Objective assessment of skin microcirculation using a smartphone camera

Skin Res Technol. 2020;00:1–7. wileyonlinelibrary.com/journal/srt  

|

|

DOI: 10.1111/srt.12919 O R I G I N A L A R T I C L E

Objective assessment of skin microcirculation using a

smartphone camera

Erik Tesselaar

_{| Simon Farnebo}

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.

© 2020 The Authors. Skin Research and Technology published by John Wiley & Sons Ltd

1_{Department of Radiation Physics,}

Department of Clinical and Experimental Medicine, Linköping University, Linköping, Sweden

2_{Department of Hand and Plastic Surgery}

and Burns and Department of Clinical and Experimental Medicine, Linköping University, Linköping, Sweden Correspondence

Erik Tesselaar, Department of Medical Radiation Physics, Department of Clinical and Experimental Medicine, Linköping University, 58185 Linköping, Sweden. Email: erik.tesselaar@liu.se

Funding information

This study was financed with support from the County Council of Östergötland, Sweden.

Abstract

Background: Existing techniques for assessment of microcirculation are limited by their large size and high costs and are often not so easy to use. Advances in mobile technology have enabled great improvements in smartphone sensor technology. In this study, we used SkinSight, an app for iPhone and iPad, to measure changes in skin microcirculation during physiological provocations. The system estimates changes in the concentration of hemoglobin in the skin by analyzing the reflected light emitted from the built-in light-emitting diode and detected by the camera of the smartphone. Methods: A relative hemoglobin (Hb) index was measured during a 5-min arte-rial occlusion, post-occlusive reactive hyperemia, and a 5-min venous occlusion in 10 healthy subjects, on two separate days. The index was calculated in an area of the skin from the color information in the images acquired by the phone camera. Polarized light spectroscopy imaging was used to measure changes in red blood cell concentration for comparison.

Results: During arterial occlusion, relative Hb index was unchanged compared to baseline (P = .40). After release of the cuff, a sudden 60%-75% increase in Hb index was observed (P < .001) followed by a gradual return to baseline. During venous occlusion, Hb index increased by 80% (P < .001) followed by a gradual decrease to baseline after reperfusion. Day-to-day reproducibility of the relative Hb index was excellent (ICC: 0.92, r = 0.94), although relative Hb index was consistently higher during the second day, possibly as a result of changed lighting conditions or calibra-tion issues.

Conclusion: Microvascular responses to physiological provocations in the skin can be accurately and reproducibly measured using a smartphone application. Although the system offers a handheld, easy to use and flexible technique for skin microvascular assessment, the effects of lighting on the measured values and need for calibration need to be further investigated.

K E Y W O R D S

hemoglobin, microcirculation, post-occlusive reactive hyperemia, smartphone, venous occlusion

1 | INTRODUCTION

In recent years, great improvements have been made in smartphone camera technology as a result of rapid advances in camera optics, image sensors, illumination, sensitivity, and image processing capa-bilities. For the majority of regular consumers, the advantages of dedicated cameras do not longer outweigh the burden of having to carry an extra device, and therefore, the smartphone has become their primary camera. This tendency is also increasingly seen in medi-cine, where bulky, professional-grade, offline diagnostic instruments are falling out of grace and being replaced by more mobile, smart devices. Examples include smartphone-based ECG monitors such as the KardiaMobile from AliveCor, ultrasound scanners such as the Philips Lumify, dermascopes (DermLite, Virtual Dermatoscope), oph-thalmoscopes (D-Eye, iExaminer), and even mobile DNA sequencers (SmidgeION).

