AI in Wound Care: Shaping the Present and Future

Artificial intelligence, or AI for short, has been implemented into many facets of healthcare. From aiding with disease diagnosis and designing individualized treatment plans to analyzing large data sets of medical information. But in practice, what does AI in wound care look like? 

In the not-so-distant past, wound care professionals had to rely primarily on manual visual assessments to gather information on wounds. Today, AI wound assessments can be done using smartphones and more to diagnose, assess, and even treat injuries. We’ll dive into the evolution of AI wound healing developments and how they can be used to improve care outcomes. 

AI Is Reshaping How We Approach Wound Assessments

One area where artificial intelligence has been incredibly beneficial is wound assessments. These assessments are important metrics for tracking wound surface area changes over multiple weeks so practitioners can determine whether the therapy is effective. Clinical guidelines encourage regular documentation of key details, such as wound size and the rate of healing. However, this procedure is time-intensive, laborious, and imprecise without AI. 

Traditionally, wound care specialists used a ruler to assess the length and width of a wound. While easy to do, it’s not necessarily reliable because taking measurements in this manner is challenging. Following that, they’d place a cotton swab into the deepest part of the injury to determine its depth. Yet, this, too, could lead to inaccuracies when approximating the wound area. As you might imagine, these approaches also carry a risk of infection. 

AI Wound Assessment Innovations

Computer processing advancements in the last decade have enabled the integration of AI systems into different medical imaging technologies. For instance, they’ve been implemented into magnetic resonance imaging (MRI), ultrasound, X-ray, and even computed tomography (CT) scans. 

There are different forms of AI, but the ones primarily used in assessing medical images are deep learning and machine learning. So far, research has shown that they offer great reliability and accuracy. But before we provide some examples of that, let’s clarify what these AI terms mean.

Machine learning: A branch of AI that allows a system to autonomously learn and improve when provided with large amounts of data.

Deep learning: A subset of machine learning that shows computers how to process data similar to the human brain. These models can realize complex pictures, data patterns, and more to provide insights and predictions.  

Deep-Learning Image Analysis Pipeline

One of the first examples of AI in wound care we’ll examine is the deep learning-based image analysis pipeline. This tool was created to process non-uniform wound images to identify their type and size. It can remove important data, such as the wound location, conduct image cropping, and even calculate metrics regarding the size of the wound perimeter over time. 

Researchers found that this pipeline was effective, required very little human intervention, and could roughly calculate wound sizes even when up to 50% of the reference images weren’t available. 

Employing a Fully Automated Wound Segmentation Tool

A cohort study created and evaluated a deep learning network for fully automated objective tissue identification and measurement using a smartphone. The 2022 study comprised a dataset of 58 unidentified wound images of various forms of chronic wounds. In addition, five practitioners took the time to label four different tissue types using a browser-based image annotation tool.

Two deep convolutional neural network (CNN) architectures were made, and they were trained using over 450,000 and 17,000 image-label pairs, respectively. To date, this is the largest and most diverse dataset reported for training deep-learning AI models for wound and wound tissue segmentation. 

Upon study completion, they assessed that the algorithm did offer objective wound tissue measurement and identification. This helped wound care specialists document the wound more precisely while also being easy to use, cost-effective, and transportable. 

AI App for Wound Assessment and Management

A recent study looked to assess the effectiveness and utility of a machine learning AI application. The study design involved utilizing retrospective data obtained from patients evaluated using the AI app and later compared to individuals who were assessed using traditional approaches. 

The app could analyze images using machine learning algorithms and computer vision methods. It could identify wound elements such as depth, color, type, and whether granulation or necrotic tissue was present. The app then gives suggestions on wound management strategies, such as what topicals and wound dressings could be used or whether a referral to a specialist was needed. 

At the conclusion of the study, the authors discovered the following: 

• The app demonstrated better accuracy and consistency of wound assessment than the traditional methods
• AI app use decreased the need for face-to-face consultations
• The AI app was user-friendly and was accepted by providers and patients 

AI in Skin and Wound Classification 

Wound assessments are only half the battle. Another critical piece is proper wound classification. NIH explained that correctly classified wounds “help predict the likelihood of surgical site infections, postoperative complications, and reoperation.”

Few studies have been done on AI in wound classification, but we found a particularly interesting example of one. 

Burn Depth Classification with AI

In a 2021 study, researchers wanted to evaluate the effectiveness of AI in determining burn depth. To do so, they looked at pediatric scald models and different burn wound depths, which are: 

• Superficial 
• Superficial to intermediate 
• Medium to deep 
• Deep to full thickness 

A total of 100 burn images were obtained, 17 of which included all four burn depths. Those were used to train the convolutional neural network. The AI model was able to identify those four levels of burn severity using semantic segmentation of light photography images. Interestingly, the AI model was more accurate than expert clinicians, achieving 92% accuracy. Comparatively, the clinicians achieved 75% accuracy. 

