Smarter AI Enhances Spinal Cord Injury Detection from CT Scans

Spinal Cord Injury (SCI) represents a significant clinical challenge, often triggering a cascade of secondary health issues such as blood pressure fluctuations and impaired temperature regulation. While machine learning offers diagnostic hope, earlier models struggled with inconsistent imaging data. Addressing this, researchers have introduced a novel deep learning solution: the Siamese Convolutional WideRes Network (SCWRes-Net).

The process begins by feeding Computed Tomography (CT) images into the system. A specialised tool, the Mask Regional Convolutional Neural Network (MRCNN), isolates the spinal cord from the rest of the image, followed by an active contour approach for precise disc localisation. The system then extracts critical geometric and connectivity features from these isolated areas.

By integrating two distinct neural network architectures—Siamese and Wide Residual networks—the SCWRes-Net model effectively overcomes technical hurdles like class imbalance. The results are promising; the model demonstrated an impressive accuracy of 92.56% and a True Positive Rate of 93.33%, marking a significant step forward in automated medical imaging analysis.