1.

Introduction

The effectiveness of vascular lesion treatment is usually evaluated by visual inspection by a dermatologist. Color images are taken before and after treatment in order to qualitatively assess the results of the treatment; they also serve as evidence in case of complaints.

Several studies of techniques for quantitative assessment of therapy¹ have been performed by laser speckle imaging,² laser Doppler perfusion imaging,³ optical coherence tomography,⁴ optical Doppler tomography,⁵ reflectance confocal microscopy,⁶ pulsed photothermal radiometry,⁷ diffuse reflectance spectroscopy, and spectral imaging.^8,9 Skin color image analysis has been offered as a simple and fast method for treatment evaluation,^10,11 but there are some limitations due to errors associated with surface reflection and the effects of skin chromophores other than hemoglobin.¹²

Multispectral imaging can be used for noncontact assessment of skin chromophore (e.g., oxy-/deoxyhemoglobin, melanin) distribution; it reflects the overall skin condition and may facilitate better pathology diagnostics.^13,14 However, commercial multispectral imaging cameras are bulky and expensive, thus limiting their clinical implementation. A digital color RGB camera can be regarded as a low-cost alternative. It acquires three spectral images [red (R), green (G), and blue (B)] simultaneously and can be used as a simple and fast spectral imaging device. In combination with specific narrow-band spectral light sources, RGB imaging could become competitive for some specific applications including the assessment of hemoglobin concentration.^15–18

Erythema index (EI) is a useful parameter for estimating the cutaneous hemoglobin content and quantitatively characterizing vascular lesions. Simplified methods for determining EI are based on comparing the skin optical density (OD) in green (high hemoglobin absorption) and red (low hemoglobin absorption) spectral ranges.¹⁹ Sometimes EI is attributed also to blood contrast²⁰ or tissue viability.^16,21 Derma Spectrometer® and Mexameter® are commercially available instruments for point measurements of EI.²² The contribution of melanin in the skin absorption processes exploited for EI calculations is usually neglected. However, the difference in OD between the blue and green spectral ranges²⁰ or a weighting factor applied to the absorption index in the red band²³ can be used to correct for the melanin absorption.

EI difference ($\mathrm{\Delta }\mathrm{EI}$) images representing difference between the port-wine stain (PWS) and normal skin areas were analyzed before and after laser therapy, suggesting that $\mathrm{\Delta }\mathrm{EI}$ may be used as an effective parameter to monitor PWS treatment.^24,25 Imaging techniques support visual inspection of the treatment result and allow spatial analysis of the skin lesion, for example, to compare the area of certain $\mathrm{\Delta }\mathrm{EI}$ threshold level before and after treatment.²⁴

Image analysis still requires user-friendly interface improvements before routine clinical use.¹ Dermatologists are interested in a simple, fast, and low-cost technique that supports visual inspection and evaluation of the treatment result.

A handheld prototype RGB imaging system for mapping and monitoring of the hemoglobin distribution in the skin has been designed and tested. The previous study showed that this system can be used to monitor hemoglobin concentration changes during specific provocations such as arterial/venous occlusions and heat test.^17,18 The aim of this study was to clinically test the handheld RGB imaging system for evaluating the vascular lesions’ laser and intense pulsed light (IPL) treatment results. The contrast of EI (CEI) was proposed as a parameter for simple and fast quantitative characterization of vascular lesions. The multispectral imaging system was used as a reference.¹³

2.

Materials and Methods

2.1.

Data Acquisition

Measurements were taken with a homemade handheld RGB imaging system [Fig. 1(a)] that consists of a commercial RGB CMOS sensor (USB uEye LE from IDS Imaging Development Systems, GmbH, Germany) within a ring of RGB light-emitting diodes (LEDs). Two orthogonally oriented polarizers were placed in front of the CMOS camera and RGB LEDs, respectively, in order to suppress specular reflection from the skin surface.²⁶ The system was adapted for a working distance of 3 cm, providing a $3\times 3\mathrm{cm}$ field of view and a spatial resolution of $0.05\times 0.05\mathrm{mm}$ per pixel. The imaging camera was connected to a PC via a USB interface. All settings were manual and were maintained constant during the measurements. Measurements were taken in darkness to reduce artifacts caused by the ambient light. A single measurement lasted $<10\mathrm{s}$; most of the time was spent for positioning and less than a second for taking a snapshot.

