Autonomous Closed-Loop Bioelectronic Wound Dressings for Chronic Wound Therapy

1. Problem Statement

Chronic wounds — defined as wounds that fail to progress through the normal healing cascade within 30 days — affect 8.2 million Medicare beneficiaries annually in the United States and impose a direct treatment cost of $28 billion per year on the Medicare system alone (Nussbaum SR et al., Value in Health, 2018;21(1):27–32). The three primary chronic wound types — diabetic foot ulcers (DFUs), venous leg ulcers (VLUs), and pressure injuries — share a common pathophysiology: persistent inflammation, bacterial biofilm colonization, and impaired angiogenesis that stall the wound in the inflammatory phase indefinitely.

Diabetic foot ulcers represent the most economically devastating subset. An estimated 1.6 million new DFUs occur annually in the United States, with 19–34% of the 38.4 million Americans living with diabetes developing a DFU in their lifetime. The recurrence rate is 65% at 3–5 years. One in six DFUs leads to amputation, and DFUs account for 83% of all significant lower-limb amputations. Five-year mortality following major amputation ranges from 50% to 70%. The average Medicare cost per DFU episode that heals primarily is $4,830, but costs escalate with complications: $13,580 for a single minor amputation, $31,835 for multiple minor amputations, and $73,813 for a major amputation.

The current standard of care relies on periodic clinical assessment — typically weekly or biweekly office visits — where a clinician visually inspects the wound, subjectively estimates healing progress, and adjusts treatment. Between visits, wound deterioration goes undetected. Bacterial infection produces measurable biochemical changes (elevated nitric oxide, hydrogen peroxide, pH shift, temperature increase) 1–3 days before clinical symptoms become visible. By the time redness or purulent discharge prompts a patient to seek care, the infection has established and treatment options narrow. A system that continuously monitors wound biochemistry, detects early infection signatures, and autonomously intervenes would fundamentally change the trajectory of chronic wound management.

2. State of the Art

Four research programs have converged to make autonomous closed-loop wound therapy technically feasible. None has produced a deployable commercial system combining all three required capabilities: multiplexed biosensing, ML-based classification, and autonomous therapeutic intervention.

Multiplexed sensing with closed-loop treatment (Caltech)

Wei Gao’s laboratory at Caltech developed a stretchable wireless bioelectronic system that simultaneously monitors uric acid, lactate, pH, and temperature in the wound bed while delivering combination therapy — antibiotic drug release and electrical stimulation — through the same flexible substrate. Published in Science Advances in 2023, the system demonstrated substantially accelerated wound healing in a rodent chronic wound model. In April 2025, Gao’s group published the iCares system in Science Translational Medicine, testing a microfluidic wound monitoring platform on 20 human patients with diabetic foot ulcers and venous leg ulcers. The device measures six biomarkers through a nanoengineered sensor array. An ML algorithm classifies wound severity and predicts healing time with accuracy comparable to expert clinician assessment. The 2025 human study validated monitoring and classification but did not test closed-loop drug delivery in humans.

Wireless closed-loop sensing and stimulation (Stanford)

Yuanwen Jiang, Zhenan Bao, and Geoffrey Gurtner at Stanford University developed a wireless closed-loop smart bandage that continuously monitors skin impedance and temperature, then delivers electrical stimulation to accelerate healing. Published in Nature Biotechnology in 2023 (41(5):652–662), the system demonstrated 25% faster wound closure and 50% enhanced dermal remodeling versus untreated controls in a mouse wound model. Transcriptomic analysis confirmed activation of proregenerative M2 macrophage gene expression. Gurtner subsequently enrolled 83 patients in a clinical trial at the University of Arizona. Preliminary results report 90% positive predictive value for wound complications using only 2 of 9 sensors.

ML-driven closed-loop bioelectronic therapy (UC Santa Cruz)

Marco Rolandi at UC Santa Cruz, with UC Davis and Tufts, developed the a-Heal platform under DARPA BETR funding. Published in npj Biomedical Innovations in 2025, a-Heal integrates an onboard camera, an ML model they term an “AI physician” that diagnoses wound stage, and a bioelectronic actuator delivering fluoxetine plus electrical field therapy. Preclinical testing showed approximately 25% faster healing versus standard care. The ML model runs on an external computer, not on-device.

The gap between these research systems and a deployable product is fourfold: (a) no existing system runs ML inference on-device; (b) closed-loop drug delivery has been demonstrated only in animal models; (c) all prototypes are laboratory-fabricated with no DFM analysis; (d) no group has engaged with FDA on a regulatory strategy for a combination product with an adaptive ML algorithm.

3. Foundational Research

4. Competitive Landscape

No commercial entity offers a product combining wound biomarker monitoring, ML-based classification, and autonomous closed-loop drug delivery or electrical stimulation. Every existing product addresses one or at most two functions.

Vomaris Innovations (acquired by Arthrex). Tempe, AZ. $13.2–$16.2M raised. Products: Procellera, JumpStart, PowerHeal — passive bioelectric antimicrobial dressings using silver-zinc microcell batteries. FDA 510(k) cleared (K160783). No sensors, no monitoring, no drug delivery.

