Quantification metrics for telangiectasia using optical coherence tomography

Abstract

Hereditary hemorrhagic telangiectasia (HHT) is an autosomal dominant disorder that causes vascular malformations throughout the body. The most prevalent and accessible of these lesions are found throughout the skin and mucosa, and often rupture causing bleeding and anemia. A recent increase in potential HHT treatments have created a demand for quantitative metrics that can objectively measure the efficacy of new and developing treatments. We employ optical coherence tomography (OCT)-a high resolution, non-invasive imaging modality in a novel pipeline to image and quantitatively characterize dermal HHT lesion behavior over time or throughout the course of treatment. This study is aimed at detecting detailed morphological changes of dermal HHT lesions to understand the underlying dynamic processes of the disease. We present refined metrics tailored for HHT, developed from a pilot study using 3 HHT patients and 6 lesions over the course of multiple imaging dates, totalling to 26 lesion images. Preliminary results from these lesions are presented in this paper alongside representative OCT images. This study provides a new objective method to analyse and understand HHT lesions using a minimally invasive, accessible, cost-effective, and efficient imaging modality with quantitative metrics describing morphology and blood flow.

Figure 1

Image analysis pipeline for lesion quantification. (a) Enface image generated by DSDLI, (b) filtered and binarized image, (c) isolated lesion vasculature, (d) dilated and closed image, (e) structural slice image, and (f) shows the depths from where the deep and superficial images are constructed from. The axis in each image represents a 1 mm × 1 mm bar.

Figure 2

Progression at the superficial (a-d) and deep (e–h) slice-stacks, and the structural cross-sections (i-l) over the course of four imaging sessions P2L1. The axis in each image represents a 1 mm × 1 mm bar.

Figure 3

Three lesions from different patients demonstrating very different morphologies classifiable by VALVe. (a-c) is a deep slice example of a dense lesion, P2L2, with a VALVe = 0.903, (d-f) is a superficial slice example of an intermediate lesion, P1L1, with VALVe = 0.512, and (g-i) is a superficial slice example of a diffuse lesion, P3L1, with a VALVe = 0.290. The axis in each image represents a 1 mm × 1 mm bar.

Figure 4

Two lesions from different patients demonstrating EFs and classifications. (a) is a superficial lesion, P2L2, with EF = 0.46 and (b) is the surface photograph of P2L2. (c) is a deep lesion, P1L1, with EF = 0.93 and (d) is the surface photograph of P1L1. The axis in each image represents a 1 mm × 1 mm bar.

Figure 5

Graph of results for (a,b) vascular fraction, (c,d) VALVe, (e,f) Fractal Dimension, and (g) epidermal fraction, with (a,c,e) being superficial and (b,d,f) being deep, with (h) as a legend for all graphs.

Figure 6

Vascular images directly from the VivoSight showing the same lesion (a) upon arrival, (b) after heating, (c) after cooling. The axis in each image represents a 1 mm × 1 mm bar.