A Time-dependent Machine-Learning-based Prediction Model for Progression of Knee Osteoarthritis

In the current hospital flow, it usually takes 2 years or even more for patients to queue up for treatment from public hospitals. The lack of self-management practices drives them to be dependent on the non-personalised KOA treatment plans designed by doctors, they are formulated according to the patient’s current severity without considering future progression, and this has led to poor treatment outcomes. For example, a patient with low level of severity but high progression risk in the future two years will only be classified as a low priority group by the doctors under current diagnostic procedure, so they can’t receive proper treatment before their knee conditions deteriorate.

In light of this, our AI is developed on a large database from the U.S and will further include the Hong Kong database to customise to the local population. It leverages both clinical record and medical imaging data for a holistic multimodal analysis of the progression risk. We aim to implement the system at both the community and hospital level by taking different approaches. In the community level, our target customer will be the family doctor, we hope to provide the locals with fast and low-cost preliminary screening and achieve patient self-management. While for the specialised clinic and hospital level, the target customers are public and private hospitals, our main objectives will be the construction of an effective triage system and the facilitation of personalized treatment.

Team Member(s)

LI Ho Hin Toby
CHAN Lok Chun
Wong Ka Yan
Ng Ching Yee