IDDF2024-ABS-0261 Mass data-driven gastrointestinal tract screening using machine learning and VDO capsule endoscopy (TeleGI): a study protocol

<h3>Background</h3> Efficient screening for gastrointestinal (GI) pathologies, including colorectal cancer (CRC), is crucial for early detection. Traditional methods are often invasive and uncomfortable. The TeleGI Study, in collaboration with the Department of Health Service Support and over 1 million Village Health Volunteers (VHVs), employs machine learning (ML) and non-invasive video capsule endoscopy to improve screening efficiency and patient compliance. This study evaluates the feasibility and effectiveness of this innovative CRC screening protocol through a randomized controlled trial in Nonthaburi province, Thailand. <h3>Methods</h3> Phase I Collects and analyzes gastroscopy and colonoscopy data from Phra Nang Klao Hospital to develop machine learning (ML) insights and assess CRC screening practices. Phase II Develops ML models using Phase I data to identify high-risk individuals by analyzing demographics, medical history, and GI symptoms. The models detect various pathologies such as cancers, polyps, diverticula, and ulcers across multiple GI tract sections, validated using methods like split-validation and cross-validation for accuracy. Phase III Conducts a clustered randomized controlled trial (RCT) with 240 high-risk individuals, randomized into three groups across six districts: 80 in conventional stool tests, 80 in capsule endoscopy, and 80 in combined methods, with positivity rates expected to yield approximately 6, 32, and 40 cases, respectively. Positive cases undergo diagnostic colonoscopies at Phra Nang Klao Hospital to confirm GI pathologies, anticipating 78 positives and confirming 61 based on historical data. The analysis focuses on the efficacy of detecting multiple pathologies across different GI locations, comparing detection rates, accuracy, patient compliance, and cost-effectiveness. The three phases of the study design are presented in Figure 1 (IDDF2024-ABS-0261 Figure 1. Study design) <h3>Results</h3> The use of ML is expected to enhance the precision of risk stratification, promoting early and targeted interventions. Capsule endoscopy is anticipated to significantly improve patient compliance due to its non-invasive nature, leading to higher detection rates of early-stage pathologies. <h3>Conclusions</h3> This innovative approach aims to transform CRC screening practices in Thailand by leveraging the extensive network of VHVs and the latest technological advancements in ML and endoscopy. By enhancing early detection and optimizing resource allocation, the TeleGI Study hopes to set new standards in GI pathology screening, ultimately improving patient outcomes and healthcare efficiency.