Deep learning-based artificial intelligence for prostate cancer detection at biparametric MRI

Abstract

Purpose: To present fully automated DL-based prostate cancer detection system for prostate MRI.

Methods: MRI 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.

Results: 525 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.

Conclusion: We 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.

Keywords: Artificial intelligence; Deep learning; Magnetic resonance imaging; Prostate cancer.

Fig. 1

68-year-old male with serum PSA 10.57 ng/ml who underwent targeted biopsy of MR findings, revealing Gleason 3 + 3 in right midline peripheral zone lesion and Gleason 4 + 4 in left anterior transition zone lesion. Both UNet and AHNet produced true positive segmentation in left anterior transition zone lesion and false negative result in region of right midline peripheral zone lesion. In addition to biopsy positive lesions, AHNet produced 1 false positive and overall Dice score of 0.80922. UNet overall Dice score was 0.72229

Fig. 2

57-year-old male with serum PSA 6.68 ng/ml who underwent targeted biopsy of MR findings, revealing Gleason 3 + 3 in all targeted cores (n = 3 lesions, 2 shown in this figure). UNet results demonstrated false negative on the patient-level (Dice = 0, n = 3 False Negative lesions), while AHNet underestimated the volume of disease (Dice = 0.0385, n = 1 True Positive lesion, n = 2 False Negative lesions)

Fig. 3

55-year-old male with serum PSA 10.97 ng/ml who underwent targeted biopsy of MR findings, revealing Gleason 3 + 4 in one lesion located in left anterior transition zone, which was targeted at biopsy. UNet results demonstrated true positive detection with 1 false positive penalty (DSC = 0.2057). AHNet produced 1 true positive detection with 3 false positive penalty lesions (DSC = 0.4776)

Fig. 4

Outcome of AI predictions in targeted biopsies with benign result