Fig. 1

RGB imaging system: color image (a) and spectral sensitivity of each color channel (b).

The multispectral imaging system was used as the reference. It consisted of a multispectral imaging camera (Nuance from Cambridge Research & Instrumentation Inc., Woburn, Massachusetts), a halogen lamp ring light, and a pair of crossed polarizers for reducing the specular reflectance. The system was adjusted for a spatial resolution of $0.75\times 0.75\mathrm{mm}$ per pixel and a spectral resolution of 15 nm (bandwidth of the Nuance liquid crystal tunable filter). The spectral operating range was 450 to 950 nm.¹³ The measurement lasted several minutes—most of the time was spent for positioning and $\sim 1\min$ to acquire the spectral image cube (51-image stack in steps of 10 nm). The multispectral data were further used for extracting absorption spectra from the regions of interest.

3.

Results

Figure 2 shows a typical vascular lesion telangiectasia (v1) that appears darker and redder compared to the surrounding tissue (n1). A distinctly darker spot can be observed in the center of the malformation. The optical density spectra [Fig. 2(a)] indicate that the increased absorption corresponds to the characteristic bands of hemoglobin between 500 and 600 nm.

Fig. 2

Optical density spectra (a) before (full line), directly after (dashed line), and a month after (dotted line) treatment of telangiectasia (red) and normal surrounding tissue (black), and corresponding optical density maps (b) at six different wavelengths.

The redness of the skin increased in the lesion and its surrounding tissue (n2) after the laser therapy (1064 nm Nd:YAG laser, $160\mathrm{J}/{\mathrm{cm}}^2$, 15 ms). OD spectra indicate increased hemoglobin concentration. The center of the telangiectasia (v2) appears darker and changes color to gray or black. Blood vessels are destroyed by phototherapy in this skin region; increased OD values in the range of 600 to 800 nm indicate increased deoxyhemoglobin and methemoglobin concentration levels.²⁷

A significant improvement is normally observed within a month after the treatment. The telangiectasia (v3) is barely noticeable to the naked eye, but still has a significantly different spectrum in the range of 450 to 600 nm compared to the surrounding tissue (n3).

The color images of skin before, immediately after, and a month after laser treatment along with the corresponding OD maps are presented in Fig. 2(b). Telangiectasia appears as a spot with a higher OD value. The best contrast between the vascular lesion and the surrounding tissue can be achieved in the green ($\sim 550\mathrm{nm}$) spectral range, while in the red ($\sim 650\mathrm{nm}$) spectral channel, absorption and contrast are significantly reduced. Therefore, the difference in OD values between the green and red spectral channels is used to calculate the EI. This parameter represents the cutaneous hemoglobin content but does not provide information about different hemoglobin types and oxygenation.

RGB images with the corresponding EI maps and CEI values of the same case of telangiectasia are shown in Fig. 3(a). The CEI decreased by 50% a month after treatment, and the vascular malformation was no longer visible to the naked eye. Decrease in CEI (for 25%) was also observed directly after treatment, but it was mainly caused by irritation of the surrounding skin.

Fig. 3

Examples of telangiectasia (a) and hemangioma (b) before, directly after, and a month after the laser treatment: color images with the corresponding erythema index maps and contrast of erythema index values.

An example of hemangioma treatment is shown in Fig. 3(b). If compared to the telangiectasia before the treatment, the hemangioma had a contrast of EI three times higher. The treatment was also performed with the 1064-nm Nd:YAG laser ($160\mathrm{J}/{\mathrm{cm}}^2$, 20 ms), resulting in a 72% reduction of the CEI value. Irritation of the surrounding skin that caused 46% decrease in CEI was also observed.

