ADVERTISEMENT
Is Artificial Intelligence the Future of Wound Care?
The term “artificial intelligence” (AI) was first used in a proposal for a conference at Dartmouth College in 1955, and AI applications were commonly seen in health care in the early 1970s.1 There is wide-reaching potential for AI to transform health care delivery by introducing new and innovative ways to assess patients, streamline care, simplify administrative processes, and collect and store data. Medical evaluations that previously required human interaction may in the future be accomplished by leveraging artificial intelligence. Machine learning (ML) is a type of artificial intelligence that utilizes various algorithms as a form of programming or “learning.” The ability to adjust to experiences and inputs is what allows ML to perform human-like tasks.
Global markets have seen the widespread adoption of AI in the arenas of finance and information technology, but its adoption in health care has lagged. Ethical and safety considerations are likely the reason as health care professionals are more cautious to adopt new technologies when human lives are at stake.2 Despite this slower integration, industry experts predict that widespread utilization of AI will improve the quality of care and health outcomes for patients by decreasing human errors.3 Additionally, clinicians will potentially become free from routine and repetitive tasks allowing more time to be focused on complex responsibilities.3 Still, human factors, such as trust, perceived usefulness, and privacy, play an important role in the acceptance and adoption of new technologies such as AI in health care.4
There is, however, a growing need for personalized medicine, rising demand for value-based care, growing datasets of patient health-related digital information, and advancements in health care IT. Pervasiveness of smartphone apps, widespread internet connectivity, and a shortage of health care providers are propelling the acceptance of AI amongst patients and health care workers alike. The artificial intelligence in health care market is estimated to generate $6.9 billion in 2021 and reach $67.4 billion by 2027; it is projected to grow at a compound annual growth rate of 46.2% during 2021–2027.5 The primary factors driving the demand for artificial intelligence in the health care sector include the increasing concern towards cost optimization of health care and management of the data, the increasing number of alliances between public and private organizations, and the accelerated regional budget towards the health care sector.5
Can these technological advances be harnessed to better predict healing outcomes, monitor wound progression, and measure responses to treatments? Innovative AI prediction models are in development helping to identify wound patterns not obvious or observable to even the most trained clinicians.
How We Manage Wounds Today
It is estimated that even with specialized wound care in hospital-based outpatient wound centers, more than one-third of patients may never have resolution of their wounds.6 A complete patient history and physical (H&P) is the cornerstone of effective wound management. The H&P enables the clinician to identify any underlying disease states such as diabetes, autoimmune disorders, vascular impairment, inflammatory conditions, anemia, kidney disease, history of radiation therapy, and malignancy, which are known factors contributing to wound chronicity.7 Testing, lab work, and specialist referrals may be recommended as part of this process.
Additionally, a detailed wound assessment is performed to collect wound-specific parameters to facilitate the creation of an appropriate evidence-based treatment plan optimizing the wound environment to aid in wound healing.
To date, there is still a paucity of basic science and clinical research aimed at developing clinical practice algorithms to quickly and easily identify wounds that have or may become chronic. Current best practices in wound management include ongoing patient and wound assessments to facilitate optimal wound healing. There is a need for better predictive models utilizing more precise and individualized wound assessment tools. The cost of implementing new technology paired with patient complexity and variability have been the primary barriers to wide adoption of new point-of-care diagnostic devices. Can AI fill this void? As technology evolves, it seems that AI is poised to be a tool of great value in the wound care space.
Historically, wound assessments, biological indicators and biomarkers have been used to identify and stratify subsets of healing and nonhealing wounds of varying etiologies. Examination of tissue samples revealed the first molecular markers correlating with healing interruptions. Certain biologic indicators such as matrix metalloproteinase (MMP) levels and pathogenic microbiota have been studied to evaluate their predictive effect of wound healing. MMPs are critically important enzymes found in every phase of wound healing.8
The key to MMP utility in wound healing is having the right amount at the right location for the right duration of time. Elevated levels of MMPs are often found in wounded tissues of prolonged duration. Unregulated levels of MMPs lead to disorganization and degradation of the extracellular matrix, thus contributing to the chronicity of nonhealing wounds.8 Currently, there is not a rapid, point-of-care test available to allow clinicians to obtain information on the type and level of MMPs present in the wound tissue, so therefore this biomarker is currently not widely utilized as a predictor of wound trajectory.