Measuring changes in skin microcirculation is of interest in vari-ous experimental and clinical settings, such as in testing endothelial

function1 _{and in the assessment of burns}2,3 _{and free tissue}

trans-fers.4,5 _{Changes in the skin microvasculature can be induced using}

physiological and pharmacological provocations6 _{and can reveal}

vascular pathologies including endothelial dysfunction and vascular

congestion,4,7 _{but also skin damage as a result of sunburn,}8 _thermal

burns,3 _{or ionizing radiation.}9

Various techniques for assessment skin microcirculation are in use today. Optical techniques such as laser Doppler flowmetry, laser Doppler imaging, laser speckle contrast imaging, and optical coherence tomography, are the most common techniques for study-ing microvascular function in the skin. All these techniques provide measures of microvascular perfusion by combining estimates of the red blood cell concentration and the flow of the red blood cells in the skin. Although proven reliable and accurate, these techniques all require relatively large and bulky equipment that are cumbersome to use at the bedside or in an operating room. More portable tech-niques based on spectroscopy imaging have recently been found to be sensitive in detecting vasoconstriction and venous stasis and can assess a relatively large area of skin, which could make them particularly suitable for use in burn assessment or in monitoring of tissue transplants during and after reconstructive surgery as well as continuous wound assessment over time. Miniature thermal imaging cameras FLIR ONE™ (FLIR Systems, Wilsonville, Ore) have also been used in the evaluation of patients undergoing perforator-based free tissue transfer and limb perfusion with a high degree of accuracy

and reliability.10

We developed a mobile phone application for iPhone (Apple Inc, Cupertino, CA, USA) that uses the built-in camera and flash of the phone to measure changes in skin microcirculation, using an image analysis algorithm based on the spectral properties of he-moglobin in the skin. We hypothesized that microvascular changes during arterial occlusion, venous occlusion, and post-occlusive reactive hyperemia could be registered accurately and in an easy way using this app. Thus, the aim of this study was to evaluate the feasibility of using a smartphone as a device for microvascular

measurement. To compare the smartphone measurements with established measures of microvascular reactivity, we used polar-ized light spectroscopy imaging.

2 | METHODS

2.1 | Subjects

Ten healthy subjects (5 women), mean age 26 (range 22-37) years, mean height 1.77 (SD = 0.10) m, mean weight 67.1 (SD = 9.9) kg,

mean BMI 21.3 (SD = 1.7) kg/m2_{, mean systolic blood pressure}

116.8 (SD = 9.4) mm Hg, were included in the study after they had given written informed consent. The subjects were asked not to eat or drink anything that contained caffeine on the day of the experiment. Reasons to exclude subjects were if they had cardiovascular disease, diabetes, skin diseases, used nicotine or medication regularly (except oral contraceptives) or if they had a systolic blood pressure of > 150 mm Hg, or a diastolic blood pres-sure of > 90 mm Hg. Blood prespres-sure was meapres-sured before and after the experiment. The study conformed with the Declaration of Helsinki and was approved by the regional ethics review board at Linköping University, Sweden.

2.2 | Smartphone imaging

A mobile application, “SkinSight,” for iOS (Apple Inc, Cupertino, CA, USA) was developed to capture images using the camera of an iPod Touch, (6th generation, Apple Inc) under white-light illumination using the built-in light-emitting diode (LED) flash. The exposure time, lens aperture, and white balance were kept constant while focus was automatically adjusted by the camera with each exposure. Images were captured at full resolution and were stored on the device. After the experiments, the images were transferred to a separate

com-puter and analyzed using ImageJ.11

Images were divided into their red, green, and blue image planes,

and for each pixel, a hemoglobin index (CHb) was calculated using

a modification of the technique presented by Liu and Zerubia.12 _In

short, changes in intensity in the green and blue images planes are largely caused by changes in the concentration of hemoglobin and melanin, while changes in the red plane mainly depend on changes in the concentration of melanin. By combining the information from the three image planes and based on absorption coefficients and the penetration depth of the illuminating light, the concentration of

hemoglobin (CHb) can be estimated from the red (R), green (G), and

blue (B) pixel values for each pixel in the image using the following algorithm:

The mean CHb was then calculated in a region of interest of the

skin.