AI Wound Healing Predictions

Chronic wounds are one of the industry’s greatest challenges. They can greatly affect a patient’s overall quality of life. We anticipate they’ll become more common with the aging population, requiring additional healthcare resources. Because of this, researchers are looking at AI to help address some of these concerns. 

AI to Predict Patient Risk of Chronic Wounds

In a 2022 study, Berezo et al. created machine learning models using EHR data to predict patients at risk of having unhealed wounds within four, eight, and 12 weeks from the start of treatment. These AI models were trained on three data sets of over 1.2 million wounds. They also included 187 covariates sharing information such as: 

• Wound attributes, like type, size, depth, location, infection presence, etc. 
• Demographics, like age, gender, tobacco use, etc. 
• Comorbidities, like coronary artery disease, congestive heart failure, hypertension, etc. 
• Other clinical factors 

The machine learning models then analyzed these inputs and developed an estimated timeframe for the wound to heal. Following completion of the study, the authors observed that this approach had three major benefits.

1. The machine learning methodologies swiftly and correctly processed vast amounts of data, recognizing patterns and predicting outcomes humans might not have predicted. 
2. The study’s use of a large dataset decreased the risk of overfitting, which is a machine learning problem that occurs when the model is unable to generalize and too closely mirrors the training dataset. 
3. The AI tool could identify wounds at risk of not healing or those that may take longer to heal, which helped practitioners issue better treatment plans. It also gives patients more realistic expectations, which in turn may reduce expenses and boost patient outcomes. 

AI Models to Produce Quantitative Wound Assessment 

Today, the standard of care involves using wound assessment tools to help determine prognosis. However, most also require manual assessment of several wound attributes and adept evaluation of multiple factors. This approach does still have its flaws: it slows down the wound prognosis process, may lead to misinterpretation, and can lead to significant variability. 

A 2023 study examined the impact of an AI-based prognostic model trained using only image-based objective features. Researchers focused on four wound types: pressure injuries, venous ulcers, diabetic wounds, and arterial ulcers. They selected those types of wounds because they are the most common in North American skilled nursing facilities and home health settings.  

The deep-learning AI models were tasked with producing objective measurements and predictions, such as the rate of healing, the probability of successful healing, and the projected time to complete healing. This quantitative data allows clinicians to track progress better, make well-considered choices, and customize treatment methods based on the AI model’s predicted outcomes. 

Why is AI in Wound Care Important?

Wound care management has long been challenging, as multiple factors can impact the wound healing process. 

With AI implementation, we may be able to address many of these concerns. 

AI Mitigates the Cost of Wounds and Aligns with Value-Based Care

It’s no secret that the United States health expenditures have increased over the years, but what about just on wound care management alone? In 2019, the U.S. spent more than any other country on wound care, exceeding $125 billion. China and Japan followed distantly behind, spending approximately $26 billion and $18 billion, respectively. 

To combat this, the U.S. healthcare system is looking for ways to decrease costs while maintaining optimal clinical outcomes. Research suggests that achieving this may require implementing advanced technologies, like AI. 

In addition, these goals largely align with value-based care, which incentivizes clinicians who prioritize outcomes-driven treatment plans and efficiency. While value-based wound care programs are in their infancy, providers who hope to improve care and be rewarded for doing so in the future shouldn’t ignore AI. 

AI Helps Promote Objectivity in Assessments

While clinicians aim to be as objective as possible, it’s understood that this is much easier said than done. The National Institute of Health (NIH) stated, “Each clinician will have widely differing and distinct opinions on wound therapy depending on prior experiences.” 

If we look at wound assessments, for example, there are elements that practitioners may answer differently based on their own personal interpretation, such as: 

• Wound size
• Degree of injury
• Drainage, pertaining to color, consistency, and amount  

This subjectivity in assessment, diagnosis, and treatment can result in oversights that increase one’s risk of infection and complications. However, AI, as demonstrated in a previously mentioned study, may be able to help with this. Although it’s not bias-proof, since it learns from large datasets resulting from human decision-making, it still pushes providers in the right direction. 

AI Streamlines Clinician Tasks 

To outsiders, wound management might seem like just looking at the injury and jotting down its appearance and odor. Yet, you know that it involves so much more than that. 

Proper wound management entails considering a patient’s overall health and determining whether they have any underlying diseases, such as diabetes or kidney disease, that may impair wound healing. It also includes assessing a patient’s nutritional status, as deficiencies can compromise their ability to heal.

Each of these components and several more should be considered, but this is not always simple, especially with a heavy caseload. 

Thankfully, artificial intelligence can help leverage human capabilities, allowing wound care specialists to maximize their time and provide better care. This is especially true when it’s paired with other technology, like a wound-care specific EHR.

AI in Wound Care: A Necessary Tool for Superior Care

The future of wound care management can no longer rely solely on the scalpels and rulers that made wound assessments possible in the past. It also cannot depend on clinicians utilizing only their clinical knowledge to make treatment decisions. If wound care experts hope to provide high-quality care while also maximizing their time and boosting their bottom lines, they’ll need AI-based tools.