Deep Learning for Pneumonia Diagnosis  ,Accomparative Study of Xception,Efficientnetb4 ,and Efficientnetv2s

                                             Vijaya lakshmi                         Pilla G S S R Krishna Patrudu

Assistant Professor &                      M-Tech Final Semester

Head of the Department   ,                 Masters of Technology,

Sanketika Vidya Parishad Engineering College, Vishakhapatnam, Andhra Pradesh, India

ABSTRACT

Pneumonia continues to be one of the most common and life-threatening lung infections around the world, especially among children, elderly people, and individuals with weak immunity. Early and accurate detection is crucial to prevent serious health complications and save lives. Traditionally, doctors diagnose pneumonia by examining chest X-ray images, but this manual process can be time-consuming and sometimes prone to human error—especially in busy hospitals or rural areas with limited medical staff.

To address this problem, our project presents an automated pneumonia detection system using deep learning, a powerful branch of artificial intelligence. The system is trained to analyze chest X-ray images and identify whether a person has pneumonia or not. We used three advanced deep learning models—Xception, EfficientNetB4, and EfficientNetV2S— to build and evaluate the system.

The training was done using thousands of chest X-ray images collected from publicly available datasets. After testing all three models, the Xception model achieved the highest accuracy of 95%, proving to be the most effective in detecting pneumonia. The other two models, EfficientNetB4 and EfficientNetV2S, also performed well but with slightly lower accuracy.

This automated system can assist doctors by providing quick and accurate results, reducing their workload and improving patient care. It is especially helpful in remote and under- resourced areas where experienced radiologists may not be available. The system supports early detection, faster decision-making, and can be integrated into hospital systems or mobile healthcare tools in the future.

Pneumonia is a serious respiratory infection that causes inflammation in the lungs. It mainly affects the small air sacs called alveoli, which fill with pus and fluid, making it difficult for a person to breathe and take in enough oxygen. Pneumonia can be life- threatening, especially for children, elderly people, and individuals with weakened immune systems.

Doctors usually use chest X-ray images to detect pneumonia. However, examining these X-rays manually can be time-consuming, requires expert knowledge, and may lead to errors, especially when large numbers of patients need to be diagnosed quickly. In rural or under-resourced areas, experienced radiologists may not be available, making accurate diagnosis even more difficult.

To solve this problem, our project aims to develop an automated pneumonia detection system using deep learning, a powerful branch of artificial intelligence. Deep learning models, especially Convolutional Neural Networks (CNNs), are excellent at analyzing images and finding hidden patterns that may not be easily visible to the human eye.