LUNG DISEASE DETECTION

Lung disease including pneumonia, tuberculosis, chronic obstructive pulmonary disease (COPD), and lung cancer account for significant risks of morbidity and mortality globally. Generally, early recognition and diagnosis will improve treatment outcomes and costs to the health service systems. Typically, identification and diagnosis consist of traditionally diagnostic testing, primarily chest x-ray and computerized tomogram (CT). These tests also extenuate significant reliance upon the experience of the Radiologists and the interpretation of these tests is completely reliant upon human expertise and error, which is time-consuming. Recently exponential advancements in deep learning (specifically Convolutional Neural Networks (CNN)) have been shown to extensively automate the detection of lung diseases with nearly perfect accuracies. Because CNN will enable automated extraction and learning of complex features from various medical images, practitioners will benefit from decreased time-consumption and erroneous assessments leading to vastly improved patient diagnosis and care. The introduction of lung disease detection in daily practice is exceedingly accessible using deep learning techniques by applying image pre-processing, features extraction, and classification. Experimental results in this paper reported improved accuracy from the deep learning models compared to traditional methods in chest x-ray diagnosis using four categories of lung disease. These large scale models can provide a robust, faster, and non-inferior diagnostic support model in lung disease detection and management, with the capacity for ongoing updates and developments, a considerably cheaper option over subsequent medical imaging, and is ultimately transportable and scalable for integration into a clinical decision support system.