Detection of Abnormalities in Spinal Cord Injury MRI Using Image Processing and Machine Learning

To enhance the analysis of radiological spine imaging and improve clinical decision-making for traumatic Spinal Cord Injury (SCI), existing Machine Learning (ML) algorithms require refinements in accuracy and clinical applicability. This study implements Deep Learning (DL) models for the automated detection of SCI in MRI images resulting from trauma, such as falls and accidents. A retrospective analysis was conducted on 890 MRI images categorized into four severity levels: Highly Severe (A), Less Severe (B), Mildly Severe (C), and No Injury (D). The DL framework utilizes EfficientNet-B0, Swin Transformer V2-S, and DenseNet121 architectures. Multi-class prediction models were developed for the four considered severity grades, and the results demonstrate that the proposed approach outperforms state-of-the-art methods across several metrics, including Accuracy, F1-Score, Precision, Recall, Jaccard Index (IoU), and Dice Coefficient.