Angiographic optical coherence tomography imaging of hemangiomas and port wine birthmarks

A current therapeutic challenge of vascular lesions is that they do not always respond effectively to laser treatment. Information on targeted vessels could potentially be used to guide laser treatments. Optical coherence tomography (OCT) is a useful tool for the non‐invasive imaging of tissues, including skin hemangiomas and port wine birthmarks. Dynamic OCT is able to rapidly characterize cutaneous blood vessels. The primary goal of this study was to demonstrate the ability of bedside OCT to image (i) overall vessel pattern; (ii) individual vessel morphology, diameter and depth; and (iii) total vessel density as a function of depth in infantile hemangiomas and port wine birthmarks (PWB).


INTRODUCTION
Treating cutaneous vascular lesions was the first indication for the medical use of lasers. Early therapeutic use sought to remove vascular birthmarks, especially port wine birthmarks (PWB), on the skin of children [1].
PWB are congenital cutaneous capillary malformations composed of ecstatic vessels. PWB's have often previously been called Port Wine Stains. However, a relatively recent survey of patients [2] indicated that they preferred the terminology PWB, so this term will be used in this manuscript. PWBs may be resistant to laser therapy. Despite many studies showing efficacy of lasers for the treatment of PWB, less than 25% of lesions have complete clearing after multiple laser sessions [3]. Methods to improve the treatment effect need to be sought.
Infantile hemangiomas (IH) are benign vascular tumors that are made of progenitor stem cells and disorganized blood vessels [4]. Various options have been explored for the treatment of hemangiomas including pharmacologic, laser and surgical interventions. Several beta-blockers have been evaluated although propranolol is currently standard of care [5,6]. While treatment with beta blockers is a current standard of care, adjunctive laser treatment can provide benefit [7].
The goal of laser therapy for vascular birthmarks is to selectively target aberrant blood vessels present in these lesions. The goal of laser therapy is to induce endothelial cell apoptosis in the vessels that create the hemangioma. In order to selectively destroy vessels, both vessel diameter and depth of lesion must be considered when determining laser settings.
Optical coherence tomography (OCT) is a non-invasive, high resolution imaging method, which can provide real- time information at the bedside. OCT was first described for retinal imaging [8] as early as 1991, but has since been applied to many other parts of the body including the skin [9,10]. OCT uses a tightly focused visible or nearinfrared laser beam to interrogate the surface layers of tissue, and applies the principles of interferometry to enable imaging to depths of 0.5-1.5 mm (depending on the tissue). Resolution is considerably higher than ultrasound, typically on the order of 10-20 mm (in all axes) and OCT provides greater depth penetration than confocal microscopy. Conventional OCT images reveal the layered structure of epidermis, upper dermis, and skin appendages, and  also the presence of blood vessels (as dark areas in the image due to the strong light absorption by water), but it is challenging to have high enough resolution to determine blood vessel characteristics. Attempts were also made to detect blood flow using Doppler-OCT [11], but the technique suffers from the fact that the blood flow direction is orthogonal to the Doppler sensing direction.
In 2010, FDA-cleared commercial OCT imaging systems were introduced (Vivosight, Michelson Diagnostic, UK) using multiple laser beams scanned simultaneously at multiple depths to push the resolution to 7.5 mm over a 6 mm scan width. These devices have been mainly studied in the context of non-invasive, non-melanoma skin cancer diagnosis, and margin mapping. Jemec and co-workers [12] first reported OCT imaging of telangiectasia with this method.
OCT imaging has been applied to evaluation of vascular birthmarks in the past, but results were limited due to technology. In 2001, Nelson et al. [13] were the first to demonstrate that Optical Doppler Tomography (ODT) could image blood vessels in PWB and proposed it as a method for laser therapy optimization. This study measured the vascular response post laser treatment, noting that at low light doses, PDL had only temporary effects on the vasculature of PWB and reperfusion occurred with resultant blood flow to pre-irradiation levels. It was theorized that higher light dosages may be more effective [13]. In 2010, Zhao and co-workers [14] used conventional OCT to scan 41 PWB patients and obtained an average vessel diameter of 95 AE 20 mm and depth of 360 AE 50 mm, and suggested that OCT measurements of vessel dimensions could be very meaningful for clinical treatment dosage. In 2013, the University of California Irvine group demonstrated a 3D Doppler OCT system with enface and projection images of a PWB on two patients and reported that the vessel diameters and morphology varied with depth and between subjects [15]. In 2015, Latrive and co-workers [16] used Doppler OCT to image a PWB and a few infantile hemangiomas and reported some statistics on the observed vessel diameters, noting that the vessel diameters had a wide variation. They concluded that there was a significant difference between vascular characteristics of PWB and infantile hemangiomas.
OCT angiography, otherwise known as Dynamic OCT, has been used to apply statistical analysis of intensity variation to motion, such as that of blood cells, to determine the vessel pattern in the superficial dermal plexus and upper dermal capillaries [17]. The wider potential of OCT angiography in clinical dermatology has been described elsewhere [18,19], along with preliminary validation of the quantitative results [20]. OCT angiography offers a noninvasive tool to image and quantify blood vessels rapidly at the bedside and could be applied to routine clinical use. In this study, we used bedside OCT to image (i) overall vessel pattern; (ii) individual vessel morphology, diameter and depth; and (iii) total vessel density as a function of depth in infantile hemangiomas and port wine birthmarks (PWB).

