Concurrent validation and reliability of digital image analysis of granulation tissue color for clinical pressure ulcers

Development of a Method for Clinical Evaluation of Artificial Intelligence-Based Digital Wound Assessment Tools

IMPORTANCE

Accurate assessment of wound area and percentage of granulation tissue (PGT) are important for optimizing wound care and healing outcomes. Artificial intelligence (AI)-based wound assessment tools have the potential to improve the accuracy and consistency of wound area and PGT measurement, while improving efficiency of wound care workflows.

OBJECTIVE

To develop a quantitative and qualitative method to evaluate AI-based wound assessment tools compared with expert human assessments.

DESIGN, SETTING, AND PARTICIPANTS

This diagnostic study was performed across 2 independent wound centers using deidentified wound photographs collected for routine care (site 1, 110 photographs taken between May 1 and 31, 2018; site 2, 89 photographs taken between January 1 and December 31, 2019). Digital wound photographs of patients were selected chronologically from the electronic medical records from the general population of patients visiting the wound centers. For inclusion in the study, the complete wound edge and a ruler were required to be visible; circumferential ulcers were specifically excluded. Four wound specialists (2 per site) and an AI-based wound assessment service independently traced wound area and granulation tissue.

MAIN OUTCOMES AND MEASURES

The quantitative performance of AI tracings was evaluated by statistically comparing error measure distributions between test AI traces and reference human traces (AI vs human) with error distributions between independent traces by 2 humans (human vs human). Quantitative outcomes included statistically significant differences in error measures of false-negative area (FNA), false-positive area (FPA), and absolute relative error (ARE) between AI vs human and human vs human comparisons of wound area and granulation tissue tracings. Six masked attending physician reviewers (3 per site) viewed randomized area tracings for AI and human annotators and qualitatively assessed them. Qualitative outcomes included statistically significant difference in the absolute difference between AI-based PGT measurements and mean reviewer visual PGT estimates compared with PGT estimate variability measures (ie, range, standard deviation) across reviewers.

RESULTS

A total of 199 photographs were selected for the study across both sites; mean (SD) patient age was 64 (18) years (range, 17-95 years) and 127 (63.8%) were women. The comparisons of AI vs human with human vs human for FPA and ARE were not statistically significant. AI vs human FNA was slightly elevated compared with human vs human FNA (median [IQR], 7.7% [2.7%-21.2%] vs 5.7% [1.6%-14.9%]; P < .001), indicating that AI traces tended to slightly underestimate the human reference wound boundaries compared with human test traces. Two of 6 reviewers had a statistically higher frequency in agreement that human tracings met the standard area definition, but overall agreement was moderate (352 yes responses of 583 total responses [60.4%] for AI and 793 yes responses of 1166 total responses [68.0%] for human tracings). AI PGT measurements fell in the typical range of variation in interreviewer visual PGT estimates; however, visual PGT estimates varied considerably (mean range, 34.8%; mean SD, 19.6%).

CONCLUSIONS AND RELEVANCE

This study provides a framework for evaluating AI-based digital wound assessment tools that can be extended to automated measurements of other wound features or adapted to evaluate other AI-based digital image diagnostic tools. As AI-based wound assessment tools become more common across wound care settings, it will be important to rigorously validate their performance in helping clinicians obtain accurate wound assessments to guide clinical care.

Skin tear classification using machine learning from digital RGB image

AIM

Skin tears are traumatic wounds characterised by separation of the skin layers. Severity evaluation is important in the management of skin tears. To support the assessment and management of skin tears, this study aimed to develop an algorithm to estimate a category of the Skin Tear Audit Research classification system (STAR classification) using digital images via machine learning. This was achieved by introducing shape features representing complicated shape of the skin tears.

METHODS

A skin tear image was separated into small segments, and features of each segment were estimated. The segments were then classified into different classes by machine learning algorithms, namely support vector machine and random forest. Their performance in classifying wound segments and STAR categories was evaluated with 31 images using the leave-one-out cross validation.

RESULTS

Support vector machine showed an accuracy of 74% and 69% in classifying wound segments and STAR categories, respectively. The corresponding accuracy using random forest were 71% and 63%.

CONCLUSION

Machine learning algorithms revealed capable of classifying categories of skin tears. This could offer the potential to aid nurses in their management of skin tears, even if they are not specialised in wound care.

Skin colour assessment of replanted fingers in digital images and its reliability for the incorporation of images in nursing progress notes

AIMS AND OBJECTIVES

To determine whether nurses can accurately assess the skin colour of replanted fingers displayed as digital images on a computer screen.

BACKGROUND

Colour measurement and clinical diagnostic methods for medical digital images have been studied, but reproducing skin colour on a computer screen remains difficult.

DESIGN

The inter-rater reliability of skin colour assessment scores was evaluated. In May 2014, 21 nurses who worked on a trauma ward in Japan participated in testing.

METHODS

Six digital images with different skin colours were used. Colours were scored from both digital images and direct patient's observation. The score from a digital image was defined as the test score, and its difference from the direct assessment score as the difference score. Intraclass correlation coefficients were calculated. Nurses' opinions were classified and summarised.

