Background: Composition of tissue types within a wound is a useful indicator of its healing progression. Tissue composition is clinically used in wound healing tools (eg, Bates-Jensen Wound Assessment Tool) to assess risk and recommend treatment. However, wound tissue identification and the estimation of their relative composition is highly subjective. Consequently, incorrect assessments could be reported, leading to downstream impacts including inappropriate dressing selection, failure to identify wounds at risk of not healing, or failure to make appropriate referrals to specialists.Objective: This study aimed to measure inter- and intrarater variability in manual tissue segmentation and quantification among a cohort of wound care clinicians and determine if an objective assessment of tissue types (ie, size and amount) can be achieved using deep neural networks.Methods: A data set of 58 anonymized wound images of various types of chronic wounds from Swift Medical’s Wound Database was used to conduct the inter- and intrarater agreement study. The data set was split into 3 subsets with 50% overlap between subsets to measure intrarater agreement. In this study, 4 different tissue types (epithelial, granulation, slough, and eschar) within the wound bed were independently labeled by the 5 wound clinicians at 1-week intervals using a browser-based image annotation tool. In addition, 2 deep convolutional neural network architectures were developed for wound segmentation and tissue segmentation and were used in sequence in the workflow. These models were trained using 465,187 and 17,000 image-label pairs, respectively. This is the largest and most diverse reported data set used for training deep learning models for wound and wound tissue segmentation. The resulting models offer robust performance in diverse imaging conditions, are unbiased toward skin tones, and could execute in near real time on mobile devices.Results: A poor to moderate interrater agreement in identifying tissue types in chronic wound images was reported. A very poor Krippendorff α value of .014 for interrater variability when identifying epithelization was observed, whereas granulation was most consistently identified by the clinicians. The intrarater intraclass correlation (3,1), however, indicates that raters were relatively consistent when labeling the same image multiple times over a period. Our deep learning models achieved a mean intersection over union of 0.8644 and 0.7192 for wound and tissue segmentation, respectively. A cohort of wound clinicians, by consensus, rated 91% (53/58) of the tissue segmentation results to be between fair and good in terms of tissue identification and segmentation quality.Conclusions: The interrater agreement study validates that clinicians exhibit considerable variability when identifying and visually estimating wound tissue proportion. The proposed deep learning technique provides objective tissue identification and measurements to assist clinicians in documenting the wound more accurately and could have a significant impact on wound care when deployed at scale.

Chronic wounds affect millions of people worldwide every year. An adequate assessment of a wound's prognosis is a critical aspect of wound care since it assists clinicians in understanding wound healing status, severity, triaging and determining the efficacy of a treatment regimen, thus guiding the clinical decision making. The current standard of care involves using wound assessment tools, such as Pressure Ulcer Scale for Healing (PUSH) and Bates-Jensen Wound Assessment Tool (BWAT), to determine wound prognosis. However, these tools involve manual assessment of a multitude of wound characteristics and skilled consideration of a variety of factors, thus, making wound prognosis a slow process which is prone to misinterpretation and high degree of variability. Therefore, in this work we have explored the viability of replacing subjective clinical information with deep learning-based objective features derived from wound images, pertaining to wound area and tissue amounts. These objective features were used to train prognostic models, that quantified the risk of delayed wound healing, using a dataset consisting of 2.1 million wound evaluations derived from more than 200,000 wounds. The objective model, which was trained exclusively using image-based objective features, achieved at minimum a 5% and 9% improvement over PUSH and BWAT, respectively. Our best performing model, that used both subjective and objective features, achieved at minimum an 8% and 13% improvement over PUSH and BWAT, respectively. Moreover, the reported models consistently outperformed the standard tools across various clinical settings, wound etiologies, sexes, age groups and wound ages, thus establishing the generalizability of the models.

