IOS Mobile APP for Tuberculosis Detection Based on Chest X-Ray Image

Tuberculosis is one of the serious illness in the world. This disease still prevalent in underdeveloped and developing countries. To screen tuberculosis, chest X-ray (CXR) and sputum smear have been used widely. Traditionally, both methods have to be managed by doctors and technicians. To improve the effectiveness of loading of mass CXR interpretation sputum smear reading, we therefore developed an algorithm with the aid of artificial Intelligence (AI). We converted CXR and sputum smear into digital images. Subsequently, using image processing methods, Chest X-Ray (CXR) and sputum smear images automatically read. In this abstract, we only report the CXR reading with custom- designed AI analysis algorithm. The tuberculosis infection usually will result in some white dots in CXR. Caffe Frame work with GoogLeNet Network were used to create a model for CXR classification. The datasets were formed by normal CXR, Tuberculosis Suspect CXR and plural effusion CXR images. The GoogLeNet network accuracy was 98.39%. coreML tools was used to convert the Caffe model into coreML model for IOS mobile App. IOS mobile app successfully classified CXR image on IOS devices.