Grapheal. Grenoble, France (CNRS spin-off, 2019). EUR 1.9M seed. WoundLAB graphene-on-polymer smart bandage — pH sensing and electrostimulation via NFC. Pilot clinical trials with diabetic patients. No autonomous drug delivery, no on-device ML.

Accel-Heal. UK. Disposable electrical stimulation device for hard-to-heal wounds. 12-day microcurrent protocol. CE marked (Class IIa). Treatment only — no sensing, no monitoring, no adaptive dosing.

MolecuLight. Toronto, Canada. FDA De Novo 2018 for i:X handheld fluorescence wound imaging device. 350-patient clinical trial, 3-fold increase in bacterial load detection sensitivity. Diagnostic imaging tool used during clinic visits — not a wearable, no treatment, no continuous monitoring.

ConvaTec. $2B+ revenue incumbent. Launched ConvaNiox (April 2025) — NO-generating wound dressing for DFUs, EU Class III. 63% improvement in healing at 24 weeks. Passive chemical release — no sensors, no feedback, no adaptive dosing.

The absence of direct competitors reflects structural barriers: (1) closed-loop wound dressings are Class III combination products requiring PMA; (2) incumbents manufacture roll-goods, not flexible electronics; (3) smart dressings that accelerate healing cannibalize the consumables business model; (4) no CPT/HCPCS code exists for autonomous wound monitoring with treatment. These barriers will take competitors 3–5 years to overcome, creating a first-mover window.

5. Total Addressable Market

Bottom-up calculation (US chronic wound treatment)

• Medicare beneficiaries with chronic wounds: 8.2 million/year
• Annual Medicare wound spending: $28.1B (conservative estimate)
• DFU patients: 1.6 million new cases/year
• Average annual per-patient cost: $33,000
• Avoidable amputation cost from early detection: $1.4–$7.4B annually
• Autonomous wound dressing as 30-day prescription at $500/unit, prescribed for 400,000 high-risk DFU patients/year, 4 units/year: $800M US SAM for DFU
• Expanding to VLUs (2.5M patients) and pressure injuries (2.5M patients): Total US SAM: $2.0–$3.5B/year

Top-down cross-check

• Smart bandage market: $926M–$1.75B (2025), projected $2.5B–$3.7B by 2030–2035, CAGR 15–17% (Precedence Research)
• Advanced wound care: $13.37B (2025) to $19.32B by 2030, CAGR 7.6% (MarketsandMarkets)
• The autonomous closed-loop segment — which does not yet exist commercially — represents the highest-value tier

Reimbursement

No dedicated CPT/HCPCS code exists for autonomous wound monitoring with integrated treatment. Near-term strategy combines RPM codes (CPT 99453, 99454, 99457, 99458) with wound e-stim codes (HCPCS G0281/G0282) and dressing supply codes (A6xxx). Stacking yields $150–250/patient/month. A new Category III CPT code would be the strategic long-term path.

6. Research Gap & HHA Contribution

What has not been done

No research group has integrated multiplexed electrochemical biosensing (Gao), ML-based closed-loop treatment decisions (Rolandi), autonomous drug delivery (Gao animal model), and electrical stimulation (Jiang) into a single manufacturable system with on-device ML inference. Each group solved components independently. Gao demonstrated sensing and treatment separately across two publications. Rolandi demonstrated ML-driven decisions but runs ML off-device. Jiang demonstrated closed-loop sensing+stimulation without drug delivery. The integration across all four components remains undemonstrated.

The specific technical gap

Four gaps exist between published results and a deployable solution:

1. On-device ML inference — every system requires external compute (smartphone or laptop). A truly autonomous dressing must run classification on ARM Cortex-M class microcontrollers (256 KB–2 MB SRAM). This requires model compression (quantization, pruning, knowledge distillation) that preserves clinical-grade accuracy.
2. Human closed-loop treatment — drug delivery + e-stim validated only in rodent models. Bridging requires FDA IDE and pivotal clinical trial for a combination product.
3. Scalable manufacturing — all prototypes are hand-assembled via photolithography and e-beam deposition. No DFM analysis exists. The first R2R smart dressing production (Purdue, 2025) covered only passive colorimetric sensors.
4. Regulatory strategy — no FDA Pre-Sub, IDE, or classification request has been filed for a closed-loop wound dressing with drug delivery.

Why HHA is positioned to close this gap

Hass Dhia (Co-PI): MS Biomedical Sciences with medical school background provides the wound biology domain knowledge that engineering labs lack. Understanding the wound healing cascade (hemostasis, inflammation, proliferation, remodeling), biomarker significance (NO = inflammation, H₂O₂ = infection, pH shift = healing trajectory), and clinical workflow integration. His experimental design capability enables correct IRB protocol design, endpoint definition, and animal-to-human translation study structure. His AI infrastructure architecture experience maps directly to building the data pipeline from wound sensor readings through ML classification to treatment actuation commands — the entire inference pipeline that currently runs on external devices and must be miniaturized for on-device operation.

