Objectively quantifying facial erythema in rosacea aided by the ImageJ analysis of VISIA red images

Abstract

Background

Facial erythema, a prominent characteristic of rosacea, causes concern to both the patient and doctor. In clinical practice, commonly used erythema severity subjective assessment tools lack objectivity and are less comprehensive. Even with images taken by the VISIA^® system, diffused erythema is difficult to segment and evaluate fully due to the automatic threshold segmentation method. This study aimed to explore a more objective and scientific erythema quantification tool with the aid of the ImageJ software analysis of the red area images taken by the VISIA^® system.

Materials and methods

Patients with rosacea were enrolled and assessed for the clinical severity of their illness using various stools—the standard grading systems (SGS) for rosacea, investigator's global assessment (IGA), and clinician's erythema assessment (CEA). Facial images in the red area mode of the VISIA^® system were further analyzed by the ImageJ for the relative intensity of redness and percentage of erythema area; the correlation with the scores of the subjective grading systems was evaluated.

Results

This study included 201 patients (195 females and 6 males). The relative intensity of redness was positively correlated to the SGS, IGA, and CEA scores (0.688, 0.725, and 0.718, respectively) (p < 0.001). The percentage of erythema area was positively correlated to the SGS, IGA, and CEA scores (0.615, 0.666, and 0.656, respectively) (p < 0.001).

Conclusion

We demonstrated a more objective and precise method of assessing the severity of facial erythema rosacea, which could comprehensively assess the severity by both the area and intensity of facial erythema.

1. INTRODUCTION

Rosacea is a common chronic inflammatory skin disease of unknown etiology, ¹ with a reported incidence ranging from 0.09% to 22% and an average of 5.46%. ² Facial erythema is the most common and vital manifestation of rosacea. ³ It is considered to be caused by skin inflammation, vasodilation, and changes in vascular structure which result in the abnormal increase of hemoglobin in the papillary dermis. ⁴ Rosacea is a chronic relapsing skin condition that requires long‐term management. Therefore, it is crucial to assess the severity of facial erythema comprehensively and precisely.

The standard grading system for rosacea (SGS), ⁵ clinician erythema assessment (CEA), ⁶ patient's self‐assessment (PSA), ⁷ and investigator's global assessment (IGA) ⁸ are currently widely used in dermatological practice. However, their intrinsic deficiencies include lack of objectivity and accuracy, ⁹ existing inter‐ and intraobserver differences, ⁶ and difficulty in precise quantification. ⁴ , ¹⁰ Therefore, more objective assessments without these deficiencies are required to evaluate the severity of facial erythema in rosacea accurately and consistently. ¹¹

In some dermatological clinical trials, some devices are used to detect and quantify erythema, including the chromameter, polarized light spectroscopy, laser speckle contrast imaging, and computer‐assisted imaging system such as the VISIA^® Complexion Analysis System. Among these, the VISIA^® system is more commonly used in clinical practice. It collects images by standard white light, ultraviolet light, and polarized lights and then analyzes the photos using a threshold approach to quantify the number and severity of skin lesions. The red area image measures the skin's redness according to the hemoglobin distribution, showing erythema, dilated capillaries, and vascular lesions. However, in clinical practice, erythema cannot be fully segmented or recognized by the VISIA^® system, especially when presented in a diffuse or gradient manner, making the data imprecise ⁴ (Figure 1).

FIGURE 1.

Blue area: erythema area segmented in the VISIA^® system. Green arrow: diffuse or gradient erythema is not thoroughly segmented in the VISIA^® system red image. Purple arrow: erythema is not automatically demarcated in the red image by the VISIA^® system.

In this study, we aimed to find a more precise and comprehensive quantitative assessment tool for rosacea erythema using the ImageJ^® software (http://imagej.nih.gov/ij) based on the red area images taken by the VISIA^® system. This software further explores the feasibility of replacing the subjective classical grading systems of rosacea.

2. MATERIAL AND METHODS

2.1. Study participants

We enrolled 201 patients with rosacea with Fitzpatrick skin type III to IV in this study. They were recruited between July 2020 and February 2022 at the Department of Dermatology, the First Affiliated Hospital of Nanjing Medical University. The Institutional Research Committee on ethics of this facility approved the study (No. 2021‐SR‐326). All patients provided written informed consent. The key inclusion criterion was a clinical diagnosis of rosacea by two independent senior physicians according to the guidelines for rosacea. ¹² The medical history and clinical characteristics of all patients were recorded. Facial images were taken using the VISIA^® 6.0 complexion analysis system (Canfield, NJ, USA), and the red area images were further analyzed.

