The Reliability and Validity of Color Indicators Using Digital Image Analysis of Peristomal Skin Photographs: Results of a Preliminary Prospective Clinical Study

Accurate assessment is necessary to evaluate peristomal skin condition, but objective methods are lacking. The purpose of this prospective, repeated-measures study was to evaluate the reliability and validity of color indicators using digital image analysis of peristomal skin photographs. The 6-month study was conducted among 21 patients (mean age 65.1 years old, 15 men) with ostomies (14 colostomies, six ileostomies, and one urostomy) at four outpatient clinics. Photographs taken by nurses of the peristomal area using point-and-shoot cameras were processed using digital image analysis, which involved color calibration, image processing, and indicator calculation. An erythema index (EI), melanin index (MI), and hypopigmentation index were created to represent increased degrees of red, black, and white color, respectively, and their average values in the peristomal region of an image were calculated relative to values for intact skin. Reproducibility was evaluated using the interclass correlation coefficient (ICC). ICCs of color indicators for intact skin were >0.7 between baseline and the end of follow-up for the 16 participants with two or more clinic visits. Differences in these indices between peristomal and intact regions were evaluated using a linear mixed model. The EI and MI of peristomal skin were significantly higher than those of intact skin (n = 42, P <0.001). All color indicators in adjacent regions and areas where adhesive was applied were associated with the discoloration severity score and visual analogue pain score (all P <0.05). This objective and simple method had adequate reproducibility and criterion-related validity and may be useful for peristomal skin assessment. Further research is warranted.

Peristomal skin frequently is exposed to various negative stimuli, including mechanical stress secondary to repeated removal of adhesives, irritants associated with leakage of feces and urine, allergens in stoma care products, moisture from increased sweating under adhesives, and an increased biological burden from micro-organisms or fungi.¹ These factors interdependently increase the risk of peristomal skin complications such as erosion, moisture-associated contact irritation, allergic dermatitis, folliculitis, and infection.^1,2 Such complications are observed in 13% to 60% of people with a stoma depending on the stoma type and time since surgery.^3-6 Peristomal skin complications often result in physical symptoms such as pain and itching, compromised pouch attachment, increased medical costs, and decreased quality of life for patients.^6-9 Therefore, prevention, early detection, and early treatment of peristomal skin complications are crucial issues in stoma care.

Accurate assessment of the peristomal skin condition is indispensable in selecting the most appropriate treatments among the various options available in stoma care.² In addition, clinicians should monitor skin conditions to determine the effects of the care provided. For these objectives, a standardized indicator to evaluate, diagnose, and classify peristomal skin is required.¹⁰

Color is regarded as an important indicator in peristomal skin assessment, because peristomal skin complications are frequently accompanied by color changes such as erythema or hyperpigmentation.^1,2,11 Clinicians assess color in the peristomal region to determine the severity of skin complications, evaluate the extent of discoloration and the areas involved, and monitor improvements as abnormal color disappears. Although several standardized tools such as the Ostomy Skin Tool and Studio Alterazioni Cutanee Stomali instrument include items related to discoloration,^12,13 these evaluations depend solely on subjective classification of discoloration, possibly resulting in inadequate inter-rater agreement.¹⁴ Furthermore, these tools may have insensitive responsiveness in the monitoring of longitudinal peristomal skin changes such as decreased erythema and increased hyperpigmentation. An objective method would contribute to more accurate, precise, and continuous assessment of peristomal skin discoloration.

Digital image analysis recently emerged in dermatology and wound care as a novel method to evaluate color objectively.^15-17 This method is also available for peristomal assessment, because peristomal photography is a common protocol in stoma care for accurate documentation of the stoma condition.¹³ Digital image analysis of color has several advantages in stoma care: it uses a noncontact device, objective and quantitative measures, and two-dimensional assessment, all of which facilitate evaluation of both the severity and area of discoloration. Therefore, digital image analysis may be more suitable than reflectance spectrophotometers and tristimulus colorimeters that have been generally used to evaluate a limited area of skin or wound.¹⁸

Considering the increased interest in wound care in home care settings using telemedicine or assessment via the internet, photograph and image analysis of peristomal skin will become essential techniques.¹⁹ In such situations, nurses in outpatient clinics must depend on visual assessment because they cannot directly assess peristomal burning, swelling, or abnormal smells, which usually are detected by palpation or olfaction. It is crucial that visual assessment maintains accuracy across descriptions and personnel. Color indicators in digital image analysis of peristomal skin may be one solution to increase inter- and intra-rater reliability. The reliability and validity of this method have not been studied.

