Detection of knee osteoarthritis and prediction of its severity using x-ray image analysis and patients assessment data: a hybrid design

• Main Authors: Mohafez, Hamidreza, Sayed, Hamza, Hadizadeh, Maryam, Wee, Lai Khin, Ahmad, Siti Anom
• Format: Conference or Workshop Item
• Published: Springer Science and Business Media Deutschland GmbH 2022
• Subjects: TA Engineering (General). Civil engineering (General)
• Online Access: http://eprints.um.edu.my/43381/; https://www.scopus.com/inward/record.uri?eid=2-s2.0-85129327174&doi=10.1007%2f978-3-030-90724-2_16&partnerID=40&md5=044b997dbc2534b3af7935cb07d17357

Knee osteoarthritis (KOA) is a degenerative disease associated with cartilage loss, causes limitations in the range of movement, and known to be one of the most disabling age-associated diseases around the world. It is vital to predict its presence and severity at early stage to tailor the interventions and treatments properly. Traditionally, X-ray Images are graded by radiologists to quantify KOA severity; however, this approach suffers from high levels of subjectivity due to the semi-quantitative nature of grading systems. Numerous attempts have been made to recruit automated X-ray image analysis to quantify KOA severity, but few studies have used pertinent assessment data such as symptoms and medications being used to establish accurate predictive model. So, we proposed a statistical model built on combination of features extracted from X-ray images and patients’ data using ordinal regression analysis. The results revealed that the developed model based on combination of KOA X-ray key features and patient assessment data is able to predict the severity of KOA with high level of accuracy (89.2) and acceptable level of inter-rater reliability with quadratic weighted Cohen’s Kappa coefficient (QWK) of 0.8337. The study outcomes suggested that variables showing impaired knee functions are the best indicators to quantify knee OA presence and severity that may be used in conjunction with X-ray biomarkers for developing intervention and targeted treatment. © 2022, Springer Nature Switzerland AG.