2.2. Subjective assessment of the severity of rosacea

The severity of rosacea was scored by two blinded senior dermatologists, according to SGS, IGA, and CEA. The grading systems used were provided in the Supplementary Materials.

2.3. Objective quantification of erythema

Two parameters—the relative intensity of redness (RIR) and percentage of erythema area (PEA)—were used to quantify the severity of erythema in the red area images taken by the VISIA^® systems.

2.3.1. The relative intensity of redness

A digital image is composed of pixels and colors determined by color spaces; the RGB color space is one of the most commonly used. ⁴ RGB color space consists of three color channels—red (R), green (G), and blue (B)—each of which has a separate histogram showing the number of pixels at each specific density value. On the histograms’ X‐axis, the leftmost and rightmost dot represents pure black, numbered 0 and 255, respectively, and between represents gray (numbered between 0 and 255). ¹³ An average value representing the average intensity value for that channel can be recorded by computing the pixel distribution for each channel. ⁴ Therefore, the RIR value of the red channel can be employed to evaluate the severity of facial erythema.

The ImageJ software was used to analyze the red area images recorded by the VISIA^® system. The “Analyze ‐ Histogram‐RGB” plug‐in in ImageJ was used to obtain the mean values of the R, G, and B color channels. Finally, as reported by Xu et al., ¹⁰ Formula (1) is used to calculate RIR values.

$$\mathrm{RIR}=\frac{\mathrm{Red}\mathrm{channel}\mathrm{mean}}{\mathrm{Green}\mathrm{channel}\mathrm{mean}\mathrm{and}\mathrm{Blue}\mathrm{channel}\mathrm{mean}}$$ (1)

2.3.2. Percentage of erythema area

Facial erythema is caused by skin inflammation, vasodilation, and changes in vascular structure, resulting in the abnormal increase of hemoglobin in the papillary dermis. This abnormal increase manifests as deeper erythematous areas relative to other parts of the skin on red area images taken by the VISIA^® systems. We used the image processing and threshold control—which can separate the high background (or normal) level of the erythema area—methods in ImageJ.

In addition to the RGB, the CIE Lab color space is another widely used color space. The erythema region is separated with a = 165 as the threshold in the CIE Lab color space mode. According to Nischik et al.’s inflammation severity method, erythema's severity is reflected by the pixel number of the different areas. ¹⁴ The larger percentage of the affected facial area indicates more severe erythema.

Red area images recorded by the VISIA^® system were analyzed by ImageJ software. The “image‐adjust‐color balance” in ImageJ was used to select “Color Space” as lab mode. We can segregate erythema by setting the threshold to a = 165. The lip area was excluded because of vermilion. Hence, PEA was calculated by the formula (2). The step‐by‐step protocol for erythema quantification using ImageJ, used in this study, is shown in Table 1.

$$PEA=\frac{erythemaarea-liparea}{facearea}$$ (2)

TABLE 1.

The step‐by‐step protocol for erythema quantification using the ImageJ

Drag the VISIA Red image in ImageJ
RIR 1. Select the face area: Image‐Adjust‐Color Thresholding‐ Color Space as “Lab” mode‐Set a* = 0‐Select (photo 1) 2. Measure RIR: Analyze‐Histogram‐RGB. Obtain R, G, and B channel mean values and calculate RIR through Formula 1
PEA 1.Measure face area: Analyze‐tools‐ROI Manager‐Add‐Measure‐Results save 2.Select the erythema area: Image‐Adjust‐Color Thresholding‐Color Space as “Lab” “mode‐Set a*= 165‐Select” (photo 2) 3. Measure the erythema area: Analyze‐tools‐ROI Manager‐Add‐Measure‐Results save 4. Select the lip area: double‐click “”—Tolerance = 30‐OK—Outline lip (photo 3) 5. Measure the lip area: Analyze‐tools‐ROI Manager‐Add‐Measure‐Results save 6. Measure PEA: calculate PEA through Formula 2

Note. The middle bold font in the table is the operation key provided by Image J software.