Objectives: This time-motion study explored the amount of time clinicians spent on wound assessments in a real-world environment using wound assessment digital application utilizing Artificial Intelligence (AI) vs. manual methods. The study also aimed at comparing the proportion of captured quality wound images on the first attempt by the assessment method. Methods: Clinicians practicing at Valley Wound Center who agreed to join the study were asked to record the time needed to complete wound assessment activities for patients with active wounds referred for a routine evaluation on the follow-up days at the clinic. Assessment activities included: labelling wounds, capturing images, measuring wounds, calculating surface areas, and transferring data into the patient’s record. Results: A total of 91 patients with 115 wounds were assessed. The average time to capture and access wound image with the AI digital tool was significantly faster than a standard digital camera with an average of 62 seconds (P<0.001). The digital application was significantly faster by 77% at accurately measuring and calculating the wound surface area with an average of 45.05 seconds (P<0.001). Overall, the average time to complete a wound assessment using Swift was significantly faster by 79%. Using the AI application, the staff completed all steps in about half of the time (54%) normally spent on manual wound evaluation activities. Moreover, acquiring acceptable wound image was significantly more likely to be achieved the first time using the digital tool than the manual methods (92.2% vs. 75.7%, P<0.004). Conclusions: Using the digital assessment tool saved significant time for clinicians in assessing wounds. It also successfully captured quality wound images at the first attempt.

A 57-Year-Old Man with Type 1 Diabetes Mellitus and a Chronic Foot Ulcer Successfully Managed with a Remote Patient-Facing Wound Care Smartphone Application - Article abstract #933879

Introduction: Clinical signs and symptoms (CSS) of infection are a standard part of wound care, yet they can have low specificity and sensitivity, which can further vary due to clinician knowledge, experience, and education. Wound photography is becoming more widely adopted to support wound care. Thermography has been studied in the medical literature to assess signs of perfusion and inflammation for decades. Bacterial fluorescence has recently emerged as a valuable tool to detect a high bacterial load within wounds. Combining these modalities offers a potential objective screening tool for wound infection.Methods: A multi-center prospective study of 66 outpatient wound care patients used hyperspectral imaging to collect visible light, thermography, and bacterial fluorescence images. Wounds were assessed and screened using the International Wound Infection Institute (IWII) checklist for CSS of infection. Principal component analysis was performed on the images to identify wounds presenting as infected, inflamed, or non-infected.Results: The model could accurately predict all three wound classes (infected, inflamed, and non-infected) with an accuracy of 74%. They performed best on infected wounds (100% sensitivity and 91% specificity) compared to non-inflamed (sensitivity 94%, specificity 70%) and inflamed wounds (85% sensitivity, 77% specificity).Discussion: Combining multiple imaging modalities enables the application of models to improve wound assessment. Infection detection by CSS is vulnerable to subjective interpretation and variability based on clinicians' education and skills. Enabling clinicians to use point-of-care hyperspectral imaging may allow earlier infection detection and intervention, possibly preventing delays in wound healing and minimizing adverse events.

Summary: Objective: Less than 20% of hospitals in the US have an electronic health record (EHR). In this qualitative study, we examine the perspectives of both academic and private physicians and administrators as stakeholders, and their alignment, to explore their perspectives on the use of technology in the clinical environment.Methods: Focus groups were conducted with 74 participants who were asked a series of open-ended questions. Grounded theory was used to analyze the transcribed data and build convergent themes. The relevance and importance of themes was constructed by examining frequency, convergence, and intensity. A model was proposed that represents the interactions between themes. Results: Six major themes emerged, which include the impact of EHR systems on workflow, patient care, communication, research/outcomes/billing, education/learning, and institutional culture. Academic and private physicians were confident of the future benefits of EHR systems, yet cautious about the current implementations of EHR, and its impact on interactions with other members of the healthcare team and with patients, and the amount of time necessary to use EHR’s. Private physicians differed on education and were uneasy about the steep learning curve necessary for use of new systems. In contrast to physicians, university and hospital administrators are optimistic, and value the availability of data for use in reporting.Conclusion: The results of our study indicate that both private and academic physicians concur on the need for features that maintain and enhance the relationship with the patient and the healthcare team. Resistance to adoption is related to insufficient functionality and its potential negative impact on patient care. Integration of data collection into clinical workflows must consider the unexpected costs of data acquisition.