Human skin contains a bionomic mix of symbiotic, mutualistic, and pathogenic microorganisms known as the microbiome. When injury to the skin occurs, there is a break into the natural barrier leading to microbial migration into the wounded tissues. The complexity of chronic wound microbiota has led to the belief that colonization is most commonly polymicrobial, consisting of both planktonic and those present in the biofilm construct. Persistent wound biofilm has been associated with poor healing outcomes and prolongation of the inflammatory cycle.9 Although contaminated with bioburden, chronic wounds may not show clinical signs and symptoms of infection. It is often clinically difficult to easily identify if pathologic levels of problematic bacteria are present in the wound. Historically culture or tissue biopsy is needed to identify various microbes, but these tests may underrepresent the numbers of many bacterial strains.10 Microbes that may persist under various testing conditions do not always reflect the true diversity of the chronic wound microbiome and therefore do not truly illustrate the entirety of the microbiota footprint. Therefore, information gained from routine cultures may not represent the best predictors of clinical outcomes of wound healing.11
The predictive value of wound area measurements has been recognized as an important tool in wound management for over a century.12 Wound area is a very common reported data point in wound care clinical trials. Many practitioners still rely on simple wound measurements of length, width, and depth obtained manually with paper rulers. The long-standing belief that these measurements adequately and consistently represent wound area has been called into question. In fact, one published study concluded that standard, manual (l x w) measurement of cutaneous wounds inaccurately overestimated wound area by roughly 40%.13 For wound measurements to be valid predictors of wound progress, this information must be accurate, standardized, and reproducible. Technology does exist by way of planimetric imaging devices that promise to increase accuracy and precision of wound area. Unfortunately, due to cost and accessibility, these devices are not widely utilized by most health care providers. Admittedly, although not always accurate, serial wound measurements are currently the best predictor of wound progress available to most wound care clinicians.
A published, validated predictive model of wound healing based on clinical information including wound age, wound size, overall number of active wounds present, presence of infection, patient age, wound grade, renal impairment, and peripheral vascular disease does exist.14 The Wound Healing Index (WHI) was established to assist in clinical practice, research analysis and clinical trials to help determine healing potential in patients with diabetic foot ulcers (DFUs).14 However, the index’s creators admit that clinical data input into electronic medical records can be lacking, and important information used to calculate wound trajectories via the WHI may be missing or inaccurate. Additionally, simple scoring tools to assimilate this data into clinical practice are not readily available. Therefore, this tool has not gained wide utilization by health care practitioners. Reflecting on what current tools and assessments are at hand, many wound care clinicians are asking—is there a better way?
How Can AI Help Us Manage Wounds Better?
Deep learning-based image analysis software does exist to extract relevant wound information such as the location of interest, wound only image crops, and wound periphery size-over-time metrics.15 As previously discussed, measuring wound size changes over time allows clinicians to gain valuable insights such as wound rate of closure. Historically wound care clinicians have devoted time to collecting clinical biomarker and biologic data such as manual wound measurements as previously described to identify disruptions in healing patterns.
ML wound image analysis takes these assessments to the next level. By analyzing wound photos and comparing them to numerous similar cases in established databanks, AI can identify trends sooner and alert clinicians of abnormalities. Innovations such as Microsoft’s Computer Vision Inner EYE initiative are being used by clinicians to diagnose and treat tumors.16
How can clinicians leverage AI to better manage the patients we care for? Let’s discuss some use cases. Identifying wound etiology is key for selecting the most appropriate therapies and paramount for assuring accurate documentation. Misdiagnosis exposes patients to substantial risks associated with wound chronicity. When wound etiology is misclassified inappropriate treatment protocols follow and the potential to apply the wrong treatment can prolong healing time.17
Leveraging ML to analyze wound photos has the potential to optimize healing time and patient outcome. Concurrently the utilization of resources is decreased and expense to the payer is lowered. Recognizing when wound healing has stalled or when wounds are degrading allows clinicians to reassess the need for additional testing or advanced intervention is an important part of successful wound management.18 Using ML to better identify complicated wounds early so clinicians can order tests or advocate for advanced interventions could increase patient outcomes and lower health care costs. Data gained from ML can be used to support medical necessity and provide evidence for prior authorization of treatment protocols.
Chronic wound patients have a high recidivism and readmission rate. ML can leverage patient health information to correlate with health outcomes before they occur—identifying underlying patterns of disease.19 If clinicians can leverage technology to prevent re-ulceration and readmission this would translate into cost saving, which is especially important in value-based networks. There are additional opportunities in wearable devices with noninvasive biosensors offering sophisticated health measurements and monitoring various wound parameters. There is the potential to increase patient engagement and adherence with wound care protocols.
Additionally, all too often palliative care patients are identified late in patients’ journey leading to low optimization of resources and decreased quality of life.20 Harnessing the power of ML to identify potential non-healing and palliative care patients earlier to redirect to appropriate treatment protocols will support appropriate utilization of care, optimize resources (clinicians’ time, wound supplies), improve quality of life, and prevent infection/hospitalization.
In Conclusion
AI applications in health care, and in wound management particularly, have the potential to improve diagnosis and treatment. Although the clinical field is desperate for better diagnostic tools, their development is slow. For widespread adoption of AI to occur there needs to be a perceived usefulness over and above the current standard practices.
A comprehensive, unbiased method to evaluate wound trajectory is needed across patient care settings. With the power to reduce subjectivity and variability in clinical diagnosis, AI has shown promise in the wound management space. Early indications seem to illustrate that leveraging AI solutions results in significant advances in data collection to improve patient care. Identifying patients in need of advanced therapeutics early in the treatment course will allow clinicians to better tailor treatment algorithms and support better patient outcomes, additionally saving health care expenditures. AI can also provide early recognition of patients who would both fail standard of care therapy before starting treatment and would benefit from early therapeutic interventions tailored to their specific healing prognosis, ultimately resulting in improved and more cost-effective outcomes. The transition to point-of-care clinically applicable technology will necessitate a focus on simplicity, economy, and consistent implementation.
