Computer algorithms to diagnose juvenile arthritis

12 March 2019

A team of scientists at VIB and KU Leuven developed a machine learning algorithm that can diagnose arthritis in children with almost 90% accuracy, purely on the basis of a blood test. The new findings open the door to the use of machine learning for faster diagnosis and to predict which juvenile arthritis patients will react most favourably to various forms of treatment.

 

 

 

Kind toont vingers: artritis?

The algorithm was approximately 90% accurate in identifying which children suffered from arthritis.

 

 

 

 

Juvenile idiopathic arthritis is the most common rheumatoid disorder in children, but the symptoms, severity and development can vary considerably. This diversity makes it difficult to categorise patients early on and to decide on the optimum treatment. To be able to improve the diagnosis and treatment, a team of researchers at VIB, KU Leuven and UZ Leuven charted the immune systems of hundreds of children, with and without juvenile arthritis, in great detail.

Improving treatment options

"We took blood samples from more than 100 children, two thirds of whom suffered from juvenile arthritis," Erika Van Nieuwenhove (VIB-KU Leuven), the first author of the research, explained. "We analysed the children’s immune systems in more detail than ever before and subjected the data to machine learning algorithms."

The results were remarkable: the algorithm was approximately 90% accurate in identifying which children suffered from arthritis. "We based our findings solely on information concerning the immune system, excluding symptoms or other clinical data," stated Professor Adrian Liston (VIB-KU Leuven and the Babraham Institute, Cambridge, UK). "The results demonstrate that analysing immunological parameters, in combination with machine learning, offers huge potential in terms of the early diagnosis of various types of juvenile idiopathic arthritis. "

The researchers are hopeful about the impact of this research and hope that it will enable them to improve treatment options for patients going forward. "The tool still has to be validated further, but there are no other scientific obstacles when it comes to translating this approach quickly into clinical practice," Prof. Dr. Carine Wouters (UZ Leuven), who was in charge of the clinical part of this study, affirmed. "In the long term we would be able to use this kind of detailed classification information and machine learning analysis to identify which patients would react most favourably to specific treatments."

The tool still has to be validated further, but there are no other scientific obstacles when it comes to translating this approach quickly into clinical practice
Prof. dr. Carine Wouters - paediatric reumatologist

The researchers are hopeful about the impact of this research and hope that it will enable them to improve treatment options for patients going forward. "The tool still has to be validated further, but there are no other scientific obstacles when it comes to translating this approach quickly into clinical practice," Prof. dr. Carine Wouters (UZ Leuven), who was in charge of the clinical part of this study, affirmed. "In the long term we would be able to use this kind of detailed classification information and machine learning analysis to identify which patients would react most favourably to specific treatments."

Last edit: 5 May 2020