Analysis of EfficientNet Family Models by Retraining for Tuberculosis Detection from Chest X-Ray Images

This article presents a comprehensive study on the retraining of neural networks from the EfficientNet family, focusing on the detection of Tuberculosis (TB) through analysis of Chest X-Ray images. Leveraging the Shenzen and Montgomery Chest X-ray Dataset, impressive performance is showcased by the retrained models. Notably, the EfficientNet-B7 model emerges as the most effective, achieving remarkable rates of accuracy, sensitivity, specificity, and F1 score of 98.5%, 98.5%, 99%, and 98% respectively. The potential of EfficientNet architectures in medical image analysis is underscored by the results, particularly in the context of TB detection. A contribution is made to the growing body of knowledge concerning the application of deep learning techniques to enhance disease diagnosis through non-invasive imaging methods, thereby paving the way for improved early detection and patient care.