It was hypothesized that erythema, hyperpigmentation, and hypopigmentation are key parameters in the assessment of skin complications and that the degree of peristomal skin color can be evaluated using digital image analysis.¹¹ The purpose of this study was to: 1) develop color indicators using digital image analysis of photographs of peristomal skin, and 2) evaluate the reliability, criterion-related validity, and responsiveness — ie, confirmation of the change of indicators according to longitudinal color changes — in the clinical setting.

Study design. A prospective, repeated-measures study was conducted between October 2011 and March 2012. Patients with a stoma who visited outpatient clinics for regular follow-up were recruited from four institutions. The outpatient clinics were recruited by snowball sampling (additional populations — ie, clinics — were introduced by those already included in the study) and comprised one university hospital and three community hospitals. Patients with 1) any type of stoma, including colostomy, ileostomy, and urostomy; and 2) any purpose for and type of stoma construction, including permanent or temporary stomas, were eligible. Patients who did not provide consent to participate were excluded. Patients were repeatedly evaluated based on their individual regular care plan during the study period; therefore, the interval of visits varied (range 1 week to 3 months). All repeated data for a given patient at any time point were analyzed to evaluate the reproducibility and the responsiveness of image analysis to clinical changes in the peristomal skin condition. The study protocol was approved by the Ethics Committee of the Graduate School of Medicine, The University of Tokyo (#2634) and each clinic. Written informed consent was obtained from all patients.

Data collection. Patient demographics including age, gender, stoma type, underlying disease associated with stoma construction, and type of pouching system used were collected from the medical charts. In order to evaluate validity, visual peristomal skin condition and self-perception by patients including pain and itching were selected as parameters. For visual peristomal skin assessment, wound, ostomy, and continence (WOC) nurses at each institution evaluated the Discoloration, Erosion, and Tissue overgrowth (DET) score of the Ostomy Skin Tool.^12,14 This tool has been translated into Japanese, and the inter- and intrarater reliability among WOC nurses were confirmed for the Japanese version.²⁰ The severity score and the area score for discoloration, erosion, and tissue overgrowth were evaluated, and the total score (range 0–15, where 0 = none and 15 = most severe) then was calculated. In addition, patients were asked to express the degree of pain and itching of peristomal skin on a 0- to 100-mm visual analogue scale (VAS) (0 mm = none, 100 mm = most severe).

Each WOC nurse photographed the peristomal skin using a digital point-and-shoot camera. Patients sat on a chair for stoma care, and the skin/stoma photograph was taken after removal of the pouching system and following skin cleansing. The nurses followed the written manual of photography and received explanations and training from the primary author before the study. Skin was photographed approximately 10 minutes after pouch removal and cleansing.

The camera types were not strictly standardized in the data collection phase; WOC nurses used their own cameras for daily practice. Color settings were standardized by a color-calibration process (description to follow) during the image-processing phase so the difference in camera types was not anticipated to influence the result. A commercially available reference color chart with nine calibrated colors (Casmatch; Bear Medic Co, Tokyo, Japan) was placed on the surrounding skin. Photographs were taken to include the intact surrounding skin, such as that on the opposite side of the stoma. A flash was not used except in an environment that would result in extreme underexposure. The photographs were stored in JPEG format.

The photographs were processed by digital image analysis, which comprised color calibration, image processing, and indicator calculation.15 Color calibration was conducted according to the standard protocol of the color chart by image-editing software (Photoshop 6; Adobe Systems Inc, San Jose, CA, USA). The detailed protocol is described elsewhere.¹⁵ Three images reflecting different colors were obtained from a calibrated photograph using ImageJ software (National Institutes of Health, Bethesda, MD, USA); these images were the erythema index (EI), melanin index (MI), and hypopigmentation index (HI) (see Figure 1). It was hypothesized that higher EI, MI, and HI values represent increased degrees of erythema, hyperpigmentation, and hypopigmentation, respectively. The EI image was obtained from the formula logR – logG by the image calculator tool, indicating the result of subtraction of the R and G channels of the digital color images after they were logarithmically transformed.^15,16 The MI image was obtained from the inverted image of the logarithm of the R channels. The HI image was obtained by transforming the original RGB image into a 16-bit gray image that represented the degree of brightness. The image intensity was arbitrarily calibrated to standardize the intensity of the reference color (red for EI, black for MI, and white for HI) at 200. The intensity of each image ranged from 0 to 255, and a higher intensity indicated a brighter color.

