Session 10: Healthcare - \u3cem\u3eAI in Healthcare: Automated Chest X-ray Screening\u3c/em\u3e

Unstructured data, i.e. image is worth a thousand words. Image analysis has several different applications; healthcare, for instance. Fundamental image processing mechanics let us focus on how we can actually represent visual images to be processed in machine learning algorithms. More specifically, the talk aimed to provide how data scientist works with an emphasis on image processing and pattern recognition. In this context, we will present an automatic chest X-rays screening system to detect pulmonary abnormalities using chest X-rays (CXR) in nonhospital settings. In particular, the primary motivator of the project is the need for screening HIV+ populations in resource-constrained regions for the evidence of Tuberculosis (TB). The system analyzes thoracic edge map, shapes as well as symmetry that exists between the lung sections of the posteroanterior CXRs. For classification, we have used several different classifiers, such as support vector machine, Bayesian network, multilayer perceptron neural networks, random forest and convolutional neural network. Using CXR benchmark collections made available by the National Institutes of Health (NIH) and National Institute of Tuberculosis and Respiratory Diseases, India, the proposed method outperforms the previously reported state-of-the-art methods by more than 5% in terms of accuracy and 3% in terms of area under the ROC curve (AUC). On the whole, the talk will consider state-of-theart works in image analysis, pattern recognition and machine learning under the framework of healthcare and/or medical imaging. Having all these topics, we will provide/summarize how AI and machine learning have helped healthcare advance than it used to be