AI system can help ID cancerous lesions with IBD
- Inflammatory bowel disease (IBD) has no cure and affects 6 to 8 million people globally.
- People with IBD have a higher risk of developing colorectal cancer.
- Researchers from Okayama University developed an artificial intelligence (AI) system to help doctors better identify potentially cancerous lesions in the large intestine.
Inflammatory bowel disease (IBD) affects between 6 to 8 million people globally.
People with IBD have a much higher risk of developing colorectal cancer due to chronic inflammation of the large intestine or colon.
Inflammation in the large intestine makes it difficult for doctors to clearly view potentially cancerous lesions through an endoscopy. For this reason, clinicians may turn to biopsies, which are invasive and have risks.
To help provide another option, a team of researchers from Okayama University in Japan has developed an artificial intelligence (AI) system to help doctors more accurately classify these lesions in a less invasive way.
The new study was recently published in the Journal of Gastroenterology and Hepatology.
What is IBD?
IBD refers to two main conditions: ulcerative colitis and Crohn’s disease.
Ulcerative colitis is a disease that causes inflammation in the colon or large intestine. The lining of the colon becomes so inflamed that ulcers form, causing abdominal pain, diarrhea, and other symptoms.
Crohn’s disease causes inflammation in any part of the digestive tract. Symptoms of Crohn’s disease include pain where inflammation is occurring in the digestive tract as well as ulcers, diarrhea, anemia, and weight loss.
Diagnosis of IBD many times includes a colonoscopy or endoscopy.
There is currently no cure for IBD. Current treatments aim to reduce symptoms and prevent complications such as colon cancer, malnutrition, and bowel obstruction.
AI and IBD
For this pilot study, researchers from Okayama University wanted to see how well the AI system they developed measured up to standard endoscopy in classifying lesions in people with IBD. These lesions called IBD neoplasia are areas of abnormal cell growth that may or may not cause cancer.
First, the research team trained the AI system using 862 endoscopic images of 99 IBD lesions from people with IBD. The images were dated between 2003 and 2021 from two different hospitals.
Next, researchers asked endoscopists with more than 8 years of experience in gastrointestinal endoscopy to analyze the images and classify the lesions into two types. One of those types can sometimes cause a doctor to recommend a proctocolectomy — the surgical removal of the majority of the large intestine.
Through the pilot study, scientists found the AI system provided an image-based diagnostic ability of 64% sensitivity, 89% specificity, and 80% accuracy. The AI system also had a lesion-based diagnostic ability of 74% sensitivity, 85% specificity, and 80% accuracy.
Additionally, researchers found the AI system had a 79% accuracy rate when diagnosing the lesion images, while the human endoscopists had a 77% accuracy rate.
“We have successfully prototyped an AI model to determine the degree of malignancy for inflammatory bowel disease-related tumors,” Dr. Hideaki Kinugasa, assistant professor at the Department of Gastroenterology and Hepatology at Okayama University Graduate School of Medicine, Dentistry, and Pharmaceutical Sciences in Okayama, Japan, and lead author of this study, told Medical News Today. “Our AI system is valuable enough to contribute to the next generation of clinical practice.”
The next generation of endoscopic diagnosis
Kinugasa said it is important for doctors to have a tool that allows them to properly identify the severity and grade of IBD neoplasia before deciding on a course of treatment for patients with IBD.
“Overdiagnosis and underdiagnosis can be avoided and more appropriate treatment can be obtained from the beginning,” he explained. “The possibility of local treatment for inflammatory bowel disease-related tumors for which total colorectal resection is the basic treatment might be explored. By adding other annotations to this AI system, we aim to establish the next generation of endoscopic diagnosis and care.”
MNT spoke with Dr. Ashkan Farhadi, a gastroenterologist at MemorialCare Orange Coast Medical Center in Fountain Valley, California, who stated this study shows progress and another tool that can add to what doctors can offer their patients.
“The importance of finding abnormal lesions in a setting of colitis is very important because the lining is already not normal,” Farhadi explained. “So when the lining is not normal, detection of something that is abnormal on top of it is tricky and more difficult.”
Farhadi added that he would like to see how this AI system performs in a real-world setting.
“There’s no data out yet that shows how this performs in the real world (and) in the hands of real doctors — not just in a setting of research,” he said. “How in the hands of real doctors and in a setting with real patients this can perform yet needs to be seen.”
Dr. Nicholas A. Hoerter, an assistant professor of medicine at the Henry D. Janowitz Division of Gastroenterology at Icahn School of Medicine at Mount Sinai in New York, stressed as a gastroenterologist, the topic of early detection of pre-cancerous lesions in the gastrointestinal tract is of utmost importance.
“Colon cancer is the number two cause of cancer death in men and women in the United States, and patients with inflammatory bowel disease have an even higher risk than the general population,” Hoerter told MNT. “Colon cancer can potentially be prevented by early detection and removal of pre-cancerous lesions, but accurate detection in IBD is difficult and requires special expertise and training. Any technology that can augment our ability to detect these pre-cancerous lesions has the potential to save lives.”
Source: Read Full Article