To calculate the color indicators, a researcher manually selected the peristomal skin region including the stoma, the stoma region itself, and the intact skin region using a freehand tool. The inter- and intra-rater reliability of this technique have been reported previously.¹⁵ The peristomal skin region excluding the stoma — ie, the region covered by the pouch — was obtained by subtracting the stoma region from the peristomal skin region including the stoma. The peristomal skin region then was separated into the regions adjacent to the stoma and under the adhesive, because skin conditions supposedly differ between peristomal regions. The border of these regions was determined by the original image and adhesive sizes. The intact skin was determined as that on the opposite side of the stoma or that uncovered by adhesives without an operative incision on the abdomen. A histogram of the selected region was obtained. Because whiteout and blackout areas will exist in an image as noise caused by elements such as light reflectance or shadow, the intensities 0–5 and 251–255 were deleted for the purpose of noise reduction. The mean intensity of each color in the region of interest was calculated as an indicator. All indicators in the peristomal regions were expressed as relative values for those on intact skin, because relative values may be helpful in assessing the effect of skin complications independent from a patient’s inherent skin type. The representative values of each indicator for a specific patient are shown in Figure 2. In this patient, both the EI and MI ratios were increased in the region adjacent to the stoma, while only the MI ratio was high in the region under the adhesives with a slightly increased EI ratio secondary to erosion. These features corresponded to the visual features.

Statistical analysis. Descriptive data at baseline were expressed as mean (SD) for continuous variables and n (%) for categorical variables. The reproducibility or test-retest reliability of the color indicators in each patient was evaluated by an intraclass correlation coefficient (ICC) on intact skin between baseline and the end of follow-up, with the assumption that intact skin color will undergo little change. The ICC was interpreted according to common criteria; values of 0.21 to 0.40 indicate a “fair” correlation, 0.41 to 0.60 a “moderate” correlation, 0.61 to 0.80 a “substantial” correlation, and >0.81 an “almost perfect” correlation.²¹ To show the descriptive statistics of the color indicators, the least square means were calculated by a linear mixed model that included all time periods of evaluation for the same patients as a repeated variable (SAS PROC MIXED). Because the color indicators were expressed as relative values and were skewed, their logarithm values were used to make the distribution similar to the normal distribution. The same analytical procedure was applied to evaluate associations between the color indicators and DET and VAS scores, indicating criterion-related validity, and to assess the responsiveness of the color indicators against changes in the DET scores. All analyses were conducted using SAS version 9.3 (SAS Institute Inc, Cary, NC, USA). The statistical significance level was set at a P value of <0.05.

A total of 21 patients with a stoma was enrolled during the study period. The mean age (SD) was 65.1 (14.1) years, and 15 patients (71.4%) were male (see Table 1). One patient used a one-piece and all others used a two-piece pouching system. The median (range) duration since stoma construction was 14 (range 0–111) months. Sixteen (16) patients (76.2%) were evaluated twice or more. The most frequent stoma type was a colostomy (71.4%). The mean (SD) of the total DET score was 4.1 (2.9).    The reproducibility of the color indicators on intact skin was evaluated for the 16 patients who were evaluated twice or more (see Table 2). All ICCs for the EI, MI, and HI were >0.7, indicating substantial or almost perfect agreement between baseline and the end of follow-up.²⁰ In total, 42 data parameters, including repeated data parameters in a patient, were available during the study period. As descriptive statistics of peristomal skin color (see Figure 3), the EI and MI ratios were significantly higher than 1 (P <0.001), indicating a higher EI and MI in the peristomal region, the region adjacent to the stoma, and the adhesive region than in intact skin. However, the HI ratio was significantly lower than 1 in each region, respectively (All P values <0.001). The EI and MI were significantly higher in the region adjacent to the stoma than in the region under adhesives within a given patient (P <0.001).

For assessment of criterion-related validity, associations between color indicators and DET scores were evaluated (see Table 3). All color indicators in each region were significantly associated with the DET discoloration severity score and total score (all P <0.05). The EI and MI ratios showed positive relationships and the HI ratio showed a negative relationship, indicating that clinically more severe peristomal skin discoloration was accompanied by increased erythema and hyperpigmentation. For assessment of the responsiveness of the color indicators, the data at baseline and at the end of the study duration were analyzed in the 16 patients who were evaluated twice or more (see Table 4). Changes in some color indicators in each region tended to be associated with changes in the DET discoloration severity score and total score, showing the color indicators changed according to the clinical condition. Not all associations reached statistical significance, possibly because of the lack of statistical power.

