Multi-layer OCT segmentation in the presence of layer-disrupting pathology: Just-Enough Interaction Approach

Abstract

Purpose : In clinical 3D Optical Coherence Tomography (OCT) datasets, current fully automated retinal layer segmentation methods may regionally fail in locations affected by appearance-modifying retinal diseases like Age-related Macular Degeneration (AMD), Diabetic Macular Edema (DME), Retinal Vein Occlusion (RVO), and others. The only current remedy is to tediously edit the obtained segmentation in a slice-by-slice manner. This limits the use of precision medicine in the clinic. We report a new minimally-interactive approach that yields expert-approved layer segmentations in a highly efficient manner.

Methods : Automated LOGISMOS segmentation of 11 retinal layers defined by 12 surfaces is followed by visual inspection of the segmentation results and by employment of minimally-interactive 2D corrections that affect the segmentation in full 3D (Figures). The novel aspect of this "Just-Enough Interaction" (JEI) approach for retinal OCT relies on a 2-stage coarse-to-fine segmentation strategy during which the operator interacts with the LOGISMOS graph-based segmentation algorithm by suggesting desired but approximate locations of the layer surfaces in 3D rather than performing manual slice-by-slice corrections.

Results : The efficiency of achieving reliable multi-layer OCT analysis has been dramatically improved with more than 10-fold speedup compared to the traditional retracing approaches. In an initial testing set of 20 3D OCT datasets from RVO subjects, clinically accurate segmentation was achieved in all analyzed cases after 5.5±1.0 minutes/case devoted to JEI modifications. We estimate that reaching the same performance using slice-by-slice editing in the regions of local segmentation failures would require at least 60 minutes of expert-operator time for the 11 segmented retinal layers.

Conclusions : Our JEI-LOGISMOS approach to segmentation of retinal 3D OCT images is available for clinical research at http://www.iibi.uiowa.edu/content/shared-software-download .

This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.

 

(A,B) 2 slices of RVO OCT volume. (C,D) Automated layer segmentation showing local failures. (E,F) JEI interactions – only 7 inner-retinal (slice 100) and only 5 outer-retinal guiding points (slice 105) yielded regional 3D correction of all retinal layers. (G,H) Resulting layer segmentations, note the 3D effect of local JEI.

 

Regional effect of local JEI (guidance on one slice only, panel E) on 3D inner retinal thickness (top 5 layers).