Machine Learning-aided Screening of Lung DIseases in CT images.

Abstract

Chest computed tomography (CT) scan image screening for lung diseases is a laborious and time-consuming technique that can only be handled by qualified radiologists. Due to a shortage of experienced radiologists in Uganda, the workload has increased, which could eventually result in fatal diagnostic errors brought on by exhaustion. Radiologists can assist patients by using machine learning-aided lung disease screening in chest CT scan images to balance the workload and lessen the likelihood of these errors. In this study, we propose an automated decision support system with a convolutional neural network model built on the ResNet50 architecture that receives a chest CT scan image as an input and returns the probability distribution of the possible presence of any of the four categories; covid-19, healthy, lung cancer and pneumonia in the chest CT scan image. A total of 379 chest CT scan images were included in the dataset used to test the model, and they were divided into four categories; covid-19, healthy, lung cancer and pneumonia. The model achieved an accuracy of 96%, average recall of 96.25%, average precision of 96.25% and an average F1 score of 96%.