For assessment of pain and itching, 25 data parameters were available in total during the study duration. All color indicators in all regions were significantly associated with the pain score (all P <0.05) (see Table 5). The EI and MI ratios showed positive relationships and the HI ratio showed a negative relationship, indicating increased erythema or hyperpigmentation in patients who complained of a greater degree of pain in the peristomal region. However, no color indicators were significantly associated with the itching score.   Longitudinal changes in the peristomal skin condition were monitored by the color indicators; a typical case is shown in Figure 4. The patient was an 84-year-old woman who had undergone surgical construction of a descending colostomy 2 months previously. At week 1 in the outpatient clinic, the peristomal skin in the region adjacent to the stoma showed increased EI and MI ratios; whereas, the skin under the adhesive showed slight pigmentation with an increased MI ratio. At week 2, irritant contact dermatitis, possibly caused by mechanical stress, was apparent in the entire peristomal skin region with painful erythema, leading to a rapid increase in the EI ratio in both regions; the MI ratio remained constant. At this time, the types of adhesive and intervals between pouch changes were altered, and the use of moisturizing lotion was begun. After 1 week, the contact dermatitis was improved with color indicator values similar to those of the first week.

To the authors’ knowledge, this is the first study involving objective digital image analysis of color for the evaluation of peristomal skin conditions in the clinical setting. This method had adequate reproducibility for intact skin color and criterion-related validity of scores for a clinical assessment tool and patient’s perceived pain. In addition, responsiveness of the color change over time was partly confirmed. The color indicators (EI, MI, and HI) showed adequate features of reliability and validity in this study.

This image analysis using the EI, MI, and HI has several advantages. First, this is a noninvasive and quantitative method that uses routinely available photographs. Although visual assessment by clinicians is the most important method in clinical practice to assess peristomal skin complications and select appropriate care,² objective tools to support visual assessment are lacking. This simple method can be useful to quantitatively monitor changes in peristomal skin conditions. Second, three colors (red, black, and white) can be evaluated separately. A previous report¹¹ retrospectively investigated peristomal skin conditions for patients using skin-barrier products for 1 to 10 years. Results showed inflammatory changes gradually disappeared and pigmentary changes, including hyper- and hypopigmentation, became prominent at later periods over 5 years. Using multiple colors in combination may help evaluate complex skin discoloration changes. Third, this method can analyze the peristomal skin as a two-dimensional image. In dermatology, reflectance spectrophotometers and tristimulus colorimeters are widely used to measure skin color.²² However, such devices can only measure limited, narrow regions and may not be applicable to the evaluation of peristomal skin because the entire affected area and anatomical location provide other important information for peristomal assessment. Image analysis can resolve this limitation by measuring an anatomical location or area of peristomal skin.^15-17

Color indicators on a patient’s intact skin showed substantial or almost perfect reproducibility. This result indicates the image processing can standardize color circumstances for a patient even at different time points. In the clinical setting, various luminance conditions derived from the patient’s positioning and body shape, the location of the room or bed, and weather compromised the visual color assessment. Although standardized settings are required to improve the reproducibility of peristomal assessment, not all of them are adjustable in the clinical setting. Even under such conditions, the method described in this study has the advantage of improved reliability in the phase of color calibration using a standardized color chart, enabling the clinician to longitudinally monitor the skin color change. Results showed the ICC of the EI was relatively lower than those of the MI and HI. This is expected because the errors after color calibration would be greater for red than for other colors.²³ Nevertheless, this point may not be problematic, because color indicators in the peristomal region were calculated as relative to intact skin. This additional process can further reduce color indicator errors among images.

The study showed peristomal skin color is significantly different from that of intact skin unaffected by adhesives, irrespective of the presence of clinically problematic complications. This trend was particularly prominent in the region adjacent to the stoma. This result indicates a pouching system with adhesives more or less affects the peristomal skin color, including erythema and pigmentation.¹¹ Various factors including mechanical stress, chemical irritants, and ongoing occlusion may lead to altered peristomal skin barrier function.²⁴ This alteration can appear as clinically visible discoloration, although it does not necessarily represent a pathological condition. Digital image analysis may be helpful for early detection of such peristomal skin changes. Further studies using this digital image analysis will be required to quantify and understand the objective longitudinal discoloration changes after stoma construction and the detailed relationships with potential factors. The color change cutoff points to predict clinically problematic skin complications are also important to determine.

The significant correlation between the severity of discoloration score and color indicators in this study indicated the image analysis had criterion-related validity for visual color classification. Although visual assessment of peristomal skin color is recommended in clinical practice, it only can evaluate the presence or absence of skin complications. Even a DET discoloration score includes only three categories: none, slight, or severe.¹² In addition, the strict definition of color, such as “deep red,” is complicated and requires assessment expertise. This new method has the advantage of evaluating skin color quantitatively and continuously. Furthermore, results showed the color indicators in the region adjacent to the stoma had relatively stronger associations with the discoloration score than with those in the adhesive region. This result is in agreement with clinical experience showing peristomal skin complications, especially moisture-associated skin damage, frequently occur in the region adjacent to stoma at high risk for direct contact with feces or urine.⁸ In addition, responsiveness to skin color changes was partly confirmed. These results can support the construct validity for color indicators. These color indicators enable clinicians to perform accurate and sophisticated assessment of peristomal color conditions.

