Prostate cancer of magnetic resonance imaging automatic segmentation and detection of based on 3D-Mask RCNN

• DOI: 10.1016/j.jrras.2023.100636
• Publication Date: 2023-07-23T15:36:35Z
• Authors: Shu-Ting Li; Ling Zhang; Ping Guo; Hong-yi Pan; Ping-zhen Chen; Hai-fang Xie; Bo-kai Xie; Jiayang Chen; Qing-quan Lai; Yuan-zhe Li; Hong Wu; Yi Wang
• Keywords: Cancer Detection

Prostate cancer is a widespread form of that impacts men across the world. MRI plays pivotal role in detection and precise localization cancerous regions, aiding medical professionals devising effective treatment strategies for patients. As result, often used diagnosis prostate cancer. Our proposed method employs deep learning to achieve automatic segmentation single series, particularly T2-weighted imaging (T2WI). study utilized data from 133 patients at hospital, consisting 71 cases 62 benign prostatic tumors. We employed (T2WI) series 93 prostates as training set our 3D-Mask RCNN model, while remaining 40 were validation. The masks manually delineated by an experienced radiologist, with pathology serving reference standard. approach was evaluated using several metrics, such dice similarity coefficient (DSC), accuracy, sensitivity, specificity, receiver operating characteristic (ROC) curve analysis. produced promising results 3D Mask R-CNN model. yielded DSC score 0.856, sensitivity 0.921, specificity 0.961. test also successful, 0.849, 0.911, 0.931. Furthermore, model achieved AUC value 0.865 0.866, 0.875, 0.835, respectively, set. had 0.842 0.836, 0.847, 0.819, respectively. These findings demonstrate capable accurately detecting segmenting - T2WI. use tumors T2WI has been shown be highly precise. This potential greatly benefit radiologists improving accuracy efficiency diagnoses, leading more planning By automating process, this can reduce workload increase consistency diagnoses. high performance highlights techniques demonstrates significant impact these approaches have on patient outcomes.