Chest X-ray Based Pulmonary Disease Classification Using Transfer Learning and CNN

Pneumonia and Tuberculosis (TB) are the life threatening diseases caused by viral or bacterial infection of the lungs and leads to the severe abnormality of the lungs. Pneumonia infected people, who have direct contact with those having TB are at the highest risk of contracting pulmonary pneumonia tuberculosis. Early diagnosis of the pneumonia avoids the development of TB in the lungs by infection. Therefore, early diagnosis of the diseases results in successful treatment of the disease caused to the patient. Chest X-ray is the most commonly used medical imaging technique to identify the pulmonary diseases. However, lack of radiologists and pulmonologists reduces the applicability of the technique. To overcome such issues, non clinical methods such as machine learning and deep learning techniques plays a vital role in classification of pulmonary diseases. In this work, classification and prediction of pulmonary diseases using Transfer learning and CNN is proposed. The experimentation is performed by using Chest X-ray database for the classification of two types of diseases namely Pneumonia and Tuberculosis further also identify the normal lung. The proposed work is faster and employs multi-class classifier viz. transfer learning and CNN for the lung diseases classification. The MobilenetV2 transfer learning technique outer performs than convolutional neural networks. The MobilenetV2 achieved a promising performance with an accuracy of 97%.