2.3 | Tissue viability imaging

A TiVi camera system (TiVi600, WheelsBridge AB, Linköping, Sweden) was used to measure changes in the concentration of RBC in the skin. The system consists of a digital camera equipped with perpendicular

polarization filters in front of the flash and lens.13 _{The broad-spectrum}

white flashlight becomes linearly polarized when passing the first po-larization filter. Reflected light from the skin consists of both linearly polarized (directly reflected) and randomly depolarized (“subsurface”) light. A perpendicularly placed polarization filter in front of the lens prevents any directly reflected light from reaching the photo-array in the camera. The RBC in the microcirculation absorb light in the green wavelength region (about 500-600 nm) to a much higher extent than light in the red wavelength region (about 600-700 nm). By comparison, the surrounding dermis absorbs less light, and this absorption is not as wavelength dependent as that of the RBC. An image-processing algo-rithm uses this difference in absorption and produces a TiVi image, of which the pixel values are linearly proportional to the local

concentra-tion of RBC (C_RBC) in the skin. The system is relatively insensitive to the

oxygenation of RBC.13

Images were analyzed using TiVi analysis software (TiVi Version 2.1, WheelsBridge AB, Linköping, Sweden) and customized software (MATLAB R2007b, The MathWorks Inc, Natick, MA). The images contained the complete volar aspect of the lower forearm from the elbow to the wrist.

2.4 | Experimental protocol

Subjects rested semisupine in a room with a controlled temperature of 22 (0.5)°C. A pressure cuff was attached around the right upper

arm. The forearm was kept at heart level with the volar surface up-ward and supported by a pillow. The TiVi camera and the smart-phone were mounted on a camera stand positioned 30 cm above the volar forearm such that images were captured at a 90° angle (Figure 1).

After the baseline had been established, blood pressure was measured in the other arm using an automatic sphygmomanometer (M6 Comfort, Omron Healthcare, Hoofddorp, The Netherlands). The forearm was then placed level with the heart and was kept exsangui-nated for 5 min, while images were taken with both techniques at 10-second intervals.

After 5 min of occlusion, the pressure cuff was deflated, and im-ages of the forearm were taken at 1-second intervals during the sub-sequent 15 min to measure the post-occlusive reactive hyperaemic response.

After a recovery period of 15 min, the pressure cuff was inflated to 30 mm Hg above the diastolic pressure and was kept at that level for 5 min to establish venous stasis in the forearm. During this time, images were acquired at 10-s intervals. Finally, images were acquired during a 10-min recovery interval (Figure 1).

2.5 | Analysis of data

The same area on the volar side of the forearm was analyzed by

smartphone imaging and TiVi. In each image, the mean CHb and

C_RBC value of all pixels in the same 3 × 5 cm region of interest was

calculated. Data in figures are shown as mean, with error bars rep-resenting ± 1 standard deviation. Correlation between responses at different days was calculated as Pearson's (r) and intraclass cor-relation coefficients (ICC). Statistical calculations were done using

F I G U R E 1   Schematic image of the experimental setup. Provocations by arterial and venous occlusion were randomized so that in half of the study population arterial occlusion was done first followed by venous occlusion, and the opposite in the second half. All experiments were preceded by a acclimatization period of 15 min. Measurements were done with both SkinSight and TiVi for the same ROI

BaselineArterial/Venous Occlusion Recovery Baseline Recovery

–15 min 0 5 10 20 25 30 40 min Arterial/Venous

Occlusion Acclimatization

Blood pressure cuff

Microsoft Excel 2008 for Mac and GraphPad Prism version 6.0b for Mac OS X (GraphPad Software, San Diego California, USA). For all analyses, probabilities of less than 0.05 were accepted as significant.

3 | RESULTS

3.1 | Smartphone imaging (C

)

Figure 1 shows the change in CHb as measured using SkinSight,

dur-ing arterial occlusion and post-occlusive hyperemia (A) and durdur-ing venous occlusion and subsequent recovery (B). At baseline, the

in-ter-subject variation of C_Hb was 17% on day 1 and 13% on day 2.