RESULTS

The intraclass correlation coefficients for the test scores were fair. Although the intraclass correlation coefficients for the difference scores were poor, they improved to good when three images that might have contributed to poor reliability were excluded. Most nurses stated that it is difficult to assess skin colour in digital images; they did not think it could be a substitute for direct visual assessment. However, most nurses were in favour of including images in nursing progress notes.

DISCUSSION

Although the inter-rater reliability was fairly high, the reliability of colour reproduction in digital images as indicated by the difference scores was poor. Nevertheless, nurses expect the incorporation of digital images in nursing progress notes to be useful. This gap between the reliability of digital colour reproduction and nurses' expectations towards it must be addressed.

CONCLUSIONS

High inter-rater reliability for digital images in nursing progress notes was not observed. Assessments of future improvements in colour reproduction technologies are required.

RELEVANCE TO CLINICAL PRACTICE

Further digitisation and visualisation of nursing records might pose challenges.

Points to consider for skin ulcers in systemic sclerosis

This article discusses points to consider when undertaking a clinical trial to test therapy for skin ulcers in SSc. A validated definition of skin ulcers should be used if available. Defining a uniform SSc patient population, including consideration of disease duration, history of digital ulcers and capillaroscopic patterns, is important. Excluding confounding factors such as infection, calcinosis and trauma should be strongly considered, or at least accounted for, in defining patients. Outcome measures such as time to healing, prevention of new ulcers, function, pain and objective measures such as US, laser Doppler and thermography can be considered as outcome measures, although their validation has not yet been achieved and efforts may be needed to validate them before use. Likewise, biomarkers should be considered or consideration should be given to storing serum, plasma or cells for possible future analysis. A pre-planned analysis is important and should include consideration of missing data.

An improved automated type-based method for area assessment of wound surface

Accurate and precise wound measurements are a critical component of wound detection and assessment. Digital cameras are convenient and objective tools that are being increasingly used worldwide to assist with wound measurements and assessments. However, heterogeneous wounds and poor lighting conditions continue to be obstacles to wound area recognition. This study, therefore, provides an improved automated type-based wound area assessment method that is robust to lighting conditions and can distinguish between different wound tissues based on wound colors. The results of both laboratory and clinical applications of the proposed method show excellent consistency of manual area measurements. This proposed technology is expected to provide wound care specialists with more clinical information about heterogeneous wounds, thereby enabling prospective cost savings for therapy and treatment.

Reliability of tristimulus colourimetry in the assessment of cutaneous bruise colour

BACKGROUND

Bruising is one of the most common types of injury clinicians observe among victims of violence and other trauma patients. However, research has shown commonly used qualitative description of cutaneous bruise colour via the naked eye is subjective and unreliable. No published work has formally evaluated the reliability of tristimulus colourimetry as an alternative for assessing bruise colour, despite its clinical and research applications in accurately assessing skin colour. The purpose of this study was to systematically evaluate the test-retest and inter-observer reliability of tristimulus colourimetry in the assessment of cutaneous bruise colour.

METHODS

Two researchers obtained repeated tristimulus colourimetry measures of cutaneous bruises with participants of diverse skin colour. Measures were obtained using the Minolta CR-400 Chomameter. Commission Internationale d'Eclairage (CIE) L*a*b* colour space was used. Data was analysed using intraclass correlation coefficients (ICC), Cronbach's alpha, and minimal detectable change (MDC) on all three L*a*b* values.

RESULTS

The colorimeter demonstrated excellent test-retest or intra-rater reliability (L* ICC=0.999; a* ICC=0.973; b* ICC=0.892) and inter-rater reliability (L* ICC=0.997; a* ICC=0.976; b* ICC=0.982).

CONCLUSIONS

With consistent placement, the tristimulus colourimetry is reliable for the objective assessment and documentation of cutaneous bruise colour for purposes of clinical practice and research. Recommendations for use in practice/research are provided.

Nutrition-related status and granulation tissue colour of pressure ulcers evaluated by digital image analysis in older patients

OBJECTIVE

Granulation tissue colour may be an indicator for nutritional assessment in pressure ulcer (PU) care. This study evaluated the relationship between nutritional status, anaemia and diabetes, and granulation tissue colour of PUs by colour analysis of digital photographs in the clinical setting.

METHOD

The cross-sectional study included 42 older patients with 51 full-thickness PUs from 10 institutions. Patient demographics, wound status, nutritional status and dietary intakes were obtained from medical charts. From a wound image, the granulation red index was processed by computer software and the proportion of pixels exceeding the threshold intensity of 80 for the granulation tissue surface (%GRI80) was calculated.

RESULTS

Haemoglobin levels were positively associated with %GRI80 levels (p=0.007) in the crude model, but not in the adjusted model (p=0.260). The interaction term between diabetes and protein intake was significantly associated with %GRI80 levels in the adjusted models (p=0.010). At protein intakes of 0.95 g/kg or higher, diabetic wounds exhibited lower %GRI80 levels than non-diabetic wounds (p=0.002). At protein intakes of less than 0.95 g/kg, %GRI80 levels did not differ between diabetic and non-diabetic patients (p=0.247). Protein intakes of 0.95 g/kg or higher were associated with higher %GRI80 levels in non-diabetic patients (p=0.015), but not in diabetic patients (p=0.127).

CONCLUSION

Granulation tissue colour, evaluated by the objective and quantitative analysis of digital photography, is related to haemoglobin level, diabetes and dietary intakes in clinical settings.