A groundbreaking study conducted at Sheba Medical Center in Tel Hashomer has revealed that machine-learning analysis of complete capsule endoscopy (CE) videos can accurately predict the need for biological therapy in the treatment of Crohn’s disease (CD), a chronic gastroenterological disorder.
The research team, led by Prof. Uri Kopylov, Prof. Shomron Ben-Horin, and Intel engineering director Amit Bleiweiss, developed an artificial intelligence (AI) algorithm that achieved an impressive 81% accuracy when analyzing CE videos during the initial diagnosis of CD. This accuracy rate surpassed that of gastroenterologists who relied on the traditional method of analyzing inflammatory markers in stool samples (calprotectin).
Crohn’s disease is a chronic condition characterized by inflammation in various parts of the digestive system. It can affect individuals of all ages, with symptoms typically emerging during childhood or early adulthood. Common symptoms include diarrhea, fatigue, stomach aches, cramps, weight loss, and blood in the stool.
The team at Sheba Medical Center utilized a newly developed deep learning model to analyze complete CE videos from 101 CD patients, achieving the remarkable accuracy rate of 81%.
Prof. Kopylov emphasized the potential impact of AI in predicting disease progression and patient outcomes, stating, “By adopting AI in clinical practice, we can begin to use our wealth of knowledge and research in personalized medicine to drive improved patient outcomes and open the door to new possibilities for diagnosis and treatment.”
Capsule endoscopy allows for a comprehensive analysis of the entire digestive system using a tiny device equipped with a camera and transmitter. However, each capsule video generates an overwhelming amount of visual data, typically comprising 10,000 to 12,000 images, making it challenging for physicians to discern all the necessary details. This is where AI algorithms prove invaluable, as they can efficiently process and analyze the vast amount of visual information.
This study builds upon previous research conducted last year, demonstrating the AI algorithm’s ability to analyze up to 12,000 images in just two minutes. Additionally, the AI analysis exhibited high diagnostic accuracy, achieving 86% accuracy in image and data analysis compared to 68% accuracy when relying solely on experienced gastroenterologists. The AI analysis also outperformed the analysis of inflammatory markers in stool samples by experienced physicians.
Dr. Eyal Klang, head of the Sami Sagol AI Hub at Sheba’s ARC Innovation Center, emphasized the transformative potential of AI in healthcare systems, stating, “Our findings are further proof of the powerful impact that AI can have in transforming our health systems and driving positive patient outcomes.” He expressed optimism about future validations of this technology and its implementation in hospitals and clinics worldwide.
The use of AI in the diagnosis and treatment of Crohn’s disease holds great promise, paving the way for more accurate assessments and personalized approaches to patient care.