During arterial occlusion, no significant change in CHb was observed

(P = .61), but the inter-subject variation increased to 34% and 20% on days 1 and 2, respectively. After release of the cuff and reperfu-sion of the forearm, a rapid, 60%-80% increase in hemoglobin index was observed (P < .001) followed by a gradual return to baseline. The inter-subject variation of the peak response during post-occlusive

hyperemia was 9% (day 1) and 7% (day 2). During the first 4 min of

venous occlusion, CHb gradually increased to about 70%-80% from

baseline level (P < .001) after which a plateau was reached during

the final minute of occlusion. The inter-subject variation of CHb at

the end of the venous occlusion period was 15% on both days. After

release of the cuff, CHb returned to baseline in a biphasic pattern.

Within 2 min, C_Hb rapidly decreased to a level 20% above baseline.

Then, CHb continued to slowly decrease toward baseline, although

there was still an elevation from baseline 10 min after release of the cuff.

There were no significant differences in C_Hb during baseline and

arterial occlusion between day 1 and day 2, but both the post-oc-clusive hyperemic response and the venous stasis were significantly larger on day 2 than on day 1 (P < .001). The correlation between

C_Hb on the first and the second day of the experiment was 0.94, and

the intraclass correlation was 0.90 (Figure 2). The correlation coeffi-cient was 0.55 for post-occlusive hyperemia alone (P = .1) and 0.78 (P = .008) for venous stasis. The intraclass correlation was 0.82 for post-occlusive hyperemia and 0.60 for venous stasis.

F I G U R E 2   Mean microvascular response to arterial occlusion and post-occlusive hyperemia (left), during venous stasis (middle) in 10 healthy subjects as measured using the smartphone app “SkinSight” on two different days. The right panel shows the correlation between measurements on day 1 and day 2 (r = 0.94; ICC = 0.90)

0 5 10 15 20 0.0 0.1 0.2 0.3 0.4 Time (min) CHb (AU) Day 1 Day 2 0 5 10 15 20 0.0 0.1 0.2 0.3 0.4 Time (min) CHb (AU) Day 1 Day 2 0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3 0.4 0.5 C_Hb(AU) - Day 1 CHb (AU) -Day 2

F I G U R E 3   Mean microvascular response to arterial occlusion and post-occlusive hyperemia (left), during venous stasis (middle) in 10 healthy subjects as measured using polarized light spectroscopy imaging on two different days. The right panel shows the correlation between measurements on day 1 and day 2 (r = 0.68; ICC = 0.71)

0 5 10 15 20 0 50 100 150 Time (min) CRB C (AU) Day 1 Day 2 0 5 10 15 20 0 50 100 150 Time (min) CRB C (AU) Day 1 Day 2 0 50 100 150 200 0 50 100 150 C_RBC (AU) - Day 1 CRBC (AU) - Day 2

3.2 | Tissue Viability Imaging (TiVi, C

)

In Figure 3, the change in C_RBC as measured using TiVi is shown, during

arterial occlusion and post-occlusive hyperemia (A) and during venous occlusion and subsequent recovery (B). At baseline, the inter-subject

variation of CRBC was 45% on day 1 and 43% on day 2. There was no

significant change in C_RBC during arterial occlusion, compared to

base-line (P = .99). After release of the cuff and reperfusion of the forearm,

C_RBC increased rapidly to 120% of baseline level (P < .001). At the peak,

the inter-subject variation was 35% on day 1 and 28% on day 2. After

reaching a peak, C_RBC gradually returned to baseline. During 5 min of

venous occlusion, CRBC gradually increased to about 65%-95% from

baseline level (P < .001) and no plateau was reached at the end of the period of occlusion. The inter-subject variation at the end of the venous