Subject Population
Under IRB approval, a total of 49 subjects with infantile hemangiomas and PWB were enrolled in the study. Exclusion criteria were pregnancy, breastfeeding, use of oral retinoids 6 months prior to treatment, active infection, or lesions suspicious for malignancy. This is a prospective, single-arm, pilot study conducted to study the morphology, depth, and diameters of blood vessels in infantile hemangiomas and PWB. Written informed consent was obtained from each patient. Patient had OCT scans performed pre-and post-treatment with the PDL (Vbeam Perfecta, Candela/Syneron, Wayland, MA, 7-10 mm spot size; 7.0-9.5 J/cm 2 , 1.5-10 ms, 30/20-20/10 cooling). The research coordinator reviewed laser schedules and indiscriminately assigned the physician to scan a portion of our new infantile hemangioma and PWB patients to receive OCT and subsequent laser treatment over a six-month period.

OCT Procedure
Prior to laser treatment, OCT scans were captured from the lesion site using the OCT probe (Vivosight, Michelson Diagnostics, Maidstone, UK). OCT captured 120 vertical scans, across a 6 Â 6 mm area of skin. Image pixel size was 4 mm; imaging depth was 0.7-1.5 mm. OCT angiography was used to detect blood vessel structure. The "en-face" imaging mode was utilized to visualize the blood vessel plexus from above enabling an initial assessment of the vessel morphology and structure.
The dataset was filtered to remove all pixels with intensity below a threshold chosen to remove background noise (all pixels below 30% of maximum were removed). The dataset was processed with a filter (Frangi and co-workers) implemented in MatLab (MathWorks, Cambridge, UK), Fig. 3. Histograms of vessel diameters at three selected depths in hemangiomas. For example, from this data it is possible to estimate both the cross-sectional area and the volume of vessels of a particular diameter range and depth range that is to be treated by laser. which enhances and extracts vessel-shaped features in a 3D dataset. Results were then skeletonized to extract the vessel network structure. Vessel diameters and depths were extracted for all vessel segments in the resulting vessel "skeleton" to a maximum depth of 0.5 mm (at greater depths the background noise signal is too high for the Frangi filter to effectively discriminate vessel features).
The resulting datasets were amenable to statistical analysis. The total length of vessel segments per square millimeter of tissue was extracted, for each range of vessel diameters and depths. With our method, the total lengths of vessels of a given diameter at a given depth, it is possible to calculate both the cross-sectional area and the volume of vessels presented to an incident laser pulse. The median, upper and lower quartiles, and upper and lower decile vessel diameters were determined to inform the distribution of vessel diameters at each depth range, in 100 mm steps and overall.

