Automatic measurement and detection of knee osteoarthritis characteristics using ultrasound - a systematic review

Knee osteoarthritis (KOA) is a degenerative joint disease, involving various knee tissues. As it is widely recognized that treatment is most efficient in an early-stage of the disease, early diagnosis of KOA is of crucial importance. Early identification of KOA, however, has been difficult due to limitations of common clinical diagnostic methods. Ultrasound offers a solution, but acquisition and interpretation of US images requires extensive training. Automated detection and measurement of KOA-related characteristics could be of added value. This review provides an overview of the available literature on automatic measurement and detection of KOA characteristics using ultrasound. IEEE Xplore and PubMed were searched for relevant publications, followed by iterative forward and backward citation search to identify all relevant literature. One author screened for relevant publications and extracted data from included publications. The literature search in this study resulted in 10 papers. Although there is increasing recognition that KOA involves multiple knee structures rather than only cartilage, most studies still focused on cartilage (8/10 publications). Five papers present a fully-automatic algorithm, of which four were artificial intelligence (AI) based. The remaining five publications describe a semiautomatic method. Study population sizes ranged from one to 29, and only one study included four KOA patients. The small study sizes and lack of patient data make it impossible to draw conclusions on the generalizability and clinical relevance of the proposed methods. Future research should expand the focus from cartilage to multiple KOA-related characteristics and to inclusion of more subjects, including patients with KOA.