occlusion was 40% (day 1) and 28% (day 2). As with C_Hb, C_RBC returned

to baseline in a biphasic pattern after the cuff pressure was released. The microvascular response was significantly greater on day 1 compared to day 2, during all phases, including baseline, arterial oc-clusion, post-occlusive hyperemia, and venous stasis (P < .001). The

day-to-day reproducibility of the CRBC is depicted in Figure 3. The

overall correlation between C_RBC values on different days was 0.68,

and the intraclass correlation was 0.71. The correlation coefficient was 0.44 for post-occlusive hyperemia alone (P = .2) and 0.55 (P = .1) for venous stasis. The intraclass correlation was 0.46 for post-occlu-sive hyperemia and 0.51 for venous stasis.

The time-response dynamics of CHb measured using SkinSight

and C_RBC measured using TiVi were similar during both

post-occlu-sive hyperemia and venous stasis (Figure 4).

4 | DISCUSSION

The measurement of the skin microcirculation is of interest in many situations, both in the clinic and in the research. In many pathophysi-ologies, the functioning of the skin microcirculation is a measure of the viability of the tissue and of the probability of recovery. This is the case with skin burns, radiation dermatitis, microsurgical flaps,

chronic wounds, and venous insufficiency.3,4,9,14 _{In microvascular}

research, the skin is often used as a surrogate organ in which the system microvascular function can easily be assessed, and in fact,

changes in skin microcirculation are seen in a number of diseases that are associated with impaired systemic microvascular func-tion, including hypertension, diabetes, and other cardiovascular

diseases.7,15,16

Currently, assessment of the skin microcirculation is mainly done using noninvasive, laser-based techniques. One of the major techniques that has been in use since the 1980s is Laser Doppler flowmetry (LDF). LDF is based on illumination of the tissue by laser

light.17 _{When this laser light is scattered by moving red blood cells}

in the skin, its wavelength changes as a result of the Doppler effect. The scattered light is collected by a photodiode, which gives a signal that is further processed to give a measure proportional to the tissue blood flow. The laser Doppler technique has later been refined to overcome one of its shortcomings, that is, that it measures in a single point in the skin. Laser Doppler imaging and Laser Speckle Contrast imaging are similar to LDF but instead measure the skin blood flow

in an area of skin, typically up to 20 × 20 cm2_{. Besides laser-based}

techniques, other optical techniques are in use to measure the skin microcirculation, including optical coherence tomography, hyper-spectral imaging, and infrared thermography. All of these techniques are based on relatively expensive parts and require equipment that is not easily transported between patients. The data collected by the instruments are often not easily interpreted or translated into useful measures.

Tissue Viability imaging is a technique based on a commercial digital camera that overcomes many of the aforementioned

limita-tions of the common techniques for blood flow measurement.13 _{It is}

a hand-held system that is easy to use and enables quick measure-ments. However, the calculation of red blood cell concentration from the images still requires the images to be transferred to a computer for further analysis. This is a considerable drawback for the tech-nique to be used in a clinical setting.

The main result of this study is that it demonstrates that it is possible to accurately measure changes in the skin microcirculation using a mobile phone. This is interesting because it suggests that smartphone apps may replace the techniques that are currently used for skin microvascular assessments and offer a truly bedside assessment of microvascular function or tissue viability.

When comparing SkinSight with TiVi, we found that the change in hemoglobin index during arterial occlusion was somewhat lower

F I G U R E 4   Comparison in microvascular response in the skin to arterial occlusion (left) and venous occlusion (right) as measured with the smartphone app “SkinSight” (black lines) and polarized light spectroscopy imaging,

