Abstract
<jats:p>Multiple sclerosis (MS) is a chronic neurological disease whose diagnosis typically relies on resource-intensive and partly invasive procedures such as MRI and spinal fluid analysis. This study explores whether deep learning can be used to identify RBC-related changes in MS to help understand the disease process better and identify new treatment modalities. An augmented dataset of 1,944 images, comprising samples from 13 subjects (healthy controls and MS-affected), was used. Four deep learning architectures, VGG16, DenseNet201, MobileNetV2, and the Swin Transformer, were evaluated using subject-level Leave-One-Out Cross-Validation (LOOCV). The Swin Transformer achieved the highest mean accuracy of 73.53%, followed by the fully fine-tuned VGG16 at 71.02%, though a paired t-test indicated no statistically significant difference between them (p = 0.66). Explainable AI techniques (Grad-CAM and SWTformer-v1) revealed that VGG16 focused on broader structural patterns in cell groupings, while the Swin Transformer attended to more localized regions. These findings suggest that deep learning can detect subtle morphological patterns in RBCs associated with MS, and model interpretability shows that MS-related disease cues may reside in cell clustering rather than in individual cell morphology.</jats:p>