T2-weighted imaging-based deep-learning method for noninvasive prostate cancer detection and Gleason grade...

Area of Science:

• Radiology
• Artificial Intelligence
• Oncology

Background:

• Accurate prostate cancer detection and Gleason grade prediction are crucial for effective clinical management.
• Current methods may involve invasive procedures, highlighting the need for non-invasive alternatives.

Purpose of the Study:

• To develop and validate a deep-learning model for non-invasive prostate cancer detection.
• To predict the Gleason grade of prostate cancer using T2-weighted magnetic resonance imaging (MRI).

Main Methods:

• A deep-learning approach was designed for prostate cancer detection and Gleason grade prediction.
• Utilized retrospective internal datasets and external validation datasets from multiple centers.
• Performance was evaluated using the area under the curve (AUC).

Main Results:

• The deep-learning model achieved an AUC of 0.918 for prostate cancer detection in the validation set.
• For Gleason grade prediction, the model demonstrated AUCs of 0.902 (training), 0.854 (validation), 0.776 (external validation), and 0.838 (public challenge).
• The model outperformed human experts in both detection and prediction tasks.

Conclusions:

• The proposed deep-learning method effectively detects prostate cancer and predicts Gleason grade using T2-weighted MRI.
• Multicenter validation confirms the robustness of the deep-learning approach.
• Non-invasive Gleason grade prediction via this method can potentially decrease unnecessary prostate biopsies.