Background and study aims: Probe-based confocal laser endomicroscopy (pCLE) is a new imaging modality that enables
histological examination of gastrointestinal mucosa during endoscopic procedures.
Most studies have evaluated offline interpretation of pCLE images. In clinical practice,
real-time interpretation is necessary to assist decision-making during the procedure.
The aim of this pilot study was to compare the accuracy of real-time pCLE diagnosis
made during the procedure with that of blinded offline interpretation to provide accuracy
estimates that will aid the planning of future studies.
Patients and methods: pCLE was performed in patients undergoing screening and surveillance colonoscopy.
Once a polyp had been identified, one endoscopist analyzed pCLE images during the
procedure and made a provisional “real-time” diagnosis. Saved video recordings were
de-identified, randomized, and reviewed “offline” 1 month later by the same endoscopist,
who was blinded to the original diagnoses.
Results: Images from a total of 154 polyps were recorded (80 neoplastic, 74 non-neoplastic).
The overall accuracy of real-time pCLE diagnosis (accuracy 79 %, sensitivity 81 %,
specificity 76 %) and offline pCLE diagnosis (83 %, 88 %, and 77 %, respectively)
for all 154 polyps were similar. Among polyps < 10 mm in size, the accuracy of real-time
interpretation was significantly lower (accuracy 78 %, sensitivity 71 %, specificity
83 %) than that of offline pCLE interpretation (81 %, 86 %, 78 %, respectively). For
polyps ≥ 10 mm, the accuracy of pCLE diagnosis in real-time was better (accuracy 85 %,
sensitivity 90 %, specificity 75 %) than offline pCLE diagnosis (81 %, 97 %, and 50 %,
respectively).
Conclusions: These results suggest that real-time and offline interpretations of pCLE images are
moderately accurate. Real-time interpretation is slightly less accurate than offline
diagnosis, but overall both are comparable. Additionally, there was contrasting accuracy
between the two methods for small and large polyps.
