jRCT ロゴ

臨床研究等提出・公開システム

Top

Japanese

Jan. 07, 2026

Jan. 07, 2026

jRCT1032250626

Development and implementation of an AI-driven intraoperative real-time biliary structure recognition system using ICG fluorescence imaging during laparoscopic cholecystectomy

Creating an AI-based system to help doctors see bile ducts more clearly during surgery

Saiura Akio

Juntendo University Hospital

3-1-3 Hongo Bunkyoku Tokyo

+81-3-3813-3111

a-saiura@juntendo.ac.jp

Kawano Fumihiro

Juntendo University Hospital

3-1-3 Hongo Bunkyoku Tokyo

+81-3-3813-3111

fkawano@juntendo.ac.jp

Recruiting

Jan. 07, 2026

300

Observational

single arm study

open(masking not used)

uncontrolled control

single assignment

diagnostic purpose

Patients undergoing laparoscopic cholecystectomy with ICG fluorescence imaging who meet all of the following criteria:
1.Aged 18 years or older at the time of consent.
2.Diagnosed with gallstones, cholecystitis, or gallbladder polyps.
3.Able to provide written informed consent based on sufficient explanation of the study, with full understanding and voluntary agreement by the participant.

1.Patients who are unable or unwilling to provide informed consent.
2.Patients for whom indocyanine green cannot be administered at least two hours before surgery (e.g., emergency procedures).
3.Patients undergoing open surgery.
4.Pregnant
5.Minors or patients who have difficulty understanding the study in Japanese.
6.Any individual deemed unsuitable for participation by the principal investigator.

18age old over
No limit

Both

gallstones, cholecystitis, or gallbladder polyps.

gallstones, cholecystitis, or gallbladder polyps.

The performance of the AI system will be evaluated using the macro-averaged F1 score.

Methods:

The F1 score will be calculated for each class, and their unweighted mean will be used as the macro-averaged F1 score.

After the final model is fixed, evaluation will be performed once using the threshold determined from the development set.

The 95% confidence interval will be estimated using patient-level bootstrapping (1,000 iterations).

Precision
Recall
ROC-AUC
PR-AUC
Balanced accuracy
MCC
Confusion matrix
ICG visibility score
Inter-rater agreement

None
Juntendo University Certified Review Board
2-1-1Hongo, Bunkyo-ku, Tokyo, Tokyo

+81-3-5802-1584

crbjun@juntendo.ac.jp
Approval

Nov. 07, 2025

No

the United States