University of Toronto AI Helps Doctors Diagnose Rare Diseases in Children and Adjust Treatment Plans

by Super Neuro


Scene description:For juvenile idiopathic arthritis, a rare disease that only occurs in children, researchers at the University of Toronto used machine learning to accurately identify more than half of the target children, allowing them to receive better treatment options.


Keywords:Machine learning, medical assisted diagnosis

Arthritis is a common chronic disease. On average, one in 5-6 people will suffer from arthritis. Usually, only the elderly and middle-aged people with unhealthy lifestyles will suffer from arthritis.

But there is a rare type of arthritis called juvenile idiopathic arthritis (JIA) that only affects children.

JIA is an autoimmune disease, meaning that the immune system attacks its own body components due to misjudgment. JIA patients usually develop the disease before the age of 16, with the peak incidence concentrated between the ages of 5 and 7, and most of them are boys.

When children become ill, their hand and knee joints will swell and bend, seriously affecting their growth and development, and even causing high disability and mortality rates.

Unfortunately, the cause of JIA is still unclear, and there is no effective treatment method so far.

Medicine can’t solve it yet, let computer science give it a try

When it comes to arthritis, even experienced doctors cannot predict the course and severity of the disease.

With traditional treatment options, children need to use antibiotics for a long time to control inflammation, but the drugs have significant side effects and can cause drug resistance.

Children first take anti-inflammatory analgesics, such as ibuprofen, and then need to use strong antibiotics, including methotrexate (chemotherapy agent), steroids and other biological agents to suppress the immune system that has made a misjudgment. However, the harm caused by long-term use of antibiotics can also damage the immune system and cause more complications.

Although JIA is a complex disease, some patients are diagnosed withOligoarticular JIA: As people age, their symptoms will gradually improve or even disappear.This type accounts for about 50% of all JIA patients, and they are also the luckiest group.

However, even experienced doctors cannot accurately predict the development stage and severity of JIA, so in this process, the problem of overtreatment is inevitable.

It has become an important but difficult task to allow children with oligoarticular arthritis who show a trend of natural remission to stop excessive hormone treatment as early as possible. However, machine learning has recently found a breakthrough.

The promise of machine learning

Due to the complexity of the disease, the degree to which multiple joints are affected, changes over time, and the limited amount of available patient data, an approach superior to traditional models must be used for accurate analysis.

A research team from the University of Toronto has used machine learning to successfully provide good recommendations for medical visits, and their research results were published in the journal PLOS Medicine.

In the study, they used a「Multi-layer non-negative matrix classification」The machine learning technology can learn the patient's pattern information from the data and correctly classify and determine which children have oligoarthritis who can recover naturally.

To do this, they analyzed clinical data collected from all children between 2005 and 2010, in which all children underwent detailed physical examinations and were used as the basis for the analysis.

This includes recording the location of painful joints in the body, also known as "mobile joints," and the connection between mobile joints and symptoms.

The data includes seven major joint movement modes:The pelvic area, fingers, wrists, toes, knees, ankles and ambiguous patterns were analyzed to predict similarities and differences in these different activity patterns.

Their research revealed that most children had oligoarticular arthritis, and that children with polyarticular arthritis had a more difficult disease progression and took longer to enter remission than those with oligoarticular arthritis.

This is completely consistent with the hospital's observations over the years, and the system can accurately and early distinguish the types of children with JIA. 

But the researchers say better characterization of how the joints are affected is needed to predict the course of the disease and its severity, so that treatment can be more accurately targeted.