Multi Modal Smart Diagnosis of Pulmonary Diseases

Abstract

Health is an outfit that looks different on everybody. Lively health and factors on their counterpart have diseases and cures. There is a wide range of ailments, some of them being chronic and requiring timely treatment. On the grounds of the human body segments different maladies such as coronary, pulmonary, neurological disorders, and many more can be caused. Pulmonary diseases affect the lungs causing obstruction in the airflow. Pulmonary illness requires continuous monitoring of the victim under the supervision of a medical expert for a reasonable duration in order for it to be cured. Contacting the medical practitioner will not always be within boundaries of reach and timely, therefore there is a need for computerization and mechanizing the Screening Process of Pulmonary diseases namely covid, lung cancer, pneumonia, and tuberculosis(TB), and providing a pre-consult notice to the sufferer. Techs such as machine learning(ML) and deep learning(DL), mainly autoencoders(AE) along with flask to support a web interface are used. The research is distinct for it takes into consideration four different diseases and draws conclusions based on symptoms and medical scans. KNeighbors(KNN) model on symptoms gave a test accuracy of around ninety-four percent. AE on computed tomography(CT) scan and chest X-ray(CXR) has a test accuracy of about eighty-eight and ninety-two percent respectively.