LOCALIZATION OF KNEE AREA AND OSTEOARTHRITIS CLASSIFICATION USING CNN ON LEG X-RAY IMAGES

• Main Author: Amira, Rizky
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/63032
• Institution: Institut Teknologi Bandung

ABSTRACT LOCALIZATION OF KNEE AREA AND OSTEOARTHRITIS CLASSIFICATION USING CNN ON LEG X-RAY IMAGES By Rizky Amira NIM: 18316016 Undergraduate Program in Biomedical Engineering Osteoarthritis is the most common joint disease. It occurs when the ligaments and muscles that protect the bones are damaged, which can lead to cartilage breakdown. Knee osteoarthritis diagnosis is conducted by evaluating radiographs, where osteophyte, joint space narrowing, and sclerosis can be seen in the image. This process, however, can suffer from subjectivity, which can lead to diagnosis inconsistency, especially in the early stages of osteoarthritis, where discrimination of osteoarthritis severity only depends on a single feature, that is the existence of osteophyte. This problem can be solved by computational algorithms, which can perform an objective assessment. Therefore, localization and classification systems are needed to help physician evaluate radiographs. In this study, we will perform localization and osteoarthritis classification using CNN on 2263 leg x-ray images from OAI dataset. After performing localization, we will get 4526 joint area images with the size of 150×150. The grade of osteoarthritis would be known from the image. In this study, the accuracy of the localization system is 92,07% and the accuracy of the classification system is 72,77%. Keywords: Osteoarthritis, localization, classification, CNN, x-ray.