Abstract:Background Automatic detection and segmentation of intraprostatic lesions (ILs) on preoperative multiparametric‐magnetic resonance images (mp‐MRI) can improve clinical workflow efficiency and enhance the diagnostic accuracy of prostate cancer and is an essential step in dominant intraprostatic lesion boost. Purpose The goal is to improve the detection and segmentation accuracy of 3D ILs in MRI by a proposed a deep learning (DL)‐based algorithm with histopathological ground truth. Methods This retrospective study included 262 patients with in vivo prostate biparametric MRI (bp‐MRI) scans and were divided into three cohorts based on their data analysis and annotation. Histopathological ground truth was established by using histopathology images as delineation reference standard on cohort 1, which consisted of 64 patients and was randomly split into 20 training, 12 validation, and 32 testing patients. Cohort 2 consisted of 158 patients with bp‐MRI based lesion delineation, and was randomly split into 104 training, 15 validation, and 39 testing patients. Cohort 3 consisted of 40 unannotated patients, used in semi‐supervised learning. We proposed a non‐local Mask R‐CNN and boosted its performance by applying different training techniques. The performance of non‐local Mask R‐CNN was compared with baseline Mask R‐CNN, 3D U‐Net and an experienced radiologist's delineation and was evaluated by detection rate, dice similarity coefficient (DSC), sensitivity, and Hausdorff Distance (HD). Results The independent testing set consists of 32 patients with histopathological ground truth. With the training technique maximizing detection rate, the non‐local Mask R‐CNN achieved 80.5% and 94.7% detection rate; 0.548 and 0.604 DSC; 5.72 and 6.36 95 HD (mm); 0.613 and 0.580 sensitivity for ILs of all Gleason Grade groups (GGGs) and clinically significant ILs (GGG > 2), which outperformed baseline Mask R‐CNN and 3D U‐Net. For clinically significant ILs, the model segmentation accuracy was significantly higher than that of the experienced radiologist involved in the study, who achieved 0.512 DSC (p = 0.04), 8.21 (p = 0.041) 95 HD (mm), and 0.398 (p = 0.001) sensitivity. Conclusion The proposed DL model achieved state‐of‐art performance and has the potential to help improve radiotherapy treatment planning and noninvasive prostate cancer diagnosis.

Enhancing Precision in Medical Imaging: A 3D CNN Approach for Fiducial Point Detection in MRI Data

Fast and Accurate Craniomaxillofacial Landmark Detection Via 3D Faster R-CNN.

Automated Fiducial Marker Detection and Fiducial Point Localization in CT Images for Lung Biopsy Image-Guided Surgery Systems.

Development and evaluation of a deep learning based artificial intelligence for automatic identification of gold fiducial markers in an MRI-only prostate radiotherapy workflow

Automatic Localization of the Center of Fiducial Markers in 3d Ct/Mri Images for Image-Guided Neurosurgery

Automatic Detection of Craniomaxillofacial Anatomical Landmarks on CBCT Images Using 3D Mask R-CNN.

Fast and Robust Detection of Anatomical Landmarks Using Cascaded 3D Convolutional Networks Guided by Linear Square Regression.

Deep Attention Mask Regional Convolutional Neural Network for Head-and-neck MRI Multi-Organ Auto-Delineation

Deep learning-based evaluation of the relationship between mandibular third molar and mandibular canal on CBCT

The Application of Mask Region-Based Convolutional Neural Networks in the Detection of Nasal Septal Deviation Using Cone Beam Computed Tomography Images: Proof-of-Concept Study

A model-based, semi-global segmentation approach for automatic 3-D point landmark localization in neuroimages

A Systematic Approach for MRI Brain Tumor Localization and Segmentation Using Deep Learning and Active Contouring

Landmark detection in Cardiac Magnetic Resonance Imaging Using A Convolutional Neural Network

A preliminary exploration into top-down and bottom-up deep-learning approaches to localising neuro-interventional point targets in volumetric MRI

Automatic Landmark Localization in 3D Medical CT Images: Few-Shot Learning Through Optimized Data Pre-Processing and Network Design.

An XAI-enhanced efficientNetB0 framework for precision brain tumor detection in MRI imaging

Prostate cancer detection and segmentation on MRI using non‐local mask R‐CNN with histopathological ground truth

Automatic Detection of Cerebral Microbleeds from MR Images Via 3D Convolutional Neural Networks

Local and Deep Features Based Convolutional Neural Network Frameworks for Brain MRI Anomaly Detection

End-to-end prostate cancer detection in bpMRI via 3D CNNs: Effects of attention mechanisms, clinical priori and decoupled false positive reduction

Fiducial marker recovery and detection from severely truncated data in navigation assisted spine surgery