Color indicators were associated with the degree of pain. In particular, the association between the EI and pain is of importance, because the EI reflects the degree of erythema. It is hypothesized that peristomal pain is strongly associated with erythema because of acute inflammation under conditions of irritant dermatitis or infection.⁷ The results support the criterion-related validity of digital image analysis. The reason why the MI and HI were associated with pain remains unclear, because hyper- and hypopigmentation may develop after acute inflammation has disappeared; a possible explanation may be that these observations reflect postinflammatory pigmentation secondary to repeated mechanical stress and perceived pain during adhesive changes.

Color indicators did not show an association with the itching score. In fact, studies of factors and assessment methods for peristomal itching are lacking, even though itching is a frequent complaint among people with a stoma.²⁵ Further studies are required to reveal the relationships between peristomal skin conditions and itching.

Although the photograph method has clinically adequate validity and reliability, study limitations exist. Selection bias may have been present because the study population was Japanese with a particular skin type that is easily pigmented. Although this population characteristic improved the feasibility of color assessment of peristomal skin in this study, the results should be cautiously interpreted in terms of whether this method will apply to other populations, especially those with dark skin. The second limitation is the low statistical power due to the small sample size. One reason for this small sample is the different intervals of visiting outpatient clinics; some patients visited only once during the study period. Another reason is the exclusion of several photographs in which the color was difficult to accurately evaluate, such as pictures without the opposite side being visible or those affected by shade. To disseminate this method into broader settings, increased feasibility will be required in the form of easier-to-operate photography systems and supportive settings, such as optimal room illumination. Finally, the color indicators were evaluated over peristomal skin by a radial method, whereas the DET score only evaluates the proportion of the discolored area irrespective of location.¹² Another popular tool divides the peristomal region into four sites.¹³ If skin problems exist only in a part of a peristomal region but do not cover the entire region, this method may underestimate the severity because color indicators around the peristomal region are averaged. Further study will be required to compare the usefulness of various anatomical assessment techniques. Nevertheless, this method has an advantage even with respect to this point, because various indicators, such as area and morphology, can be created from image analysis using color.

This is the first study to apply digital image analysis to peristomal skin color assessment in the clinical setting. All color indicators, including the EI, MI, and HI, had good to almost perfect reproducibility in terms of the intact skin color. In addition, criterion-related validity was evident in both the region adjacent to the stoma and under the adhesive for the clinical discoloration scores and patients’ perceived pain scores when color indicators were used in the form of relative values for the intact skin. Responsiveness to color changes over time was partly confirmed.

This novel, objective method can support clinicians in their decision-making when they evaluate peristomal skin complications, select appropriate care options, and monitor the effect of care. It also should contribute to improvement of the quality of stoma care and the assessment skill of specialized nurses in daily practice.

 Potential Conflicts of Interest: This study was funded by a grant from the Japanese Society of Stoma and Continence Rehabilitation, 2011. The funding organization had no role in the design, data collection, analysis, review, or approval of the manuscript.

Dr. Iizaka and Ms. Asada are nurses, Department of Gerontological Nursing/Wound Care Management, Division of Health Sciences and Nursing, Graduate School of Medicine, The University of Tokyo, Tokyo, Japan. Ms. Koyanagi and Ms. Sasaki are wound, ostomy, and continence nurses, The University of Tokyo Hospital, Tokyo, Japan. Ms. Naito is a wound, ostomy, and continence nurse, Fujisawa City Hospital, Kanagawa, Japan. Dr. Konya is a professor and a wound, ostomy, and continence nurse, School of Nursing, Kanazawa Medical University, Ishikawa, Japan. Dr. Sanada is a professor and a wound, ostomy, and continence nurse, Department of Gerontological Nursing/Wound Care Management, Division of Health Sciences and Nursing, Graduate School of Medicine, The University of Tokyo.

Please address correspondence to: Shinji Iizaka, PhD, Department of Gerontological Nursing/Wound Care Management, Division of Health Sciences and Nursing, Graduate School of Medicine, The University of Tokyo Hospital, Faculty of Medicine Building No. 5-304, Bunkyo-ku, Tokyo 113-0033 email: iizaka-tky@umin.ac.jp.