Investigation on the Performance of DeiT Model in Medical Image Classification

Effective medical intervention relies on properly distinguishing pneumonia cases through chest X-ray examination. In this study, we assess the performance of the Data-Efficient Image Transformer model for binary classification of pneumonia versus normal cases using chest X-ray images. The model was fine-tuned with different learning rates (0.0001, 0.001, 0.01, 0.1) and epochs (5, 10, 25, 50) to evaluate its effectiveness. Across various combinations, we observed peak performance at 25 epochs with a learning rate of 0.1, achieving an accuracy (90.87%), precision (91.11%), recall (90.87%), and F1-score (90.70%). The model’s Matthews Correlation Coefficient reached 80.49%, and Cohen’s kappa attained 79.95%. The study demonstrates the efficacy of the DeiT model in capturing global relationships within the image data, offering competitive results for pneumonia classification, and highlights its potential for broader application in medical image analysis tasks.