Research on the Whole Process of AI Intelligent Management System for the Diagnosis and Treatment of Inflammatory Bowel Diseases

Overview

Inflammatory bowel disease (IBD), including Crohn's disease (CD) and ulcerative colitis (UC), is a chronic immune-mediated disorder requiring long-term management. Clinically, IBD may involve recurrent intestinal inflammation, ulcer formation, and complications such as strictures and fistulas. The etiology of IBD is associated with immune dysregulation, gut microbiome imbalance, and genetic susceptibility. Its clinical manifestations are heterogeneous; early symptoms such as abdominal pain, diarrhea, weight loss, hematochezia, or anemia often resemble gastroenteritis, irritable bowel syndrome, or infectious enterocolitis, leading to misdiagnosis and delayed diagnosis. According to international studies, the interval between initial symptom onset and confirmed diagnosis can range from several months to years, during which untreated disease progression increases the risks of hospitalization, surgery, bowel strictures, and fistulizing complications, resulting in significant impacts on patient quality of life. This study adopts a retrospective design, analyzing our hospital's electronic medical record data from 2023 to 2025.The objective is to evaluate the performance and feasibility of an artificial intelligence (AI) model-developed and incorporating natural language processing (NLP) and phenotypic recognition algorithms-in supporting early identification and diagnosis of IBD. The model has been validated in multiple European healthcare systems and is capable of recognizing high-risk phenotypic clusters from large-scale structured and unstructured medical data. This study represents the first application of this AI technology in the Taiwanese IBD population. All data processing will occur within a de-identified and secure computing environment to ensure data privacy and information security. The study will compare AI-generated diagnostic suggestions derived from medical records with actual clinical diagnoses to assess consistency and accuracy. The model's performance across different clinical characteristics, disease severity levels, and stages of illness will also be examined. In addition, statistical metrics such as precision and recall will be used to generate PRC curves for determining the optimal diagnostic threshold. The outcomes of this study are expected to validate the potential of AI technology in facilitating early recognition, accelerating diagnosis, and supporting clinical decision-making for IBD. The findings will provide essential data for developing localized AI models for IBD, ultimately enhancing diagnostic efficiency, shortening the diagnostic timeline, and improving long-term patient outcomes and quality of life. Objective 1：To retrospectively analyze the clinical characteristics and diagnostic pathways of patients with IBD (CD/UC). Objective 2：To evaluate the performance of the AI model in identifying and providing diagnostic suggestions for high-risk IBD cases. Objective 3：To compare the accuracy and consistency between AI-generated diagnostic suggestions and actual clinical diagnoses.

Primary outcome measures

• developing localized AI models for IBD [Time frame: No direct participant involvement; retrospective chart review of medical records from 2023 to 2025 only.]

Eligibility criteria

Inclusion criteria

Inclusion criteria for the IBD group:

Patients diagnosed with IBD (K50.00 to K51.919) within the specified time interval.

Inclusion criteria for the non-IBD group:

Patients never diagnosed with IBD (K50.00 to K51.919) within the specified time interval.

Exclusion criteria

Patients not within the specified time interval Deceased patients

Exclusion criteria for the non-IBD group:

Patients not within the specified time interval Deceased patients Patients with fewer than 5 hospital visits

Criteria are shown verbatim from the registry (in English). Final eligibility is always assessed by the study center.

Healthy volunteers: No

Study design

Observational model

Case-control

Study locations

Taiwan · 1 center

• Taichung Veterans General Hospital — Taichung

Publications

• Wang HH, Wang YH, Liang CW, Li YC. Assessment of Deep Learning Using Nonimaging Information and Sequential Medical Records to Develop a Prediction Model for Nonmelanoma Skin Cancer. JAMA Dermatol. 2019 Nov 1;155(11):1277-1283. doi: 10.1001/jamadermatol.2019.2335. PMID 31483437
• Nguyen PA, Syed-Abdul S, Iqbal U, Hsu MH, Huang CL, Li HC, Clinciu DL, Jian WS, Li YC. A probabilistic model for reducing medication errors. PLoS One. 2013 Dec 3;8(12):e82401. doi: 10.1371/journal.pone.0082401. eCollection 2013. PMID 24312659
• Le Berre C, Ricciuto A, Peyrin-Biroulet L, Turner D. Evolving Short- and Long-Term Goals of Management of Inflammatory Bowel Diseases: Getting It Right, Making It Last. Gastroenterology. 2022 Apr;162(5):1424-1438. doi: 10.1053/j.gastro.2021.09.076. Epub 2022 Jan 4. PMID 34995529
• Turner D, Ricciuto A, Lewis A, D'Amico F, Dhaliwal J, Griffiths AM, Bettenworth D, Sandborn WJ, Sands BE, Reinisch W, Scholmerich J, Bemelman W, Danese S, Mary JY, Rubin D, Colombel JF, Peyrin-Biroulet L, Dotan I, Abreu MT, Dignass A; International Organization for the Study of IBD. STRIDE-II: An Update on the Selecting Therapeutic Targets in Inflammatory Bowel Disease (STRIDE) Initiative of the I PMID 33359090

Identifiers

NCT: NCT07590271 · CE26157A

Primary sources (government registries)

• ClinicalTrials.gov (US NIH)

View this study on ClinicalTrials.gov ↗