Clinical Trial Details
— Status: Completed
Administrative data
NCT number |
NCT05858827 |
Other study ID # |
EA-23-004 |
Secondary ID |
|
Status |
Completed |
Phase |
|
First received |
|
Last updated |
|
Start date |
May 10, 2023 |
Est. completion date |
December 20, 2023 |
Study information
Verified date |
January 2024 |
Source |
Renmin Hospital of Wuhan University |
Contact |
n/a |
Is FDA regulated |
No |
Health authority |
|
Study type |
Observational
|
Clinical Trial Summary
In this study, the EUS intelligent picture reporting system can automatically generate
reports after reading videos of EUS examinations. This function can standardize the quality
of endoscopic ultrasound image reporting and reduce the work burden of ultrasound
endoscopists.
Description:
A well-written report is the most important way of communication between clinicians,
referring doctors and patients. Reports play a key role for quality improvement in digestive
endoscopy, too. Unlike digestive endoscopy, the quality of reporting in endoscopic ultrasound
(EUS) has not been thoroughly evaluated and a reference standard is lacking. According to the
guidance statements regarding standard EUS reporting elements developed and reviewed at the
Forum for Canadian Endoscopic Ultrasound 2019 Annual Meeting, appropriate photo documentation
of all relevant lesions and anatomical landmarks should be included in EUS reports and stored
for future reference. Systematic photo documentation in EUS is an indicator of procedure
quality according to the ASGE. Systematic photo documentation can facilitate surveillance EUS
evaluations. According to an international online survey, most endosonographers used a
structured tree in the report describing either normal and abnormal findings (81%) or only
abnormal findings (7%). Therefore, it is necessary to develop a standardized endoscopic
ultrasound image report system.
The past decades have witnessed the remarkable progress of artificial intelligence (AI) in
the medical field. Deep learning, a subset of AI, has shown great potential in elaborating
image analysis. In the field of digestive endoscopy, deep learning has been widely studied,
including identifying focal lesions, differentiating malignant and non-malignant lesions, and
so on. However, rare study works on automatic photo documentation during endoscopic
ultrasound.
Our previous work has successfully developed a deep learning EUS navigation system that can
identify the standard stations of the pancreas and CBD in real time. In the present study, we
further constructed an EUS automatic image reporting system (EUS-AIRS). The EUS-AIRS can
automatically capture images of standard stations, lesions, and biopsy procedures, and label
Types of lesions, thereby generating an image report with high completeness and quality
during endoscopic ultrasonography.
We tested the performance of the EUS-AIRS by testing its performance on retrospective
internal and external data, and we anticipate determining the utility of the EUS-AIRS in
clinical practice by testing its performance in consecutive prospective patients.