2.4. Statistical analysis

Statistical analysis was conducted using the SPSS^® 20.0 software (SPSS Inc, Chicago, IL, USA). The relationship between RIR and PEA with subjective grading systems was analyzed by the Spearman's rank correlation. Statistical significance was set as p < 0.05.

3. RESULTS

A total of 201 patients (195 females and 6 males) were enrolled in the study. The average age was 29.43 (ranging from 17 to 68). Fitzpatrick skin type is II for 92 patients, III for 98 patients, and IV for 11 patients. One hundred and seventy‐two (85.6%) and 85 (14.4%) patients were diagnosed with erythematotelangiectatic rosacea and papulopustular rosacea, respectively. The demographic characteristics of rosacea patients are listed in Table 2. The scores of SGS, IGA, and CEA for all rosacea patients are shown in Table 3.

TABLE 2.

Demographic characteristics of rosacea patients

	Gender (%)		Fitzpatrick skin type (%)			Rosacea subtype (%)
Age (years)	F	M	II	III	IV	ETR	PPR
29.43 ± 8.142	195 (97.0)	6 (3.0)	92 (45.8)	98 (48.8)	11 (5.5)	172 (85.6)	29 (14.4)

SGS, standard grading system for rosacea; F, female; M, male.

TABLE 3.

Subjective assessment of the severity of rosacea by different grading systems

CEA	n (%)	IGA	n (%)	SGS	n (%)	n (%)
0	2 (1.0)	0	2 (1.0)	Grading	ETR	PPR
1	53 (26.4)	1	53 (26.4)	Absent	3 (1.5)	0 (0.0)
2	76 (37.8)	2	74 (36.8)	Mild	115 (66.9)	9 (31.0)
3	64 (31.8)	3	66 (32.8)	Moderate	53 (30.8)	16 (55.2)
4	6 (3.0)	4	6 (3.0)	Severe	1 (0.6)	4 (13.8)

SGS, standard grading system for rosacea; IGA, investigator's global assessment; CEA, clinician erythema assessment; ETR, erythematotelangiectatic rosacea; PPR, papulopustular rosacea.

RIR values of 201 patients ranged from 0.67 to 1.28, with an average of 0.81. The correlation of RIR to the SGS, IGA, and CEA scores is shown in Figure 2. Comparing RIR to SGS, IGA, and CEA grading by senior dermatologists, the correlation coefficient was 0.688, 0.725, and 0.718, respectively (p < 0.001).

FIGURE 2.

Spearman's correlation analysis of the conventional assessment scores vs. the relative intensity of redness. (A) The standard grading system (SGS) for rosacea scores vs. the relative intensity of redness (RIR). (B) Investigator's global assessment scores (IGA) vs. the relative intensity of redness (RIR). (C) Clinician erythema assessment scores (IGA) vs. the relative intensity of redness (RIR).

PEA in 201 patients ranged from 0.22% to 86.28%, with an average of 28.64%. The correlation of PEA to the SGS, IGA, and CEA scores is shown in Figure 3. The correlation coefficient between the PEA and SGS, IGA, and CEA grading scores was 0.615, 0.666, and 0.656, respectively (p < 0.001).

FIGURE 3.

Spearman's correlation analysis of the conventional assessment scores vs. the relative intensity of redness. (A) The standard grading system for rosacea (SGS) scores vs. the percent erythema area (PEA). (B) The investigator's global assessment scores (IGA) vs. the percent erythema area (PEA). (C) The clinician erythema assessment scores (CEA) vs. percent erythema area (PEA).

The correlation coefficients between RIR, PEA, and subjective grading systems are shown in Table 4.

TABLE 4.

Correlation coefficients for SGS, IGA, and CEA of RIR and PEA

		Spearman's rank correlation coefficient (RIR)	Spearman's rank correlation coefficient (PEA)	p Value
SGS	ETR	0.691	0.594	<0.001
	PPR	0.586	0.591	<0.001
	Total	0.688	0.615	<0.001
IGA		0.725	0.666	<0.001
CEA		0.718	0.656	<0.001

RIR, the relative intensity of redness; PEA, percent erythema area; SGS, standard grading system for rosacea; IGA, investigator's global assessment; CEA, clinician erythema assessment; ETR, erythematotelangiectatic rosacea; PPR, papulopustular rosacea.

4. DISCUSSION

Erythema is one of the principal clinical features of rosacea. To better manage rosacea, it is important to evaluate clinical severity comprehensively and scientifically.

