[1572] Optical Coherence Tomography (OCT) for Intra-Operative Style Interpretation of Endometrium

Ioan C Cucoranu, Jeffrey L Fine. UPMC, Pittsburgh, PA

Background: Optical coherence tomography (OCT) offers direct microscopy of tissue without glass microscope slides. In addition to “room temperature” frozen sections or guided grossing, OCT is also being developed for in vivo imaging with pathological diagnosis. This is a pilot study using a new OCT system to look at endometrium, which is difficult to sample intra-operatively and could be amenable to in vivo imaging.
Design: Signed out, formalin fixed hysterectomy specimens were selected from 2 weekly signout rotations of one of the authors; cases were excluded if the specimen was unsuitable for post-signout imaging (e.g., morcellated uterus, endometrium totally submitted for sampling, etc.). 33 samples from 21 cases were imaged using OCT (Light CT, LLTech, Paris, France); the tissue was submitted for histology, H&E slides were prepared and digitized in whole slide image format (XT, Aperio, California, USA). Regions of interest from OCT and H&E were organized into two powerpoint presentations: training images (10 sets of matched OCT and H&E) and test images (32 sets of OCT only). Three expert pathologists (<5 years experience) reviewed the images on 24” computer monitors. Data captured included time estimates and frozen section style diagnoses (first line with benign/malignant/defer; second line with diagnosis). Data was correlated with H&E diagnoses from a fourth subspecialist pathologist. Inter-observer agreement was calculated by using Fleiss's Kappa statistics.
Results: Pathologists spent 5-10 minutes on training images and 30-60 minutes on test images. Mean deferral rate was 10.66% (range 8-12%); benign/malignant discrepancy rate was 10.44%; 7 cases were not used (no endometrium (4), OCT scan was “enface” (3)). The overall inter-observer agreement between the three subject pathologists was moderate (κ=0.4329).
Conclusions: OCT-novice expert pathologists were able to interpret endometrium without traditional glass microscope slides! In highest-quality images different patterns were plainly visible (e.g. adenocarcinoma, proliferative or secretory patterns (Figure)). Image quality improved with experience. Tissue fixation impaired enface scanning. Enface images may permit wider-area screening intra-operatively, potentially even in vivo, therefore studies with fresh tissue are important.


Category: Informatics

Monday, March 4, 2013 11:45 AM

Proffered Papers: Section H2, Monday Morning

 

Close Window