Establishment of a Diagnostic Model for Hidden Fractures of the Knee Joint Based on EasyDL
Abstract：Objective To establish a diagnostic model for hidden fractures of the knee joint using the artificial intelligence open platform EasyDL, and to evaluate the model. Methods A total of 98 patients with hidden fractures of the knee joint treated at the 964th Hospital of the Joint Logistics Support Force from October 2022 to May 2024 were selected, including 65 males and 33 females, with an average age of (32.5 ± 15.8) years, ranging from 6 to 87 years. A total of 1,482 magnetic resonance imaging (MRI) images were selected and divided into four categories: sagittal T1-weighted images, sagittal T2-weighted images, coronal T2-weighted images, and axial T2-weighted images. Each category of image was randomly assigned, with 80% used as the training set and 20% as the test set. The model's accuracy, recall rate, and F1 value were evaluated. Additionally, the diagnostic results of the test set were assessed by attending physicians and resident physicians, and the consistency between the model's diagnosis and that of physicians of different levels was evaluated. Results The overall accuracy of the model was 84.4%, the recall rate was 94.8%, and the F1 value was 89.3%. The Kappa values between the AI diagnostic model and attending physicians was 0.717, and between the AI diagnostic model and resident physicians was 0.527. Conclusion The AI diagnostic model established based on the EasyDL platform has strong detection capability and balance, and its diagnostic ability for hidden fractures of the knee joint is roughly equivalent to that of attending physicians.
Keywords：EasyDL platformKnee joint occult fractureMagnetic resonance imagingDiagnostic model
Publication Date：2025-07-25
Online Publishing Date：2025-08-26(First online date of this platform, not the publication date of the document)
Pages：4( 296-298,320 )