A Deep Transfer Learning Approach for the Detection of Silicosis Using Chest Radiography

Silicosis is a fibrotic lung disease which is caused by inhaling high amount of silica dust. The silica particles deposit in the lungs causing scarring. This prevents the scarred tissue from absorbing oxygen. The impact of the silica dust persists even after the exposure of the silica dust ceased. It cannot be cured, but when found in early stages, appropriate treatment can be given to the patients which can extend their lifetime. So, detecting them at an earlier stage is very important. In this work, we are using transfer learning techniques to create a model that can classify silicosis infected lung images and normal lung images. When feature extraction model was created from the ResNet model, it produced a high accuracy of 99.68 percentage on the training images.