The average contrast of EI values of all 31 lesions and their surrounding normal skin were calculated [Fig. 4(a)]. The contrast of EI is much higher (on average four times) for vascular malformations compared to the surrounding normal skin. The variation of CEI for vascular malformations is also higher due to different types of lesions. The CEI and its variation tend to decrease after laser treatment by 45%, on average. Related-samples Friedman’s two-way analysis of variance test shows that CEI values of vascular lesions before, directly after, and a month after treatment have statistically significant differences ($p<0.0001$). However, the CEI remained stable (low and small standard deviation) for the surrounding normal skin, except directly after treatment, when irritation of the surrounding tissue occurred. It is possible to determine a clear boundary in CEI between normal skin and vascular lesions ($\mathrm{CEI}=0.03$) for this study.

Fig. 4

Average contrast of erythema index (CEI) values of 31 lesions before, directly after, and a month after treatment (a) and comparison of the CEI values before and a month after treatment for all examined lesions (b). The lesions are colored in red, normal skin in black.

A comparison of the CEI values before and a month after treatment is presented in Fig. 4(b). Normal skin measurements (black dots) cluster around a CEI value of $\sim 0.02$, but vascular malformations (red dots) have a much higher CEI with values that are widely scattered. The threshold level indicated by the horizontal line at $\mathrm{CEI}=0.03$ (after treatment) separates the lesions that reached the contrast of normal skin.

Improvement in skin recovery can be presented as a percentage decrease in the CEI by comparing the CEI values before and a month after treatment. A histogram of the improvement in CEI for all 31 lesions is shown in Fig. 5(a). One case [Fig. 5(b1)] did not show improvement; this lesion was a pathology that appeared as a PWS. Up to 10% improvement in CEI was found in three cases [two of them are shown in Figs. 5(b2) and 5(b3)]. The dermatologist rated these four cases as no improvement. Slight improvement (10 to 30%) was found in six cases and significant improvement (30 to 60%) in seven cases. Seven treatments showed very good improvement ($>60\%$), and of these, the largest value for improvement in CEI was 79%. The same ratings were received from the dermatologist [Fig. 5(a)]. In the case of seven lesions, phototherapy was fully successful, and the vascular lesion totally disappeared; measurements a month after treatment could not be made.

Fig. 5

Histogram of improvements in the CEI of 31 lesions and the corresponding ratings from the dermatologist (a) and color images of unsuccessful treatment samples: port-wine stain (b1), superficial venous network (b2), and telangiectasia (b3).

Correcting for the melanin absorption^20,23 does not fully remove the influence of melanin on EI. Reducing melanin does help to better highlight vascular features in the EI maps. Along with the previous studies,^24,25 we assume that the epidermal melanin content does not change between the measurements. Patients are usually recommended to avoid tanning and other skin irritating procedures after the treatment. All the subjects studied had Fitzpatrick skin type II, so the skin melanin content was assumed to be similar. Moreover, the CEI of the surrounding normal skin had remained stable with low values and small standard deviation. The influence of different melanin levels on the CEI parameter should be studied further. Additional wavelengths in the infrared spectral range could be used for more precise measurements and for correcting the melanin content.²⁸

Our results showed that the measured improvement in CEI corresponded well with the dermatologist’s rating and thus, supplemented a qualitative measure of the effectiveness of phototherapy with a quantitative parameter that can be obtained easily from the RGB image data. Further improvements could be the implementation of an RGB imaging device that can be connected to a common clinical device or smartphone.

4.

Conclusions

The developed RGB imaging system appeared to be useful for quantitative evaluation of the effectiveness of phototherapy of vascular lesions. The EI was chosen as a parameter for estimating the cutaneous hemoglobin content. The CEI was proposed as a parameter for quantitative characterization of vascular lesions; it represents the homogeneity of the hemoglobin content within the region of interest. The degree of skin recovery is reflected as a percentage decrease in the CEI values before and after phototreatment.

This study demonstrates the potential of a simple RGB imaging technique for implementation in routine clinical praxis. It is a fully noninvasive technology; other advantages of the RGB imaging are small instrument size, good spectral performance, and low equipment cost. The simplicity of computing parameters allows nearly real-time monitoring. The RGB imaging system and algorithm described above could be further implemented as a connection kit to some common clinical device or smartphone.