TiVi (blue lines) 0 5 10 15 20

–50 0 50 100 150 Time (min) CHb (AU ) SkinSight TiVi 0 5 10 15 20 –50 0 50 100 150 Time (min) CHb (AU ) SkinSight TiVi

than the change in red blood cell concentration. The measured changes in hemoglobin index and red blood cell concentration were similar between both techniques during venous occlusion. The in-ter-subject variability of SkinSight was, however, significantly lower than of TiVi, both during baseline, during post-occlusive hyperemia, and after 5 min of venous occlusion. Skin color is considered the largest source of variability between subjects with polarized light

spectroscopy imaging.13 _{The two techniques used in this study}

uti-lize different algorithms to correct for melanin in the skin, but both are based on the fact that melanin absorbs light more or less equally for the red and green channels. The variability between subjects of the smartphone-based hemoglobin measurement was much lower than for polarized light spectroscopy imaging, particularly for the peak post-occlusive hyperemia. This suggests that the SkinSight al-gorithm is less sensitive to variations in skin color.

There was a difference between the measured baseline levels of

C_Hb and C_RBC on different days. This day-to-day variation was largest

for TiVi, and interestingly, the differences were opposite for both

techniques. Mean C_Hb at baseline was 10% larger on day 2, while

mean CRBC was 32% smaller on day 2 of the experiment. The peak

C_Hb response after arterial occlusion was 14% higher on day 2, while

CRBC was 15% lower. These differences in results may be explained

by differences in background lighting between the two measure-ment days. Both techniques are based on illumination of the tissue by white light emitted by the light source of the system, either the smartphones integrated LED or the light-emitting LED ring on the TiVi system. Environmental background lighting, such as ceiling lights or sunlight, affects the measurement as they will change the apparent color of the skin during the measurement and thereby the calculation of the hemoglobin index and red blood cell concentra-tion. This problem could be mitigated by keeping the background lighting as low as possible, and by reducing the distance between the skin and the system's light source. However, this is not always feasible in clinical practice, and to obtain clinically relevant circum-stances, we did not attempt to dim or turn off all the background lighting during the experiments in this study. Despite the day-to-day variations, we found that the correlation between measurements made with SkinSight on day 1 and day 2 was very good (r = 0.94; ICC = 0.90).

SkinSight does not use any polarizing filters between the light source and the camera. Polarizing filters are used in other optical imaging techniques to suppress specular reflections from the skin surface, which do not contain any information about the hemoglobin content in the skin. During the development of SkinSight, we tested the system both with and without polarizing filters. Although the images acquired without polarizing filters do contain a specular com-ponent, we found that the algorithm used in SkinSight effectively suppresses this component and that the use of polarizers did not significantly change the measurement. It should be noted that this could be different when specular reflections make up a large part of the total signal, and diffusely reflected light from within the skin might become undetectable. In experiments like those performed in

this study, on dry and intact skin, specular reflections do however not seem to be problematic.

This study has limitations. Human skin is characterized by variable concentration in melanin, ranging from very low in light Caucasian skin, to very high in black African skin. As the subjects included in this study had a Caucasian skin, we do not know the sen-sitivity of the algorithm in darker skin types. Although the algorithm is tuned to specifically measure hemoglobin content, a high concen-tration of melanin may change the peneconcen-tration depth of the light and the calculations of hemoglobin concentration. Whether this can be adjusted for by including melanin as a parameter in the algorithm needs further attention.

In the current study, the image analysis was done on an external computer for practical reasons. However, the computational power of modern smartphones is sufficient, and it will be no problem to do the calculations directly on the device.

5 | CONCLUSIONS

Changes in the skin microcirculation during physiological provoca-tions can be reliably measured using a modern smartphone technol-ogy. This offers a handheld, easy to use, and flexible technique for skin microvascular assessment, although the effects of lighting on the measured values and need for calibration need further atten-tion. Smartphone apps have the potential to at least in some fields replace bulky and expensive techniques that are currently used for skin microvascular assessments and offer a truly bedside assessment of microvascular function or tissue viability.

CONFLIC TS OF INTEREST None declared.

ORCID

Erik Tesselaar https://orcid.org/0000-0002-8387-0583

REFERENCES

1. Roustit M, Cracowski JL. Assessment of endothelial and neurovas-cular function in human skin microcirculation. Trends Pharmacol Sci. 2013;34(7):373-384.