The more commonly used subjective grading systems used in clinical practice are the SGS and CEA. The SGS ⁵ grades the primary and secondary characteristics of rosacea into a standard classification system that includes the patients’ evaluation of the disease; this makes the scoring process complicated. However, some researchers reported that the classification of patients’ characteristics in SGS does not correspond well to the specific symptoms of rosacea, especially in dark‐skinned patients. ⁹

CEA is considered to be a reliable scale for assessing facial erythema in patients with rosacea ⁶ ; it focuses more on erythema intensity and not the area of the erythema. ¹⁵ Other scales, such as the flushing assessment tool (FAST) ¹⁶ and the global flushing severity scale (GFSS) ¹⁷ are self‐assessment tools for facial flushing (including skin redness, heat, tingling, and/or itching) in a 24 h period; also, FAST records the patient's facial flushing over a week. This makes FAST and GFSS assessment scales subjective and complicated to use in practice.

All the scores of these scales depend on the visual assessment of the patient's skin color, which cannot quantify the mild differences in erythema. ¹⁸ Color is a subjective and nonlinear sensory perception because the eye's sensitivity to visible light depends on the colors’ wavelength, which varies within the individual. ¹⁹ Furthermore, the subjective scales are not uniform and standardized and are therefore more dependent on the clinical experience of the dermatologist. Additionally, skin color is dependent on the mixture of “tanning” by epidermal melanin and “redness” of the skin due to blood flow; it is difficult to differentiate these components by visual examination alone. ²⁰ Therefore, objective methods are needed to quantify the skin erythema of rosacea.

Some devices are used to detect and quantify erythema, including chromameter, the VISIA^® system high‐frequency ultrasound, and laser speckle contrast imaging. ²¹ , ²² , ²³ , ²⁴ The VISIA^® system is used for the objective evaluation of rosacea erythema. It provides data on feature counts, absolute scores, and percentiles for red area images. The feature count is the total number of discrete instances of erythema, regardless of their size or severity. The absolute score provides an overall assessment of the erythema, including number, size, area, and intensity. However, the data provided do not accurately represent the actual situation of patients with facial erythema ²² since the VISIA^® system cannot fully segment or recognize erythema, especially when presented in a diffuse or gradient manner. This makes data imprecise.

In our study, the high‐resolution images captured by the VISIA^® system were further optimized for segmentation and calculation as RIR and PEA, making the evaluation of the overall severity of facial erythema more objective and comprehensive.

Xu et al. ⁴ validated the feasibility of using RIR values in quantifying facial skin erythema using a simulated erythema gradient. However, RIR only describes the intensity of erythema in the face and cannot provide an overall assessment of the erythema area. Meanwhile, they only studied simulated images of the erythema gradients and did not validate images from clinical patients. In a study by Kim et al., ¹⁰ PEA was used to evaluate the erythema area of patients with rosacea. The RIR and PEA are both related to the severity of erythema in rosacea patients.

Logger et al. ¹⁹ used the CIE Lab a* value to quantify facial erythema in white light images shot with a digital camera by ImageJ. However, the correlation coefficient between CIE Lab a* of facial erythema and CEA score was only 0.3 in their study, which did not show a strong correlation between the a* value and CEA score.

We are the first to describe that values of RIR and PEA calculated by ImageJ from the red area images of the VISIA^® system correlated well with the scores of SGS, IGA, and CEA, suggesting that they can be utilized in monitoring facial erythema in rosacea objectively and quantitatively since the process incorporated both the intensity and areas of erythema.

The limitation of our study was in the relative complexity of using ImageJ. It would be preferable if there was a more convenient analysis software that is available, simple to use, and free at all times.

5. CONCLUSIONS

RIR and PEA values calculated from the images taken by the VISIA^® system favorably correlated well with subjective grading systems—SGS, IGA, and CEA—suggesting a more objective and accurate method for evaluating the severity of facial erythema rosacea in clinical practice.

CONFLICT OF INTEREST

None declared.

ETHICAL STATEMENT

This study received approval from the local ethical committees.

Supporting information

Supporting Information

Tao M, Li M, Zhang Y, et al. Objectively quantifying facial erythema in rosacea aided by the ImageJ analysis of VISIA red images. Skin Res Technol. 2023;29:1–7. 10.1111/srt.13241

DATA AVAILABILITY STATEMENT

The data sets generated during and/or analyzed during the study are available from the corresponding author upon reasonable request.