Evaluating the performance of commercial liver MRI AI software in detecting malignancy in post-treatment lesions.

Motivation: Inexperienced junior radiologists faced difficulty in identifying the activity of liver post-treatment lesions with multiple therapeutic techniques. Goal(s): Commercial liver MRI AI software is promising in improving the accuracy of junior radiologists in judging the activity of lesions after treatment. Approach: Two senior radiologists used the 5-point scale to evaluate the malignancy of liver lesions as the reference standard. A junior radiologist was evaluated without and with the assistance of AI software to test the diagnostic performance. Results: AI software performs better in sensitivity and negative predictive value. With the help of AI, the diagnostic efficacy of junior radiologists has been significantly improved. Impact: Accurate identification of liver lesion malignancy is essential for determining effective treatment regimens. AI software can support junior radiologists in assessing malignancy in post-treatment lesions, regardless of the familiarity with specific treatment techniques.