Machine learning in the differential diagnosis of ulcerative colitis and Crohn's disease: a systematic review.

Inflammatory bowel disease (IBD) includes two main types: ulcerative colitis and Crohn's disease. While both cause chronic inflammation in the digestive system, they affect different areas and require different treatments. Doctors sometimes struggle to tell them apart, which can delay proper care. This review examined 31 studies involving over 15,000 patients to see if computer-based machine learning could help doctors make more accurate diagnoses. The researchers found that machine learning programs, particularly those using "random forest" and "deep learning" methods, showed promise in distinguishing between these two conditions. These computer models were especially accurate when analyzing data from colonoscopy images and stool samples. Most studies were published recently (2021-2023), suggesting this is an emerging field. While this research doesn't directly involve acupuncture, it's important for patients with IBD to understand their exact diagnosis, as treatment approaches differ significantly between ulcerative colitis and Crohn's disease. Some patients with IBD explore complementary therapies like acupuncture to help manage symptoms such as abdominal pain, stress, and inflammation. Traditional Chinese medicine views digestive disorders as imbalances in the body's energy systems, and acupuncture may help regulate immune function and reduce inflammation. If you have IBD and are considering acupuncture as part of your care plan, discuss it with your gastroenterologist first, as it should complement rather than replace conventional medical treatment. To explore acupuncture safely, seek a licensed acupuncturist with experience treating digestive conditions.

This systematic review analyzed 31 studies (n=15,140) examining machine learning applications for differentiating ulcerative colitis from Crohn's disease. The review searched six databases (PubMed, Web of Science, Embase, Cochrane Library, Scopus, Ovid) for studies published between January 2000 and November 2024, with protocol registered in PROSPERO (CRD42024543036). Most included studies were retrospective (87%), with 65% published between 2021-2023. Random forest was the most frequently utilized algorithm (32%), followed by support vector machines (29%). Quality assessment employed QUADAS criteria. Key findings indicate machine learning, particularly deep learning and random forest models utilizing endoscopic imagery and fecal biomarker data, demonstrated potential for enhanced diagnostic accuracy in IBD subtype differentiation. Clinical relevance for acupuncture practitioners: Accurate IBD subtype diagnosis is essential before developing complementary treatment protocols, as disease patterns and acupoint selection may differ between ulcerative colitis and Crohn's disease presentations. No specific effect sizes were reported in this methodological review.