Aims To evaluate the diagnostic capability of endocytoscopy (ECS) [model GIF-H290 EC,
Olympus] in detecting dysplasia in Barrett’s esophagus (BE).
Methods For each procedure, the BE segment was at first evaluated for lesion localization
by high definition-white light endoscopy. This was followed by further tissue characterization
by assessment of the vascular network through magnifying narrow band imaging. Then
the tissue of interest was rinsed with approximately 8 ml N-acetylcysteine, prior
to spraying the staining mixture of 0.05% crystal violet (10ml) and 1% methylene blue
(1ml). After 90 seconds of absorption time, ECS was performed to assess architectural
and cytological features of the tissue. A comparable procedure was conducted ex vivo for imaging the EMR specimens.
Results This on-going study of ECS in BE patients included at the moment 30 patients. We
imaged 43 sites with ECS, of which we have 32 targeted biopsies containing all stages
of dysplasia in BE. Imaging was considered to be classifiable in 56% (n=24) and unclassifiable
in 44% (n=19). Nevertheless, poor resolution of images (51%) due to patient-related
factors and low quality of staining (56%) often made it hard to interpret in vivo ECS images. Overall, we are able to classify images into three categories including
BE without dysplasia, BE with dysplasia and EAC in vivo and ex vivo.
Conclusions The ECS could provide clear images in which distinct architectural and cytological
features could be identified. However, certain patient-related as well as procedure-related
factors can trouble clear ECS imaging and thus diagnoses in BE patients. In order
to implement ECS into clinical practice, we will build an in vivo atlas with representative images of each category and investigate the use of an artificial
intelligence-aided diagnostic system that could help to enable a highly accurate diagnosis.
