Chest Disease Classification Based on Deep Convolution Neural Network (DCNN)

Abstract

Abstract— In this paper, a robust method for diagnosing chest illness from chest X-ray images is designed and implemented utilizing deep learning networks. The goals of the suggested system are to ensure that patients receive the appropriate treatment at the appropriate time, enable prompt and accurate diagnosis, and stop a patient's health from declining. We developed a neural network approach that fine-tunes convolutional layers and hyperparameters to categorize chest diseases into 14 groups. Using a U-net model for lung separation and a Res-Net50 for chest diagnosis, our approach can distinguish between 14 categories of chest illnesses and normal patients. The proposed method showed exceptional diagnostic efficiency with a 92.71% classification model accuracy and a 97% segmentation model accuracy across all classes.