Abstract
Background and Aim: Endoscopic ultrasound (EUS) is the most accurate diagnostic modality for polypoid lesions of the gallbladder (GB), but is limited by subjective interpretation. Deep learning-based artificial intelligence (AI) algorithms are under development. We evaluated the diagnostic performance of AI in differentiating polypoid lesions using EUS images. Methods: The diagnostic performance of the EUS-AI system with ResNet50 architecture was evaluated via three processes: training, internal validation, and testing using an AI development cohort of 1039 EUS images (836 GB polyps and 203 gallstones). The diagnostic performance was verified using an external validation cohort of 83 patients and compared with the performance of EUS endoscopists. Results: In the AI development cohort, we developed an EUS-AI algorithm and evaluated the diagnostic performance of the EUS-AI including sensitivity, specificity, positive predictive value, negative predictive value, and accuracy. For the differential diagnosis of neoplastic and non-neoplastic GB polyps, these values for EUS-AI were 57.9%, 96.5%, 77.8%, 91.6%, and 89.8%, respectively. In the external validation cohort, we compared diagnostic performances between EUS-AI and endoscopists. For the differential diagnosis of neoplastic and non-neoplastic GB polyps, the sensitivity and specificity were 33.3% and 96.1% for EUS-AI; they were 74.2% and 44.9%, respectively, for the endoscopists. Besides, the accuracy of the EUS-AI was between the accuracies of mid-level (66.7%) and expert EUS endoscopists (77.5%). Conclusions: This newly developed EUS-AI system showed favorable performance for the diagnosis of neoplastic GB polyps, with a performance comparable to that of EUS endoscopists.
| Original language | English |
|---|---|
| Pages (from-to) | 3548-3555 |
| Number of pages | 8 |
| Journal | Journal of Gastroenterology and Hepatology (Australia) |
| Volume | 36 |
| Issue number | 12 |
| DOIs | |
| Publication status | Published - 2021 Dec |
Bibliographical note
Funding Information:Preoperative EUS images of pathologically confirmed GB polyps and gallstones were retrospectively collected between January 2014 and May 2019 as the AI development cohort. All EUS examinations were performed with a real-time, gray-scale sector scan echoendoscope EUS (GF-UCT260 or GF-UE260-AL5; Olympus, EG3870UTK; Pentax, Tokyo, Japan) by well-trained endoscopists at the EUS unit of Gachon University Gil Medical Center and Yonsei University College of Medicine, Gangnam Severance Hospital. The diagnostic accuracy of EUS-AI may vary depending on the quality of the training EUS images. To account for this, an expert endoscopist viewed all EUS images obtained by endoscopists with at least an intermediate level of experience and selected the optimal training images. The inclusion criterion was the presence of GB polyps with maximum diameter of 7–20 mm. Multiple images from a single patient were selected from different lesions. The AI development cohort consisted of 1039 EUS images of GB lesions (836 GB polyps and 203 gallstones) from 670 patients who underwent cholecystectomy. An external validation cohort was created with 83 EUS images from 83 patients who underwent preoperative EUS and cholecystectomy between June 2019 and January 2020, which allowed additional verification of the EUS-AI algorithm.
Funding Information:
This work was supported by faculty research grant from the Gachon University Gil Medical Center (Kim Eui Joo, Grant number: FRD2019‐04) and faculty research grant from Yonsei University College of Medicine (Cho Jae Hee, 6‐2020‐0133). Financial support:
Publisher Copyright:
© 2021 Journal of Gastroenterology and Hepatology Foundation and John Wiley & Sons Australia, Ltd
All Science Journal Classification (ASJC) codes
- Hepatology
- Gastroenterology