Smart Bandage Delivers Electricity and Drugs to Heal

The smart bandage can accelerate wound healing by effectively tracking and responding to the healing process. The proof-of-concept device, called a-Heal, is designed to fit over a commercial colostomy dressing and contains a camera that takes images of the wound every two hours, as well as a wireless connection to a machine learning module that provides updated recommendations on how to stimulate healing.

An interdisciplinary research team led by Marco Rolandi, a professor of electrical engineering and computer science at UC Santa Cruz, has developed a treatment. Rolandi says the initial motivation was to reduce recovery time from battlefield injuries by 50 percent as part of a Defense Advanced Research Projects Agency initiative.

While a-Heal has been tested on a wound (on a pig) caused by an injury and can heal quickly, Rolandi says it could be applied to other types of wounds that have difficulty healing — such as wounds that are chronic or those that become infected.

Current wound dressings are a one-size-fits-all approach and cannot account for differences in how individual wounds heal.

To heal wounds faster, a-Heal monitors wounds, diagnoses the wound’s current healing stage, suggests treatments and delivers them. By repeating the process, the bandage creates a feedback control mechanism.

The machine learning algorithm, ML Physician, analyzes each image of the wound, “compares it with the training data, and decides at what stage the wound is at and whether the wound is healing in the desired way, or whether the wound requires intervention or treatment to speed up the healing process,” Rolandi says. At the recommendation of the artificial intelligence, the wearable device activates bioelectronic actuators to deliver one of two treatments, known to increase healing in animals: electrically stimulating the wound to reduce inflammation or infusing it with fluoxetine, a drug that promotes tissue growth.

The researchers tested the device on a pig because pig skin is similar in many ways to human skin, says Min Zhao, MD, a physician and professor of dermatology at UC Davis Health.

Although pads that provide electrical stimulation already exist, the a-Heal system goes a few steps further. “I’m not aware of anyone who has used photographic information with some kind of closed-loop operating system,” says Jeffrey Gortner, a physician, surgeon and professor at the University of Arizona with a joint appointment at Stanford University, who was not involved in the study. “So it certainly combines two things that people are working on in a new way.”

How smart bandages heal wounds faster

Every two hours, the bandage’s camera unit takes 11 images at different focal depths. When asked how well the camera can capture darker human skin tones, Rolandi said that if the training data is there, it should be fine, but it’s too early to tell because this version of the device wasn’t intended for human testing.

The ML Physician machine learning algorithm, located on a laptop, applies a leader-follower strategy to image analysis. Part of the algorithm, Deep Mapper, creates a predicted wound image, or “leader” — what the wound would look like if it healed perfectly. Deep reinforcement learning controllers attempt to “follow” the ideal image by controlling the amount of treatment that is delivered.

The machine learning algorithm chooses only one treatment at a time. Electrical stimulation is applied initially. Treatment automatically switches to drug delivery when the probability of the wound remaining in the inflammatory healing phase is 40 percent. (Wounds generally heal in four stages: blood clotting, inflammation, tissue growth, and maturation.)

The bioelectronic actuator used to deliver treatments is a cylindrical silicone polymer body containing eight reservoirs arranged in a circuit, four for electrical stimulation and four for drug delivery. Each tank contains an electrode, as well as a central counter electrode. The hydrogel attaches the electrodes to the wound.

Using iontophoresis, the device delivers either a saline solution for electrical stimulation or a solution of the drug fluoxetine. “We basically have an electric current from the treatment molecules instead of electrons,” Rolandi says. “Part of the circuit, if you will, is actually the wound bed. By measuring the current, we can count the number of molecules getting into the wound bed, so we have very precise control of the dose.”

The researchers tested the smart bandage attached to a belt on a pig.Image source: Hobo Li, Hsin Ya Yang, et al.

Smart bandage results

The results showed that 50% of the wound treated with the device was covered with new skin cells, compared to only 20% of the control wound. “It is encouraging that there is more re-epithelialization,” Gortner said. The device was used for the first seven days of a 22-day trial, but wounds were not completely closed at 22 days.

One gene associated with inflammation, interleukin 1 beta, was reduced by 61 percent in treated wounds. The presence of other genes associated with inflammation or anti-inflammatories is also trending in the right direction, Rolandi says. “in spite of [sample size] “If epithelialization is low and not statistically significant yet, it is good to see continued improvement in epithelialization, quality of epithelialization, inflammation, vascular growth, and wound maturation,” says Zhao.

Overall, “I think it’s a modest effect,” Gortner says.

The team hopes that improvements to the devices will lead to better clinical outcomes. “I think we have a lot of future work to do,” Rolandi says. They published their results in Biomedical innovations On September 23.

Gortner says he would have liked to see larger sample sizes, and that he would “really like to see them take the wounds all the way to closure with the machine running the whole time, and see the difference in volume time to closure.” The device was used for the first seven days of a 22-day trial, but wounds were not completely closed at 22 days.

Rolandi says the team did not follow the entire healing process because “first, intervening at the beginning of the healing phase tends to have the greatest impact on the healing itself as a whole. Second, it becomes very complex to conduct these large preclinical animal trials… For this first publication, we decided to start simple and treat the wound for just seven days.”

Going forward, the team plans to simplify the device. (The current version took a month to create, Rolandi says.) “We’re developing a flexible version. This is definitely something we knew was necessary.”