Pneumonia detection on chest X-ray images using segmentation technique and modified AlexNet-ML classifiers

Worldwide, pneumonia is a serious health issue, and detecting pneumonia quickly and accurately is necessary for improving patient outcomes. This method seeks to save human labor, increase diagnostic efficiency, and eventually enhance patient outcomes in the healthcare system by automating the pneumonia detection process. This study aims to develop a robust pneumonia detection system using modified deep learning techniques followed by different classifiers such as kNN, DT, RF, and SVM applied to chest X-ray images. In the planned model, the images are processed beforehand to enhance the quality and clarity of the image for accurate prediction and are segmented utilizing the GrabCut algorithm to eliminate unwanted portions from the image. In this work, features are extracted utilizing a modified AlexNet architecture, and pneumonia disease is classified into pneumonia and normal images using different classifiers. The proposed method achieved optimal performance based on the testing accuracy, testing sensitivity, and testing specificity of 96%, 95.59%, and 96.15%, respectively, using the chest X-ray image dataset. According to the testing results, the most effective approach among the current methods with the highest accuracy was the combination of modified AlexNet models and the SVM classifier.