Multi-pose-based Convolutional Neural Network Model for Diagnosis of Patients with Central Lumbar Spinal Stenosis

Although plain radiographs have declined in importance since the advent of magnetic resonance imaging (MRI), their diagnostic ability has improved dramatically when combined with deep learning. Previously, we developed a convolutional neural network (CNN) model using a radiograph for diagnosing lumbar spinal stenosis (LSS). In this study, we aimed to improve and generalize the performance of CNN models using multi-pose radiographs. Individuals with severe or no LSS, confirmed using MRI, were enrolled. Lateral radiographs of three postures were collected. We developed a multi-pose-based CNN (MP-CNN) model using four pre-trained algorithms and three single-pose-based CNN (SP-CNN) using extension, flexion, and neutral postures. The MP-CNN model underwent additional internal and external validation to measure generalization performance. The ResNet50-based MP-CNN model achieved the largest area under the receiver operating characteristic curve (AUROC) of 91.4% (95% confidence interval [CI] 90.9–91.8%). In the extra validation, the AUROC of the MP-CNN model was 91.3% (95% CI 90.7–91.9%) and 79.5% (95% CI 78.2–80.8%) for the extra-internal and external validation, respectively. The MP-based heatmap offered a logical decision-making direction through optimized visualization. This model holds potential as a screening tool for LSS diagnosis, offering an explainable rationale for its prediction.