Aussie app assesses hand arthritis as accurately as a rheumatologist


An Australian-devised smartphone app can identify osteoarthritis, rheumatoid and psoriatic arthritis in patient hands with accuracy similar to rheumatologists in less than three minutes, its developers say.

The HandScan smartphone app uses machine learning to assess photos of hands and survey data provided by patients to screen for arthritis diagnoses.

The app was subject to evaluation by rheumatologists at seven private practices across Australia in a study that showed consistent results of 95% accuracy for the machine learning algorithm compared to expert rheumatologist diagnosis.

“This could be used to assist primary care physicians in the assessment of patients presenting with hand arthritis, and has the potential to improve the clinical assessment and management of such patients,” the evaluation published in BMC Musculoskeletal Disorders concluded.

“It can provide a point-of-care result to primary care physicians (PCPs), reducing the need for separate investigations, including blood tests and imaging. This approach could improve accuracy compared to traditional screening methods, and significantly reduce patient and healthcare costs.”

The app took a median time of two minutes and 59 seconds — with a range from two minutes and 40 seconds to three minutes and 35 seconds — to capture the data and determine a screening result.

One of the creators of the app, the University of Notre Dame Musculoskeletal Discipline Leader, Dr Mark Reed, said the target users would be GPs in Australia and primary care physicians internationally.

“It compared favourably to diagnostic methods in primary care, namely blood test and imaging results,” he said.

“Both the physical appearance and symptom responses help to differentiate the different forms. It can also identify combined diagnoses, when patients with existing osteoarthritis develop superimposed inflammatory arthritis.”

Diagnostic accuracy and type of arthritis

The study classified accuracy as the total proportion of all correct model predictions, precision was the proportion of positive diagnosis predictions that were positive, and recall was the proportion of all the positive diagnoses that the app was able to identify.

The specificity indicator revealed the proportion of negative diagnoses that the app identified correctly.

For rheumatoid arthritis, the app achieved accuracy of 85.1%, and precision, recall and specificity of 85.1, 80.0, 88.1 and 82.7%. The corresponding results for psoriatic arthritis were 95.2, 76.9, 90.9 and 95.8%, and for osteoarthritis were 77.4, 78.3, 80.6 and 73.7%.

A pilot study of the app previously found the speed at which it could assess arthritis would be a benefit.

Dr Reed said HandScan was currently under consideration by the TGA as a Class I device, and it was hoped the app would be available for general access later this year.

Further testing with GPs is underway.

Already a member?

Login to keep reading.

OR
Email me a login link
logo

© 2022 the limbic