Detection of Pneumonia and Tuberculosis using Deep Learning approach

Pneumonia and tuberculosis are two major problems which can be diagnosed through X-ray images. However, accurate and efficient detection of these diseases is a challenge for radiologists, as it requires careful examination of multiple images. A deep learning approach can help to automate this process and improve the accuracy of disease detection. Specifically, aim to investigate the effectiveness of transfer learning and the ResNet50 model for this task. A publicly available dataset of chest X-rays from Kaggle is used. The dataset includes images from patients with and without pneumonia and tuberculosis. Transfer learning with the ResNet50 model a multilayer CNN model along with a residual block is used, for effective image classification. Model is trained using fine tuning techniques. For quantitative treatment, accuracy is used as the main metric for evaluating the model's performance, and also used precision, recall, and F1 score to assess the model's ability to detect cases of pneumonia and tuberculosis. The accuracy of the model is 96.59% for detecting cases of pneumonia tuberculosis. Additionally, the precision, recall, and F1 score for both diseases were above 0.90. These results demonstrate the effectiveness of transfer learning and the ResNet50 model for disease detection in chest X-rays.