Artificial intelligence is revolutionizing our understanding of complex diseases, offering new methods of health research that wouldn’t have been possible without it. A recent paper co-authored by Amii researchers uncovers the hidden links between bowel disease and mental health, and may lend support to a long-suspected connection between the gut and the brain.
The paper, Network analysis of extraintestinal manifestations and associated autoimmune disorders in Crohn’s disease and ulcerative colitis, was co-authored by Amii Fellows J. Ross Mitchell and Randy Goebel. It appeared in the prestigious njp Digital Medicine journal, published by Nature, earlier this year.
The project began when Mitchell was approached by Daniel Baumgart, a fellow professor at the University of Alberta and an internationally-renowned expert on inflammatory bowel diseases (IBD). Baumgart has access to a massive dataset from over 30,000 IBD patients suffering from Crohn's disease, collected over two decades through Alberta Health Services.
“He put years of work into this to collect this unbelievably huge data set that is unparalleled in the world as far as we know,” Mitchell says.
Crohn's disease is a chronic condition that can cause swelling and inflammation in the bowel and can lead to pain, diarrhea, and malnutrition. If untreated, it can have a severe impact on a person’s quality of life and can lead to an increased risk of colon cancer.
Canada, particularly Alberta, has high rates of Crohn's disease and other IBDs compared to other Western countries, Mitchell says, making the dataset Baumgart collected a valuable research tool.
Making sense of a mountain of data
Mitchellexplains that the data was anonymized, removing the patient’s personal info. But it did include the patients’ diagnostic codes. These codes are what the medical system uses to identify the diseases or conditions for which a person is diagnosed and treated: each diagnosis gets a unique code, which is recorded in the patient’s medical history.
Mitchell and Baumgart were sure there were valuable insights within the massive database, but its size would make it difficult to find them.
“I couldn't make sense of it. I was going through this — many tens of thousands of rows with many thousands of patients — it's unintelligible,” he said.
“It's too much, and so I thought, how would I tackle this as a data scientist. The natural approach would be to build and display it as a network.”
Mitchell built a tool that used a machine learning approach called network analysis to find hidden connections within the data. Each diagnostic code was turned into a node, and the tool then visualized the connections between nodes with lines. The thicker the line, the stronger the connection.
Suddenly, it was possible to see at a glance which other conditions were strongly connected with IBD. Some of those connections were expected - people with IBD often suffer from other gastrointestinal conditions, as well as musculoskeletal issues. But Mitchell says he was surprised to find the strongest connection was actually with mental health disorders. Mitchell says that of the 30,000 patients with Crohn's disease, nearly a third were also diagnosed with anxiety or depression sometime in their lifetime.
Findings spark international interest
The study is believed to be the most comprehensive and first AI-supported analysis of the ways that IBDs can manifest outside of the intestines. And it has caught the attention of others who study gastrointestinal disease.
In the medical community, there has long been suspicion that our intestinal systems and brain have a stronger connection than is immediately apparent, called the gut-brain axis. The connection seems to be a two-way street: the health of the digestive system can have a big impact on our moods and mental health, and vice versa.
Mitchell says this study supports that idea and has led to a lot of interest among other researchers examining the gut-brain axis.
“Baumgart has had a lot of communication from people around the world who see this and go, 'Oh, this is really cool because it supports some of our hypotheses that we've been talking about for years.’”
These insights could also lead to changes to the way that IBDs are diagnosed and treated. Mitchell notes that being aware of these connections could mean that physicians can keep an eye out for signs of anxiety and depression in IBD patients that might have otherwise been missed.
It could also have an impact on treatment decisions; for instance, if a prescription used to treat bowel disease might increase the risk of depression, a doctor needs to understand that connection.
Mitchell says the network analysis was only the first study done with the dataset, and there are many more insights to be gathered from it. He is planning other machine-learning projects using the data, with his ultimate goal being algorithms that could lead to faster and easier diagnosis of Crohn’s and other bowel disorders.
“There's a lot to do. What we learn from this data can help physicians all over the world conduct medicine and future experiments.