Introduction
Digital single-operator cholangioscopy (DSOC) allows an accurate diagnosis of biliary strictures. Two DSOC-based artificial intelligence (AI) models trained to identify neoplasia were proposed. DSOC-guided pCLE biliary strictures direct visualization, using the Miami and Paris classifications has demonstrated high diagnostic accuracy for neoplasia (higher than expert endoscopists at times). To date, there are no head-to-head studies comparing DSOC-AI vs DSOC-pCLE.
Aims & Methods
We aim to compare the diagnostic accuracy for identifying neoplasia in indeterminate biliary lesions using a novel DSOC-AI model against DSOC-pCLE. Historic cohort-based diagnostic accuracy study in agreement with STARD initiative including patients ≥18 years old who underwent DSOC-pCLE (06/2014 to 11/2021). Patients without available DSOC videos, biopsy, or impossibility for 12-month follow-up were excluded. For DSOC direct visualization, neovasculature constituted neoplasia. For DSOC-pCLE, a higher presence of the Miami classification’s malignancy criteria (thick white and dark bands, dark clumps, epithelium) than the Paris classification’s inflammatory criteria (vascular congestion, roughness aspect, increased inter-glandular space, and thickened reticular structure) constituted neoplasia. For DSOC and pCLE-guided biopsy, corresponding histological findings constituted neoplasia. An offline DSOC-based AI model analysis was performed using the DSOC recorded videos from the selected cases. The DSOC-AI model considered neovasculature as the criterion for neoplasia. The gold standard for neoplasia was based on further clinical evolution, imaging, or surgical specimen findings during a 12-month follow-up. Accuracy of DSOC direct visualization, DSOC-pCLE, DSOC and pCLE-guided biopsy, and DSOC-AI model was calculated in terms of sensitivity, specificity, positive and negative predictive values (PPV and NPV), observed agreement, and area under the curve (AUC). Comparison of obtained diagnostic accuracies were defined through DeLong’s test for two ROC curves.
Results
Ninety patients were selected, median age was 66.4 ± 13.7 years, 56.7% female. Tumour suspicion was the most common indication (55.6%) (Table 1). DSOC-AI reached a 97.7% sensitivity, 75% specificity, 98.8% PPV, 60% NPV and 96.7% observed agreement; whereas, DSOC-pCLE had a 94.2% sensitivity, a 100% specificity, 100% PPV, 44.4% NPV, and 94.4% observed agreement (Table 2). AUC for DSOC-AI was 0.79; for DSOC direct visualization (0.74; P=.763), DSOC-pCLE (0.72; P=.634), and DSOC and pCLE-guided biopsy (0.83; P=.809).
| Sensitivity
| Specificity
| PPV
| NPV
| OA
|
DSOC direct visualization
| 83/86; 96.51
(90.14 - 99.27)
| 3/4; 75
(19.41 - 99.37)
| 83/84; 98.81
(93.54 – 99.97)
| 3/ 6; 50
(11.81 – 88.19)
| 86/90; 95.56
(89.01 – 98.78)
|
DSOC-guided pCLE
| 81/86; 94.19
(86.95 – 98.09)
| 4/4; 100
(39.76 – 100) | 81/81; 100
(95.55 – 100)
| 4/ 9; 44.44
(13.7 - 78.8)
| 85/90; 94.44
(87.51 - 98.17)
|
DSOC and pCLE-guided biopsy
| 84/86; 97.67
(91.85 - 99.72)
| 4/4; 100
(39.76 - 100)
| 84/84; 100
(95.7 – 100)
| 4/ 6; 66.67
(22.28 – 95.67)
| 88/90; 97.78
(92.2 – 99.73)
|
DSOC-based AI model
| 84/86; 97.67
(91.85 – 99.72)
| 3/4; 75
(19.41 – 99.37)
| 84/85; 98.82
(93.62 - 99.97)
| 3/ 5; 60
(14.66 - 94.73)
| 87/90; 96.67
(90.57 - 99.31)
|
PPV: Positive predictive value; NPV: negative predictive value; OA: observed agreement; pCLE: probed-based confocal laser endomicroscopy.
|
Conclusion
The DSOC-AI model demonstrated an offline diagnostic accuracy similar to DSOC-pCLE. Larger multicentric head-to-head trials with a proportional sample among neoplastic and non-neoplastic cases are advisable for best estimation of obtained results.
References
1. Koda H, Hara K, Nozomi O, Kuwahara T, Nobumasa M, Haba S, et al. High-Resolution Probe-Based Confocal Laser Endomicroscopy for Diagnosing Biliary Diseases. Clin Endosc. 2021;54(6):924–9.
2. Pilonis ND, Januszewicz W, di Pietro M. Confocal laser endomicroscopy in gastro-intestinal endoscopy: Technical aspects and clinical applications. Transl Gastroenterol Hepatol. 2022;7:1–20.
3. Kahaleh M, Raijman I, Gaidhane M, Tyberg A, Sethi A, Slivka A, et al. Digital Cholangioscopic Interpretation: When North Meets the South. Digestive Diseases and Sciences. 2021;67(4):1345–51.
4. Robles-Medranda C, Oleas R, Sánchez-Carriel M, Olmos JI, Alcívar-Vásquez J, Puga-Tejada M, et al. Vascularity can distinguish neoplastic from non-neoplastic bile duct lesions during digital single-operator cholangioscopy. Gastrointestinal Endoscopy. 2021;93(4):935–41.
5. Robles-Medranda C, Valero M, Soria-Alcivar M, Puga-Tejada M, Oleas R, Ospina-Arboleda J, et al. Reliability and accuracy of a novel classification system using peroral cholangioscopy for the diagnosis of bile duct lesions. Endoscopy. 2018;50(11):1059–70.
6. Sethi A, Tyberg A, Slivka A, Adler DG, Desai AP, Sejpal D v., et al. Digital Single-operator Cholangioscopy (DSOC) Improves Interobserver Agreement (IOA) and Accuracy for Evaluation of Indeterminate Biliary Strictures: The Monaco Classification. Journal of Clinical Gastroenterology. 2022;56(2).
Disclosure
Carlos Robles-Medranda is a key opinion leader and consultant for Pentax Medical, Steris, Medtronic, Motus, Micro-tech, G-Tech Medical Supply, CREO Medical, EndoSound, and mdconsgroup. The other authors declare no conflicts of interest.