For a study, the researchers sought to find how well a CNN model could diagnose cervical ossification of the posterior longitudinal ligament (OPLL). Image analysis for diagnostic purposes was conducted. The study compared 50 patients with cervical OPLL and 50 patients with plain radiographs. Investigators evaluated the area under the receiver operating characteristic curve to assess the CNN model’s performance (AUC). They also compared CNN’s diagnosis sensitivity, specificity, and accuracy with general orthopedic surgeons and spine experts. The gold standard for diagnosis was computed tomography. The CNN model was trained and validated using radiographs of the cervical spine in neutral, flexion, and extension positions. Its architecture was built using the deep learning PyTorch toolkit. The model had a sensitivity and specificity of 80% and 100%, respectively, and was 90% (18/20) accurate. On the other hand, Orthopaedic surgeons had a mean accuracy of 70%, with a sensitivity and specificity of 73% (SD: 0.12) and 67% (SD: 0.17), respectively. The spine surgeons’ average accuracy was 75%, with sensitivity and specificity of 80% (SD: 0.08) and 70% (SD: 0.08), respectively. The model based on radiographs had an AUC of 0.924. In the diagnosis of OPLL, the CNN model demonstrated successful diagnostic accuracy and appropriate specificity.
Source –www.thespinejournalonline.com/article/S1529-9430(22)00007-9/fulltext
