Automation in Pneumonia Detection

• Published in: Handbook of Texture Analysis: AI-Based Medical Imaging Applications: Volume II
• Main Author: Shaharudin N.S.; Hasan N.; Sabri N.
• Format: Book chapter
• Language: English
• Published: CRC Press 2024
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85200192885&doi=10.1201%2f9780367486082-8&partnerID=40&md5=8c429adb0a9631e0ee24d489ea6f4d2b

A significant “death threat” disease in Malaysia, pneumonia has been recorded at the top of the chart after ischemic heart disease, with 11.6%, according to the Department of Statistics Malaysia (DOSM). Its infections are deadly because of its contamination in restricted areas. Accordingly, it leads to difficulty in diagnosing pneumonia because of limited vision. Thus, proper assistance is needed in medical imaging, since the inflammation takes place on a lung or both lungs. Therefore, the model system, “automation in pneumonia detection” is developed using shallow learning approach to utilize the machine learning technique classifying the radiographic images between three classes, normal and pneumonia-infected images, bacterial, and viral, efficiently. Apart from that, the system also supports a cost-effective technique, unlike other researchers. The approach also uses proper texture analysis feature for accurate classification performance, the gray level co-occurrence matrix feature. Thusly, the system is also able to categorize and identify normal and pneumonia infections, viral, and bacterial images successfully, with its final accuracy of 86.67%. Furthermore, diagnosing abnormalities might be time-consuming for the doctors and experts because of its identical appearance; therefore, the model target is to alleviate the drawbacks and support an efficient interpretation of images in medical description visually. © 2024 selection and editorial matter, Ayman El-Baz, Mohammed Ghazal and Jasjit S. Suri; individual chapters, the contributors.