Pneumonia Detection using Machine Learning

Over the last few years, strides made in artificial intelligence have greatly improved the identification of chest X-ray pneumonia. This paper examines and contrasts multiple approaches where contemporary neural networks and data-centric methods have been employed to enhance diagnostic precision. These approaches feature a combination of EfficientNetB7 and Squeeze-and-Excitation modules, a PD-Net variant based on ResNet50, and a YOLOv8 system reinforced by the generation of synthetic images using FILM and PCA methods. Every strategy has its own merits: EfficientNetB7’s feature extraction skill is unparalleled, the balance YOLOv8 maintains between the quality of detection and the speed of processing is commendable, and ResNet50 is a solid transfer learning backbone tailored for hospital settings. This​‍​‌‍​‍‌ review additionally stresses the requirement for balanced datasets, the application of explainable artificial intelligence (AI) techniques like Grad-CAM for model debugging, augmented data, and explainable AI frameworks to provide more transparency to the end-users, the clinicians. The integration of diverse and interpretable data with computational efficiency is highlighted by this study as the main factor for reliable AI systems for the detection of pneumonia in vivo clinical ​‍​‌‍​‍‌settings.