Artificial Intelligence detects ostearthritis

A recently presented study gives insight as to how artifical intelligence can identify patients with osteoarthritis using specifically trained disease classifiers.

An artificial-intelligence-driven algorithm developed by a group of researchers from Austria and Portugal outperforms the available state-of-the-art in identifying patients affected by osteoarthritis based on conventional X-rays, a recently presented study finds.

The algorithm — which offers a classifier score of 0.85 as measured by ROC AUC — could be particularly useful in an automatic identification of patients with different stages of knee osteoarthritis, said the lead author Dr. Hladuvka, a computer scientist and mathematician at Vienna-based VRVis Research Center.

“Being able to tell automatically from X-rays which individuals are affected by the diesease and which ones are not is a significant milestone for artificial intelligence in medcial imaging”.

The artificial intelligence program that the involved teams have developed could help physicians identify and assess the progression of the disease as well as potential early signs of osteoarthritis onset. Creating an effective AI algorithm involves three main steps: writing the software, training it and then testing it to see how well it works, the researchers said.

The initial step was to develop a robust semi-automatically placed layout for regions of interest (ROI), followed by computing different texture parameters in each ROI, and employ statistical and machine learning methods to evaluate feature combinations. Based on 153 high-resolution radiographs of a 10-year Portugeses study, the results identify medial femur as an effective univariate descriptor, with significance comparable to medial tibia. A linear five-feature classifier combining femur/tibia texture descriptors, achieves AUC of 0.85, outperforming the state-of-the-art.

“The radiologist or orthopedic can spend more time on deciding the appropriate management option than identifiying patients” said co-lead study author Dr. Richard Ljuhar (previoulsy part of Braincon Technologies Research).

The authors designed the algorithm to not only identify patients with or without osteoarthritis, but to also use it for the identification of patients at risk for developping the disease prior to any visible radiographic signs.

Such a study has been already completed and presented at this years ASBMR in Denver, Colorado.

Read the full paper here.