RESULTS
Forty-nine subjects (21 hemangioma subjects and 28 PWB subjects) were imaged with OCT. Patient ages ranged from 3 months to 73 years old (Table 1). Table 1 Figure 1A. diversity of age for hemangiomas and PWB along with the variety of locations these lesions are located.
The OCT images revealed that the blood vessel patterns in vascular lesions appeared dramatically different from that seen in normal skin (Fig. 1A). Figure 1A also shows two image of normal cheek skin (not area of vascular lesion) at bottom right for comparison. The red color denotes image pixels in which high dynamic signal intensity was detected, indicating motion due to blood flow. Figure 1B shows the clinical photo of the Figure 1A OCT scanned patients. Figure 2 demonstrates the steps in processing; the raw image of the scan along with a Frangi-filtered image and skeletonized image.
Some datasets exhibited artifacts due to motion of the probe or patient during the scan, which was found to be a problem particularly for very young children. For this analysis we focused on five datasets for infantile hemangiomas and five for PWB without motion artifacts. In both infantile hemangiomas and PWB, blood vessel diameters ranged from 20 to 160 mm (Figs. [3][4][5]. An example of the statistical analysis is shown in Figure 3, which is from the scan of a PWB. It shows that detected vessels ranged in diameter from 20 to 160 mm, and that the quantity of vessels of all diameters was slightly higher at 0.30 mm depth than at 0.22 mm, and slightly higher still at 0.5 mm but the largest vessels in the range 120-160 mm diameter were mainly at 0.50 mm depth. The most common vessel diameter was 60 mm; at 0.5 mm (for example), the total length of vessel segments with this diameter was found to be just over 0.2 mm for each square millimeter of tissue. We used this method of quantifying vascular density because we feel it is more useful than simply providing a count of the number of vessels detected. Scans at depths greater than 0.5 mm were not analyzed with this method because the noise present at greater depths interfered caused problems for the Frangi filter detecting vessel-like features. The histograms show a heterogeneous variation in vessel diameters in all types of lesions, but also wide variations from lesion to lesion (Fig. 6). PWB tended to have larger vessel diameters than hemangiomas but not necessarily higher density. Nevertheless, because the vessels are larger in diameter, the total blood vessel volume is likely higher for PWB (Fig. 7). Figures 8 and 9 are contour plots of the vessel distribution for both hemangiomas and PWB as a function of diameter and depth; the z-direction representing total vessel length per square millimeter of tissue. Comparing Figure 8 with Figure 9, we see that for this small sample of lesions, PWB tended to have larger diameter vessels at shallow depths and that hemangiomas exhibited higher vessel density as depth increased.
Summarizing the above data sets we note a trend that hemangiomas have significant variation in the diameter of the vessels that comprise the lesions. The median diameter of hemangioma in our data set were 50-70 mm. It was also noted there is a wide variation in overall   vessel density with hemangiomas. The hemangioma blood vessel densities increase with depth (Figures 8  and 9). The port wine birthmarks trends showed they were greater in diameter than the hemangiomas with the median diameter 70-100 mm. In addition the port wine birthmarks were thicker and had more total volume than the hemangiomas (Fig. 7).

DISCUSSION
OCT is an imaging tool, which can be applied for assessment of skin vascular lesions in vivo and in real time in a non-invasive bedside imaging procedure. This procedure is easily done and taught in clinical practice. A limitation is that the existing OCT technology does not allow statistical analysis of vessels deeper than 0.5 mm, but this may improve in the future as the technology develops.
In this study, we demonstrated blood vessel diameters range from 20 to 160 mm. There was a heterogeneous population of both small and larger blood vessels in both infantile hemangiomas and PWB. Vascular OCT may be used to give us information to individualize laser settings. Furthermore, comparing the statistics of vessel diameters and OCT images of the lesions with the clinical photographs, it is clear that the visible appearance of a PWB or hemangioma is a complex function of vessel diameter, density, and depth that requires further study. For example PWB#4 has relatively few vessels, but they are large diameter and occur at shallow depths, whereas PWB #1 has many more vessels but they are mostly much smaller diameter and tend to be deeper, yet both lesions appear clinically similar. OCT provides a powerful tool for further research into vascular lesion structure and how they present clinically.
In 1983, Anderson and Parrish [21] established the concept of the 'Thermal Relaxation Time' (TRT) for blood vessels. They explained how pulses with shorter duration than the TRT of a vessel result in the heating effect of the laser pulse to be confined to the vessel, limiting the collateral damage to the surrounding tissue. The TRT is related to the square of the vessel diameter. Our results indicate the complex nature of vessel diameter and depth distribution in hemangiomas and PWB. Currently clinicians do not know the vessel characteristics of lesions they are treating. We speculate that more complex treatment protocols are required, tailored to individual vessel diameters and depths in a lesion, which can be observed using OCT. Optimized protocols may require utilization of multiple pulse durations, in the same or in multiple sessions. It might be efficacious to start with longer pulses with pulse durations set for the largest vessels and fluence for purpura. Once the largest vessels have been destroyed, then the lesion can be re-imaged and TRT set for the next size down that are still viable. Performing multiple passes, will require the physician to carefully monitor epidermal health to avoid damage to the skin surface that could results in scarring or permanent discoloration. Based on the trends seen with our OCT images the actual diameters of the blood vessels appear larger than expected with a median blood vessel diameters ranging from 50 to 100 mm, suggesting that longer pulse durations than conventionally used may be more effective especially for the PWB. Additional OCT data with larger number of patients is needed to further evaluate these blood vessels diameters.

CONCLUSION
OCT allowed rapid, non-invasive characterization of the diameter and depth of blood vessels in individual vascular lesions. Imaged lesions consisted of a heterogeneous population of vessel sizes, morphologies, and depth. Future studies could utilize this information to assist development of individualized treatment protocols in an effort to improve vascular birthmark removal.
OCT gives us a tool to rapidly and non-invasively characterize the diameter and depth of blood vessels in individual vascular lesions. We demonstrated the vascular lesion consist of a heterogeneous population of vessel sizes, morphologies, and depth distributions. Use of this information has great potential to evaluate and individualize treatment protocols and ultimately, may improve treatment of vascular birthmarks.