Artificial Intelligence-Enhanced Identification of Incidental Findings in Prostate MRI

Area of Science:

• Radiology
• Artificial Intelligence
• Medical Imaging

Background:

• Prostate MRI scans often reveal incidental findings outside the prostate gland.
• Automated detection of these findings can improve reporting efficiency.

Purpose of the Study:

• To evaluate the feasibility of an automated AI model for detecting and segmenting common incidental findings in prostate MRI.
• To assess the performance of the AI model across multiple datasets.

Main Methods:

• A retrospective study of 465 prostate MRI exams was conducted.
• An nnU-Net model was trained to segment findings like lymph nodes, hernias, and joint changes.
• Model performance was evaluated using Dice scores, IoU, and radiologist assessments on independent datasets.

Main Results:

• The AI model achieved varying segmentation performance across different findings, with high Dice scores for sigmoid diverticulosis and hydroceles testis.
• Radiologist evaluation showed high agreement with AI-assisted detection of incidental findings.
• Accuracy varied by dataset but was generally high for most findings.

Conclusions:

• The AI model successfully segmented frequent incidental findings in prostate MRI across diverse datasets.
• This technology has the potential to support radiologists in achieving consistent and efficient reporting.
• Further research with larger, diverse datasets is warranted.