Pneumonia detection: A comprehensive study of diverse neural network architectures using chest X-rays
Wajahat Akbar, Abdullah Abdullah, Altaf Hussain, Tariq Hussain, Farman Ali, Muhammad Inam Ul Haq, Raaz Waheeb Attar, Ahmed Alhomoud, Ahmad Ali AlZubi, Reem Alsagri
Abstract
<abstract xmlns="http://www.w3.org/1999/xhtml"> Pneumonia is of deep concern in healthcare worldwide, being the most deadly infectious disease, especially among children. Chest radiographs are crucial for detecting it. However, certain vulnerable groups exhibit heightened susceptibility, emphasizing the critical nature of accurate diagnosis and timely intervention. This paper presents convolutional neural network (CNN) models for the detection of pneumonia from chest X-rays images. Among 20 different CNN models, we identified EfficientNet-B0 as the most accurate and efficient, boasting an impressive accuracy rate of 94.13%. Furthermore, the precision, recall, and F-score metrics for this model stand at 93.50%, 92.99%, and 93.14%, respectively. This research underscores the potential of CNNs to revolutionize pneumonia diagnosis. </abstract>