It is anticipated that AI can be utilized in a variety of ways that extend beyond wound healing into other diagnostic areas within the wound care space. Shifting the focus of wound care from a reactive to a preventative model by suggesting personalized treatment plans will represent a welcome sea change. Next phase research is still needed to analyze the effectiveness of various AI to determine utility. AI is an essential part of the paradigm shift from curative to preventive medicine. AI holds a strong transformative potential to enhance sustainable health care in the wearable and smart dressing market by empowering self-care, self-monitoring, and self-diagnosis. The on-demand availability, potential to improve efficiency, and reduce the cost of health care service delivery are several of the most enticing features of AI.
Windy Cole, DPM, has practiced in Northeast Ohio for over 22 years. She is an Adjunct Professor and Director of Wound Care Research at Kent State University College of Podiatric Medicine. She is board certified by the American Board of Foot and Ankle Surgery and the American Board of Wound Management. Cole is a member of the American College of Clinical Wound Specialists Board of Directors. Additionally, she holds multiple advisory and editorial positions with various medical and wound care publications, and sits on the advisory board of multiple emerging biotech companies and has been integral in collaborating on innovative research protocols in the space.
References
1. Guo J, Li B. The application of medical artificial intelligence technology in rural areas of developing countries. Health Equity. 2018 Aug 1;2(1):174-181
2. Arora A. Conceptualising artificial intelligence as a digital healthcare innovation: an introductory review. Med Devices (Auckl). 2020 Aug 20;13:223–30
3. Topol EJ. High-performance medicine: the convergence of human and artificial intelligence. Nat Med. 2019 Jan;25(1):44–56.
4. Lu Y, Xiao Y, Sears A, Jacko JA. A review and a framework of handheld computer adoption in healthcare. Int J Med Inform. 2005 Jun;74(5):409–22
5. https://www.marketsandmarkets.com/Market-Reports/artificial-intelligence-healthcare-market
6. Fife CE, Buyukcakir C, Otto G, Sheffield P, Love T, Warriner R 3rd. Factors influencing the outcome of lower-extremity diabetic ulcers treated with hyperbaric oxygen therapy. Wound Repair Regen. 2007 May-Jun;15(3):322-31.
7. Ackermann PW, Hart DA. Influence of comorbidities: neuropathy, vasculopathy, and diabetes on healing response quality. Adv Wound Care (New Rochelle). 2013 Oct;2(8):410-421
8. Michopoulou A, Rousselle P. How do epidermal matrix metalloproteinases support re-epithelialization during skin healing? Eur J Dermatol. 2015 Apr;25(Suppl 1):33–42.
9. Frank DN, Wysocki A, Specht-Glick DD, et al. Microbial diversity in chronic open wounds. Wound Rep Regen. 2009 Mar-Apr;17(2):163–172.
10. Bowler PG, Duerden BI, Armstrong DG. Wound microbiology and associated approaches to wound management. Clin Microbiol Rev. 2001 Apr;14(2):244–269.
11. Gardner SE, Hillis SL, Heilmann K, Segre JA, Grice EA. The neuropathic diabetic foot ulcer microbiome is associated with clinical factors. Diabetes. 2013 Mar;62(3):923–930.
12. Allen CW. Measurements of skin lesions. J Cutan Genitourin Urin Dis. 1897:15:438.
13. Rogers LC, Bevilacqua NJ, Armstrong DG, Andros G. Digital planimetry results in more accurate wound measurements: a comparison to standard ruler measurements. J Diabetes Sci Technol. 2010 Jul 1;4(4):799-802.
14. Fife CE, Horn SD, Smout RJ, Barrett RS, Thomson B. A predictive model for diabetic foot ulcer outcome: The Wound Healing Index. Adv Wound Care (New Rochelle). 2016 Jul 1;5(7):279-287. doi: 10.1089/wound.2015.0668. PMID: 27366589; PMCID: PMC4900227.
15. Carrión H, Jafari M, Bagood MD, Yang HY, Isseroff RR, Gomez M. Automatic wound detection and size estimation using deep learning algorithms. PLoS Comput Biol. 2022 Mar 11;18(3)
16. Microsoft. https://www.microsoft.com/en-us/research/project/medical-image-analysis/news-and-awards/
17. Hess CT. Misdiagnosis of wounds. Adv Skin Wound Care. 2019; 32(3):144.
18. Chen WY, Rogers AA. Recent insights into the causes of chronic leg ulceration in venous diseases and implications on other types of chronic wounds. Wound Repair Regen. 2007;15(4):434–449
19. IBM. https://www.ibm.com/watson-health
20. Cole W. Palliative wound care: closing the gap. Podiatry Today. 2022; 35(4):44-47.
