How SickKids researchers are using AI to reimagine the future of paediatric care

Streamlining care through Emergency Department automation

The emergency department (ED) experience in any hospital can be difficult. To improve the patient experience in the ED, a team led by Staff Physician and Clinical Artificial Intelligence and Machine Learning Lead in the Division of Paediatric Emergency Medicine at SickKids, Dr. Devin Singh is pioneering machine learning algorithms that can identify which diagnostic imaging and lab tests a patient needs before they can be assessed by a physician.

Using ED triage data collected when patients first arrive, the model can predict what tests are needed (like x-rays and urine tests) for common medical conditions. These models may also be able to automate ordering these tests early in a patient’s visit, streamlining care significantly. The team wants to take it a step further to adapt the models to make predictions using data originating from the home, before patients even come to the ED, which could revolutionize the way children access health care. 

Predicting a life-threatening event before it happens

Up to 29 per cent of critically ill children experience abnormal heart rhythms, also known as arrhythmias. Currently, arrhythmias are diagnosed by a physician looking at the ECG signal on the patient's bedside monitor. A SickKids machine learning system developed by a team led by Staff Physician in the Department of Critical Care and Co-Chair of the AI in Medicine for Kids (AIM) Initiative at SickKids Dr. Mjaye Mazwi is able to detect arrhythmias among patients in the critical care unit seconds after onset, alerting the care team and expediting diagnosis. For critically ill children, early diagnosis could mean early treatment to reduce harm and improve patient outcomes.  

Using deep learning to prioritize care and reduce stress on families

Hydronephrosis is a common prenatal ultrasound finding that involves an enlargement of the central part of the kidney where urine collects. While the majority of children with this condition will get better on their own, some will require medical or surgical intervention. These children are monitored with repeated ultrasounds that may include invasive, painful tests that take place over many clinic visits.

A team led by PhD student Lauren Erdman and Nurse Practitioner Mandy Rickard developed a deep learning tool to enable a more efficient way to help children with hydronephrosis. The tool analyzes patient ultrasounds to predict the probability a patient may need surgery, or if they will improve on their own. This aims to help decrease clinic visits and invasive testing for some, while fast-tracking surgery for those who need it – reducing stress and time spent in hospital for families. 

Detecting cancer earlier with whole-body MRIs

Detecting cancer early is critical to improving survival odds. That's why a team led by SickKids Senior Scientist and Co-Chair of the AI in Medicine for Kids Initiative Dr. Anna Goldenberg is developing a whole-body MRI-based machine learning tool to detect early-onset cancers among children. Using this machine learning tool, doctors would more easily be able to detect cancers in children early enough to spare patients the side effects of aggressive treatments.

This content originally appeared in Innovate Toronto. Learn more about how we're using AI to power the future of health-care on the AIM website.