EP52 Artificial Intelligence Accurately Predicts Patient-Reported Outcomes and Cotyloid Synovitis Using Intraoperative Arthroscopic Cotyloid Fossa Images

Abstract

Abstract Introduction Artificial intelligence (AI) has demonstrated potential in predicting patient outcomes following orthopaedic surgeries using intraoperative arthroscopic images. Studies assessing the efficacy of AI trained on intraoperative arthroscopic hip images to predict outcomes following hip arthroscopy are limited. The purpose of the present study was to evaluate the efficacy of AI in predicting patient reported outcomes (PROs) and cotyloid synovitis in patients undergoing hip arthroscopy for labral tears using intraoperative arthroscopic images of the cotyloid fossa. Methods A retrospective review of all patients aged 18-80 who underwent hip arthroscopy by the senior author for femoroacetabular impingement between 01/01/2010 and 05/01/2022 was performed. The primary outcomes were attainment of patient acceptable symptom state (PASS) and minimal clinically important difference (MCID) for postoperative modified Harris Hip Score (mHHS). Intraoperative arthroscopic images of the cotyloid fossa were used for model training. An EfficientNetV2-L model was pretrained on a small set of manually screened images, achieving 94% accuracy in identifying cotyloid fossa images. This model was used to screen 26,000 intraoperative images, resulting in the identification of 1,750 high-quality cotyloid fossa images. An EfficientNetV2-S model was then trained on a portion of these images and tested on the remaining images for different outcomes. The models’ predictions were compared against actual postoperative outcomes to assess accuracy. Results A total of 1,750 images from 742 cases were included. Overall, EfficientNetV2-S demonstrated strong performance in predicting mHHS PASS and mHHS MCID. For predicting mHHS PASS at two years post-operatively, the model demonstrated an accuracy of 0.84 and AUC of 0.88, with a sensitivity of 0.97 and specificity of 0.70. Additionally, the model predicted achievement of mHHS MCID at two years post-operatively with an accuracy of 0.82 and AUC of 0.91, and sensitivity and specificity of 0.94 and 0.74 respectively. The model showed consistent accuracy in predicting the presence of cotyloid synovitis with an accuracy of 0.89, AUC of 0.97, and sensitivity of 0.89 and specificity of 0.94. Conclusion AI and machine learning algorithms demonstrated strong proficiency in predicting PROs and the presence of cotyloid synovitis when trained on intraoperative arthroscopic images of the cotyloid fossa.

Ähnliche Arbeiten