No abstract

The chest X-ray is a quick and effective test that has been useful for decades to help doctors to view vital organs. When focused on the chest, it can help spot abnormalities or diseases f the airways, bones, heart and lungs. The chest X-ray dataset currently available consists of details regarding 14 diseases. The lack in publically available datasets creates a difficulty in providing more Computer Aided Detection(CAD) in real world medical science with chest X-rays. In this paper we present the NIH chest X-ray dataset comprised of 112,120X-ray images with disease labels from 30,000 unique patients. The labels are expected to be more than 90% accurate and we perform weakly supervised learning with this dataset. The deep convolutional neural network is used to recognize and locate the common disease patterns and we try to integrate the automated detection system with normal hospital management system.