2. Droog EJ, Steenbergen W, Sjöberg F. Measurement of depth of burns by laser Doppler perfusion imaging. Burns. 2001;27(6):561-568. 3. Mirdell R, Iredahl F, Sjöberg F, Farnebo S, Tesselaar E. Microvascular

blood flow in scalds in children and its relation to duration of wound healing: a study using laser speckle contrast imaging. Burns. 2016;42(3):648-654.

4. Zötterman J, Bergkvist M, Iredahl F, Tesselaar E, Farnebo S. Monitoring of partial and full venous outflow obstruction in a por-cine flap model using laser speckle contrast imaging. J Plast Reconstr Aesthetic Surg. 2016;69(7):936-943.

5. Holm C, Mayr M, Höfter E, Ninkovic M. Perfusion zones of the DIEP flap revisited: a clinical study. Plast Reconstr Surg. 2006;117(1):37-43. 6. Iredahl F, Löfberg A, Sjöberg F, Farnebo S, Tesselaar E. Non-invasive

measurement of skin microvascular response during pharmacologi-cal and physiologipharmacologi-cal provocations. PLoS One. 2015;10(8):e0133760.

7. Czernichow S, Greenfield JR, Galan P, et al. Microvascular dysfunc-tion in healthy insulin-sensitive overweight individuals. J Hypertens. 2010;28(2):325-332.

8. Pettersson E, Anderson CD, Henricsson J, Falk M. Validation of phototesting for estimation of individual skin ultraviolet sensitiv-ity based on a lengthwise attenuating ultraviolet B field. J Med Eng Technol. 2015;39(2):91-98.

9. Tesselaar E, Flejmer AM, Farnebo S, Dasu A. Changes in skin micro-circulation during radiation therapy for breast cancer. Acta Oncol (Madr). 2017;56(8):1072-1080.

10. Muntean MV, Achimas-Cadariu PA. Detection of perforators for free flap planning using smartphone thermal imaging. Plast Reconstr Surg. 2018;142(4):604e.

11. Rueden CT, Schindelin J, Hiner MC, et al. Image J2: ImageJ for the next generation of scientific image data. BMC Bioinformatics. 2017;18(1). https://doi.org/10.1186/s1285 9-017-1934-z

12. Liu Z, Zerubia J. Melanin and hemoglobin identification for skin dis-ease analysis. In: Proceedings - 2nd IAPR Asian Conference on Pattern Recognition, ACPR;2013. https://doi.org/10.1109/ACPR.2013.9 13. O'Doherty J, Henricson J, Anderson C, Leahy MJ, Nilsson GE,

Sjöberg F. Sub-epidermal imaging using polarized light spectros-copy for assessment of skin microcirculation. Ski Res Technol. 2007;13(4):472-484.

14. Bergkvist M, Henricson J, Iredahl F, Tesselaar E, Sjöberg F, Farnebo S. Assessment of microcirculation of the skin using tissue viability imaging: a promising technique for detecting venous stasis in the skin. Microvasc Res. 2015;101:20-25. https://doi.org/10.1016/j. mvr.2015.06.002

15. Abularrage CJ, Sidawy AN, Aidinian G, Singh N, Weiswasser JM, Arora S. Evaluation of the microcirculation in vascular disease. J Vasc Surg. 2005;42(3):574-581.

16. Burnand K, Abisi S. Microcirculation in chronic venous insuffi-ciency. In: Venous Ulcers; 2007. https://doi.org/10.1016/B978-01237 3565-2.50006 -3

17. Eun HC. Evaluation of skin blood flow by laser Doppler flowmetry. Clin Dermatol. 1995;13(4):337-347.

How to cite this article: Tesselaar E, Farnebo S. Objective assessment of skin microcirculation using a smartphone

camera. Skin Res Technol. 2020;00:1–7. https://doi.