Enabling Accurate Tuberculosis Diagnosis through Deep Learning on Patient CXR Images

Tuberculosis stands out as a severe and deadly infectious disease globally, characterized by a high incidence rate and mortality. While common tuberculosis screening methods involve diagnosing through CXR images, this approach carries a significant risk of misdiagnosis. The recent emergence of deep learning models has opened new avenues to tackle this issue. This paper is dedicated to exploring the utilization of deep learning models for tuberculosis screening using CXR images. The primary objective of this article is to leverage the latest advancements in deep learning to craft DNN network models and facilitate feature integration. This integration is based on the image features extracted by the DNN models. Furthermore, the article includes performance testing on the publicly accessible Shenzhen dataset. The experimental findings underscore the accuracy and reliability of our method in diagnosing tuberculosis with precision through CXR images. This conclusion holds immense significance for potential clinical applications.