Quantification Metrics for Telangiectasia Using Optical Coherence Tomography

This pilot study was an exploratory study aimed at developing a protocol for the acquisition, processing, analysis and quantification of OCT images of HHT skin lesions. As a result of this pilot project, an integrated protocol can be defined that will optimize patient experience, user-friendliness, image quality and reliability, and measurement accuracy.

Current practice in the analysis of telangiectasias relies mainly on visual inspection or histological imaging. In these practices, monitoring and evaluation is mostly qualitative or done by measuring approximate visible surface diameters12.13. In this article, we used OCT-A and developed specific measures capable of following lesional changes and dynamic patterns of angiogenesis and blood flow, as well as structural changes in vivo. Dynamic angiographic parameters include VF – which was able to quantify blood flow from the lesion and supporting vasculature, VLA – which provided isolated information on blood flow changes in the lesion area, LA – which showed information on the spread of the lesion and possibly indications of angiogenesis and VALVe, which describes the density of the lesions. FD described size and complexity with lower fractal dimension indicating lower complexity and reduced vascularity. EF indicated changes in the epidermis related to the lesion. Each metric provided important information about the lesion and its progression and were considered together to provide a comprehensive analysis.

All lesions showed unique variations across measurements indicating that significant morphological changes in lesion structures were detected over time. Although VF, VLA, and LA all depend on blood flow to the lesion, changes over time in each of these measures for a single lesion did not necessarily follow the same trend, indicating the importance of collecting and to analyze each of the measurements. separately. Additionally, differences between superficial and deep metrics were also noted, revealing differences in lesion response to depth and reinforcing the need to measure metrics at different depths. Large-scale, easily discernible changes are easy to observe visually, but differences in finer detail may not be noticeable, requiring quantitative measurements capable of detecting small changes in morphology and angiogenesis not visible at normal perception. The quantitative measurements produced show good potential for a careful study of the underlying dynamical processes occurring below the surface.

Few major changes or correlations could be determined from the quantitative measures used above. Some lesions seemed to have little or no response, while others changed dramatically. As shown in Figure 5, all lesions appeared to show a general decline in values ​​related to the amount of blood flowing through the vasculature; however, P3L2 seemed to show the opposite trend. It is difficult to determine whether these changes are a consequence of the treatment, natural changes in the morphology of the lesions or simply due to the variability of the external environment affecting the lesions. It is very possible that the treatment had no effect on the lesion and any changes noted in this study were due to variability in environmental factors and patient physiology. Moreover, since the lesions themselves varied in size and morphology, some measures are more effective than others in describing changes in morphology. Future studies with larger treatment and control patient cohorts are needed to validate this pipeline for use as a clinical treatment assessment tool.

Since the vascular fraction assesses all opposite imaged area, changes to the lesion and surrounding vasculature would impact this metric. Although used in analysis in other fields, this metric would be subject to error due to global changes affecting the entire imaging area, and therefore may not be representative of the dynamic processes of the lesion. Comparatively, VALVe isolates the vasculature in the lesion area and would provide a more accurate analysis of lesion evolution over time. This is more clearly visible in Figs. 5a-d, where although a decrease in both deep and superficial VF is observed, a very slight increase in superficial VALVe and a variable trend for deep VALVe are observed. Thus, this more direct approach could potentially yield better insights into morphological changes compared to more traditional measurements like VF.

This analysis also provided valuable information on potential sources of variability in a lesion, related to both environmental factors and biological processes. Throughout the study, many lesions showed a peak in parameters late in treatment. Many imaging sessions started during a colder season and the peak in metrics was noticed in imaging sessions performed during warmer seasons. From this we can predict that the dilation of the blood vessels was probably due to the change in temperature. This expansion can be observed in the sample data provided in Tables 1 and 2, as well as in Figure 2. The test performed with a single lesion under different temperature conditions, shown in Figure 6, also demonstrates this phenomenon. The same lesion appears dramatically different on arrival, after heating, and after cooling, highlighting the need for drastic temperature controls in future studies. A method of lesion characterization and clustering can also be observed through VALVe and EF. These parameters provide valuable information about lesion density and concentration, and could serve as useful tools in treatment planning. In this, we propose to classify the lesions as “dense” or “diffuse” according to the VALVe measurement, and as “deep” or “superficial” according to EF. Figures 3 and 4 help demonstrate the differences between these classifications. While a binary classification may be possible, a more graduated scale may need to be employed; however, a larger dataset with more lesions should be studied in order to numerically assign values ​​to these classifications. Nevertheless, these characterizations can help in treatment planning. For example, a deep, diffuse lesion may respond less well to topical therapies or may require adjustments in laser power to allow sufficient penetration for ablation of the lesion using laser therapies. It is also possible to apply these characterization measures in diagnostic pipelines for younger patients or to diagnose uncertain lesions; however, further studies are needed to determine threshold metrics that could successfully differentiate telangiectasia from other malformations. OCT can easily be adapted and used in a workflow to determine specific correlations not only for HHT, but beyond other dermatological and lesion analyzes that were previously performed only by qualitative visual inspection.

The image acquisition process also revealed some systematic issues that could be improved. The main issue was that image acquisition is very slow, with a single scan requiring up to 3 min. This long acquisition period made the process very prone to motion artifacts. Many collected scans were rendered unsuitable for quantification due to these large motion artifacts. This long acquisition time also limited the achievable imaging regions. Some lesions of interest found on other skin surfaces were not feasible for imaging because the patient and/or technician were unable to maintain their position throughout the acquisition period. Additionally, the current probe system has a plastic offset that defines a focusing distance between the laser source and the skin surface. Depending on the location or consistency of the skin, the thickness of the epidermis varies and the plastic offset must be modified. This can be cumbersome, with optimal offset often unavailable. Additionally, holding the probe immobile against the skin will often induce pressure around the imaging region which can affect fine blood flow, potentially impacting observed angiographic measurements. After imaging an area, there was often shrinkage on the skin surface due to the plastic shift. Overall, system improvements including faster image acquisition and adjustable probe distance would increase the application range of the current system. A fully contactless probe would be a preferable solution, but may not be feasible without significant improvements in system acquisition and bulk motion compensation mechanisms. Potentially, a real-time imaging system could alleviate much of this burden, but would require significant hardware improvements and optimization of current processing methods to be fully realized. Additionally, a previous study indicated that structural OCT image quality may be slightly affected by skin tone; however, further investigation into the limitations of OCT-A with respect to skin tone needs to be conducted.26. Future improvements in acquisition methods to combat this problem would likely benefit this study as well as many others.

About Roberto Frank

Check Also

OnePlus 10 Pro: long-term review

Over the years, OnePlus has managed to transform itself from an emerging player into a …