Figure 1

Workflow of the multi-pose model to classify severe lumbar spinal stenosis (LSS). The first part in the convolutional layer part is for the extraction of spatial features extracted by three pre-trained single-pose-based convolutional neural network models with non-trainable parameters, and the second part in multi-layer perceptron (MLP) is for classifying patients using the concatenated features. H&E Histogram equalization.