Around 500,00 people in the US have Chron's disease, a chronic inflammatory bowel disease that damages the digestive lining that results in inflammation, abdominal pain, severe diarrhea, exhaustion, weight loss, and malnutrition. Scientists are using artificial intelligence (AI) to predict the postoperative recurrence of the disease successfully.
They built a deep learning model trained to analyze histopathology images of surgical specimens in patients with and without Chron's disease recurrence.
Training AI to Predict Recurrence of Postoperative Chron's Disease
Researchers used AI to identify and categorize histopathology images that will be used in a model that predicts postoperative recurrence of Chron's disease. According to SciTech Daily, the AI has previously identified unknown differences in adipose cells and disparities in the degree of mast cell infiltration in the outer lining of the gut when comparing people with and without the recurrence.
Chron's disease has a 10-year rate of postoperative symptomatic recurrence, which could happen to about 40% of people who once had the disease. Currently, no scoring system is available to predict whether the inflammatory gastrointestinal illness will return.
More so, lead investigators Dr. Takahiro Matsui and Dr. Eiichi Morii from Osaka University Graduate School of Medicine's Department of Pathology explained that most analysis today uses AI focused on malignant tumors. However, they aimed to obtain useful information from a wider variety of hispothalogy images.
About 68 individuals with Chron's disease participated in the study who underwent bowel resection between January 2007 and July 2018. They were divided into two groups based on whether they will have postoperative recurrence after two years.
The two groups are further divided into subgroups whose histopathological images were used for training the AI and the other group used to validate the AI model. The photos were labeled for the presence or absence of postoperative recurrence and processed. The findings indicate that AI accurately classified and labeled the photos according to the presenting symptoms of the disease.
Prediction Heat Maps Accurately Identified Location of Recurrence
After training the AI, the next step for the team is to create prediction heat maps that will identify locations and histological markers that the machine learning algorithm predicted, according to a similar report from Azo Robotics. The team photographed all layers of the intestinal wall to achieve this second step.
The heatmaps showed that AI accurately predicted the subserosal adipose tissue layer, although the model was less precise in mucosal and correct muscle layers. They noticed that all the photos with the greatest predictive findings had fat tissue.
However, the recurrence group's adipose cells turned out to be smaller, flatter and had a lower center cell distance than those nonrecurrence group. It is called adipocyte shrinkage, an important histological characteristic when predicting postoperative recurrence of Chron's disease, according to the researchers.
Furthermore, those with a larger number of mast cells in the subserosal adipose tissue have cells linked to Chron's disease recurrence and adipocyte shrinkage.
The researchers noted that the data they gathered is the first to correlate the histology of subserosal adipose cells and mast cell infiltration to postoperative recurrence of Chron's disease.
They detailed the results of their study in "Deep Learning Analysis of Histologic Images from Intestinal Specimen Reveals Adipocyte Shrinkage and Mast Cell Infiltration to Predict Postoperative Crohn Disease," published in Elsevier's The American Journal of Pathology.
RELATED ARTICLE: Inflammatory Bowel Disease - Nutritional Suggestions!
Check out more news and information on Medicine & Health in Science Times.
