PNEUMONIA DIAGNOSIS USING DEEP LEARNING APPROACH

Abstract Pneumonia has been one of the most popular disease affecting the lungs in Malaysia. Many people have lost their lives to this disease in Malaysia. Currently, experts and doctors use experience and knowledge to interpret X-Ray of patients to detect presence of pneumonia. This method is obviously not fool-proof. The risks that can occur with this current method is misdiagnosis due to human error. Lack of experts to handle the cases around the world has also become a growing issue. Liabilities for doctors and hospitals can be high in case of misdiagnosis and it can be terribly damaging and life threatening for patients. The purpose of this project is to design a deep learning model which can detect pneumonia with high accuracy. There are two types of pneumonia which are common, namely bacterial and viral. Our model should be able to detect the specific types of pneumonia if present, and if not, must be able to detect a healthy X-Ray. This can leverage the advantage of technology to help the medical and healthcare sector, as well as improving human lives. The methodology used is agile methodology to take as-fast-as-possible approach for this project. The data used is taken from Kaggle for proper structured datasets. Our aim is to ensure the model learns the patterns and generalizes its learning compared to memorizing the patterns from training data. The models will be evaluated using a validation set and validation accuracy will be the measure for performance of models. The results of this project should be able to detect presence of pneumonia in a given X-Ray image using deep learning with high accuracy