Development and validation of a multimodal feature fusion prognostic model for lumbar degenerative disease based on machine learning: a study protocol

Introduction Lumbar degenerative disease (LDD) is one of the most common reasons for patients to present with low back pain. Proper evaluation and treatment of patients with LDD are important, which clinicians perform using a variety of predictors for guidance in choosing the most appropriate treatment. Because evidence on which treatment is best for LDD is limited, the purpose of this study is to establish a clinical prediction model based on machine learning (ML) to accurately predict outcomes of patients with LDDs in the early stages by their clinical characteristics and imaging changes. Methods and analysis In this study, we develop and validate a clinical prognostic model to determine whether patients will experience complications within 6 months after percutaneous endoscopic lumbar discectomy (PELD). Baseline data will be collected from patients' electronic medical records. As of now, we have recruited a total of 580 participants (n=400 for development, n=180 for validation). The study's primary outcome will be the incidence of complications within 6 months after PELD. We will use an ML algorithm and a multiple logistic regression analysis model to screen factors affecting surgical efficacy. We will evaluate the calibration and differentiation performance of the model by the area under the curve. Sensitivity (Sen), specificity, positive predictive value and negative predictive value will be reported in the validation data set, with a target of 80% Sen. The results of this study could better illustrate the performance of the clinical prediction model, ultimately helping both clinicians and patients. Ethics and dissemination Ethical approval was obtained from the medical ethics committee of the Affiliated Hospital of Gansu University of Traditional Chinese Medicine (Lanzhou, China; No. 2022-57). Findings and related data will be disseminated in peer-reviewed journals, at conferences, and through open scientific frameworks.