The fully automated deep learning system was trained to isolate the ACL and then detect structural abnormalities using multiple convolutional neural networks. The AI was retrospectively tested on the MRIs of 175 subjects with full-thickness ACL tears and 175 with intact ACLs.
The assessments of both the deep learning system and a group of five radiologists were compared to definitive arthroscopic results for each subject. As a result, the researchers reported that there was no statistically significant difference between the performances of either diagnostic method, proving the feasibility of using AI to read imaging data for ACL tears.
More articles about AI:
NIH gives biotech company $300K to use AI to develop hantavirus antibodies
FDA approves AI that detects collapsed lungs in chest X-rays
Machine learning can predict sudden death in patients with heart failure
