Deep learning-based artificial intelligence for prostate cancer detection at biparametric MRI
Sherif Mehralivand,Dong Yang,Stephanie A. Harmon,Daguang Xu,Ziyue Xu,Holger Roth,Samira Masoudi,Deepak Kesani,Nathan Lay,Maria J. Merino,Bradford J. Wood,Peter A. Pinto,Peter L. Choyke,Baris Turkbey
DOI: https://doi.org/10.1007/s00261-022-03419-2
IF: 2.4
2022-01-31
Abdominal Radiology
Abstract:PurposeTo present fully automated DL-based prostate cancer detection system for prostate MRI.MethodsMRI scans from two institutions, were used for algorithm training, validation, testing. MRI-visible lesions were contoured by an experienced radiologist. All lesions were biopsied using MRI-TRUS-guidance. Lesions masks, histopathological results were used as ground truth labels to train UNet, AH-Net architectures for prostate cancer lesion detection, segmentation. Algorithm was trained to detect any prostate cancer ≥ ISUP1. Detection sensitivity, positive predictive values, mean number of false positive lesions per patient were used as performance metrics.Results525 patients were included for training, validation, testing of the algorithm. Dataset was split into training (n = 368, 70%), validation (n = 79, 15%), test (n = 78, 15%) cohorts. Dice coefficients in training, validation sets were 0.403, 0.307, respectively, for AHNet model compared to 0.372, 0.287, respectively, for UNet model. In validation set, detection sensitivity was 70.9%, PPV was 35.5%, mean number of false positive lesions/patient was 1.41 (range 0–6) for UNet model compared to 74.4% detection sensitivity, 47.8% PPV, mean number of false positive lesions/patient was 0.87 (range 0–5) for AHNet model. In test set, detection sensitivity for UNet was 72.8% compared to 63.0% for AHNet, mean number of false positive lesions/patient was 1.90 (range 0–7), 1.40 (range 0–6) in UNet, AHNet models, respectively.ConclusionWe developed a DL-based AI approach which predicts prostate cancer lesions at biparametric MRI with reasonable performance metrics. While false positive lesion calls remain as a challenge of AI-assisted detection algorithms, this system can be utilized as an adjunct tool by radiologists.
radiology, nuclear medicine & medical imaging