrasound images. Our MedMusNet model outperformed other models, surpassing even human experts. These preliminary results suggest the potential for aiding urologists in prostate cancer diagnosis via biopsy and treatment decision-making.