Multicenter study on predicting postoperative upper limb muscle strength improvement in cervical spinal cord injury patients using radiomics and deep learning

This study uses MRI scans and machine learning to predict how well patients with cervical spinal cord injuries will recover muscle strength in their upper limbs after surgery. The method, called radiomics, involves extracting detailed information from MRI images and using deep learning to analyze this data and make predictions. The study found that combining radiomics with a deep learning model called ResNet34 was particularly effective in predicting patient outcomes.

• 1
  A combined model using ResNet34 and radiomics demonstrated excellent performance with an accuracy of 0.800 and an AUC of 0.893 in predicting long-term motor function outcomes.
• 2
  Deep learning features were found to be more significant than radiomics features in the combined model.
• 3
  The Grad-CAM visualization method highlighted that regions mainly concentrated on the edematous spinal cord, deformed vertebrae, and even the edematous ligaments are critical for prognosis prediction.