Automated Arthritis Detection Using Artificial Intelligence on Smartphone Photographs

Over the past 4 years the investigators have aimed to help the early detection of arthritis leveraging artificial intelligence. This project aims to detect arthritis based on smart phone photographs of joint areas that make it scalable and available in the community. This group first developed a compelling proof-of-concept pipeline and models using 100 patients. (published in Frontiers in Medicine, Nov 2023, wherein they demonstrated that this technology works with reasonable accuracy in the lab, viz Technology Readiness Level currently stands at 3-4). They followed with a newer paper (submitted for publication, available on preprint server MedRxiv) that trained two different CNNs, a screening CNN on uncropped hands that distinguishes patients from controls followed by joint specific detections.

The system involves supporting infrastructure that will enable efficient detection of arthritis. This includes

• Collection of photos in a standardized manner using custom designed boxes
• Using and testing a browser pipeline
• The CNN models will be trained on the dataset of photographs taken in this and results will be deployed to doctors in the community. This ensures a doctor in the loop that can later take action on the results for further confirmatory tests or management.
• Understanding knowledge, attitude of patients and doctors towards AI in clinical decision making algorithms

This is a Prospective, non-interventional study and this project only involves an investigator taking a smartphone photograph of some joint areas kept in standardized positions. This involves no risk to the patient.

Inclusion criteria

• Inflammatory arthritis of any etiology

Exclusion criteria

• Severe deformity that hampers standardization of photographs