Prospective Intestinal Ultrasound Study for Artificial Intelligence Algorithm Development

Inflammatory Bowel Disease (IBD)

Intestinal Ultrasound, Artificial Intelligence, Machine Learning, Bowel Wall Thickness, Crohn Disease, Ulcerative Colitis, Inflammatory Bowel Disease, Ultrasound

This observational study will use intestinal ultrasound images to help develop computer programs that may make it easier to assess bowel inflammation. The study will include adults who are healthy or who are having an intestinal ultrasound as part of care for inflammatory bowel disease, such as Crohn disease or ulcerative colitis. Researchers want to find out whether a computer program can measure bowel wall thickness from ultrasound images as accurately as experienced ultrasound clinicians. They will also explore whether the program can identify other changes in the bowel. Participants will have an intestinal ultrasound as part of their usual care. People who agree to take part will allow the research team to use de-identified ultrasound images and a small amount of related health information. No medication, experimental treatment, or additional ultrasound procedure will be given for this study. The computer program will not be used to make decisions about a participant's care, and participants will not receive individual results from the program.

Inflammatory bowel disease, including Crohn disease and ulcerative colitis, requires reliable assessment of intestinal inflammation to support clinical monitoring and treatment decisions. Intestinal ultrasound is a non-invasive, point-of-care imaging method that can assess bowel inflammation without radiation exposure. However, its interpretation may vary by operator experience and local scanning practice. This prospective, multicentre, observational study will collect de-identified intestinal ultrasound images, cine loops, and limited associated metadata from adults aged 18 years and older. Eligible participants may be healthy individuals with no known inflammatory bowel disease or individuals undergoing intestinal ultrasound for inflammatory bowel disease screening, monitoring, or another clinical reason. Intestinal ultrasound examinations will be performed as part of routine clinical care. The study does not add a treatment intervention, investigational drug, experimental device, or clinically directed diagnostic procedure. Following informed consent, ultrasound images and cine loops generated during the routine examination will be exported in de-identified DICOM format. Limited information relevant to artificial intelligence development and validation, such as age, sex, self-reported race, pregnancy status, eligibility confirmation, and relevant scan-related or pre-existing clinical information, may also be recorded. Direct personal identifiers and personal health information will be removed before the data are transferred outside the clinical site. The de-identified data will be reviewed for appropriate de-identification and formatting before secure transfer to Dova Health Intelligence Inc. The data will be annotated and used to develop, train, tune, and test computer vision algorithms designed to identify bowel structures and measure bowel wall thickness on intestinal ultrasound. The study will also assess the feasibility of developing algorithms to evaluate bowel wall layer stratification, luminal diameter, bowel wall flow, bowel wall scarring, gastrointestinal motility, and mesenteric fat proliferation. The primary objective is to evaluate whether the computer vision algorithm prototype can measure bowel wall thickness with performance comparable to manual measurements by experienced sonographers. Secondary objectives include evaluating the quality, diversity, and usability of the collected ultrasound data for artificial intelligence development and assessing the feasibility of measuring additional intestinal ultrasound features. The algorithms developed in this study are investigational and will not be used in routine clinical care during the study. No participant-level algorithm findings will be returned to participants, and participation is not expected to provide a direct clinical benefit. The primary foreseeable risk is a loss of confidentiality; this risk is mitigated through de-identification at the study site, verification of de-identification before transfer, secure file transfer, controlled access, and secure cloud-based storage. The study plans to enroll up to 95 participants internationally, including up to 30 participants at the British Columbia site.

No related papers found yet.

Canada