Haedar Hadi (Lead PI): MS Computer Science (Boston University, Information Systems focus) provides the on-device ML engineering that all existing prototypes lack. The single biggest technical gap — moving ML inference from smartphones/laptops to microcontrollers embedded in wound dressings — requires TinyML model design, quantization-aware training, model compression, and adaptive control algorithms for drug delivery optimization. His evaluation methodology expertise is directly applicable to designing the validation protocol that demonstrates ML-driven treatment decisions match expert clinician judgment — the core evidence requirement for FDA PCCP submission. His cloud infrastructure expertise supports the patient monitoring dashboard that clinicians use to oversee autonomous dressing operation across patient cohorts.

Ahmed (Key Team Member, Director of Manufacturing): This is the most critical commercialization gap across all published prototypes. Every prototype is hand-assembled in a research cleanroom. Ahmed’s flexible electronics manufacturing expertise addresses the transition from lab-grade fabrication (photolithography, e-beam deposition) to production-grade manufacturing (roll-to-roll printed electronics, flexible PCB assembly, microfluidic channel bonding). His 3D printing scale-up capability is directly relevant — Gao’s iCares uses 3D-printed biocompatible polymer that must transition from single-unit lab printing to statistical process-controlled production. His GMP compliance expertise (21 CFR Part 820 quality system regulation) ensures the manufacturing process meets FDA requirements from day one, not as a retrofit.

The lab-to-production bridge

Most research proposals for bioelectronic wound dressings end at “it works in the lab.” This proposal includes explicit DFM milestones at every phase, ensuring that prototype decisions consider production scaling, tolerance analysis, and quality systems from day one. Ahmed’s manufacturing engineering expertise addresses the valley of death between TRL 4–5 prototypes and TRL 7+ deployable systems — the gap where most funded bioelectronic research stalls. The originating labs have not closed this gap because: (1) academic labs have zero manufacturing expertise or mandate — they publish papers, not products; (2) Gao’s lab has no manufacturing partner despite producing two high-profile publications; (3) DARPA BETR funding (Rolandi) explicitly targets TRL-5 but the transition from TRL-5 prototype to TRL-7 production requires manufacturing engineering that no BETR performer team includes; (4) wound care incumbents (Smith+Nephew, ConvaTec, 3M/Solventum) have not entered the space because their manufacturing infrastructure is optimized for passive materials, not active electronics.

Capability gaps and how funding addresses them

The team lacks: (1) an electrochemical sensor specialist — grant funds support a postdoctoral hire or academic collaborator with nanoengineered biosensor expertise; (2) a regulatory affairs professional with Class III combination product PMA experience — funded via subcontract to a specialized regulatory consulting firm; (3) a clinical wound care collaborator (dermatologist or vascular surgeon) as clinical PI for the human trial — recruited from a partner wound care center.

7. Comparable Funded Projects

Government agencies have committed substantial funding to closed-loop bioelectronic wound therapy, validating both technical feasibility and clinical urgency.

These awards total over $60M in government investment (2020–2025), with DARPA alone committing over $60M across BETR and BEST. The progression from BETR (2020, research prototypes) to BEST (2025, TRL-5 with FDA transition) demonstrates accelerating funder urgency.

8. Opportunity Assessment

TRL evidence chain

TRL 4. Closed-loop drug delivery + e-stim validated in rodent models by two independent groups (Gao 2023, Jiang 2023). ML-based wound classification validated in 20 human patients (Gao 2025). ML-driven treatment decisions demonstrated preclinically (Rolandi 2025). TRL 5 requires first-in-human closed-loop treatment.

Technical risks

Regulatory pathway

Combination product (device + drug) under 21 CFR Part 3. CDRH-led PMA pathway if using already-approved drug molecule. Predicates: MolecuLight i:X (De Novo 2018, wound imaging), Vomaris Procellera (510(k) K160783, bioelectric dressing), Dexcom G7 (510(k), continuous wearable biosensor). If ML algorithm is locked after training, standard software V&V applies. If adaptive (learning from patient data), FDA PCCP framework applies. Over 53 devices have FDA-authorized PCCPs as of 2025, but none in wound care — first-mover PCCP would establish regulatory precedent and competitive moat. Recommended: initial submission with locked algorithm, PCCP supplemental for adaptive capability post-launch. Regulatory timeline: 3–5 years (IDE + pivotal trial + PMA review). Regulatory barriers framed as competitive moat: competitors who begin development after clinical trials start face the same 3–5 year regulatory pathway.

CPT/HCPCS reimbursement

Near-term stacking of RPM codes (99453/54/57/58) + wound e-stim (G0281/G0282) + dressing supplies (A6xxx) yields $150–250/patient/month. New Category III CPT code application for long-term dedicated reimbursement.

Proposed experimental approach (first 6 months)

Months 1–3: Acquire wound sensor validation datasets (collaborate with Gao lab for iCares training data access). Begin TinyML model architecture selection (CNN vs. lightweight transformer). Design microcontroller target (ARM Cortex-M4F or M7). Begin electrochemical sensor characterization in simulated chronic wound fluid (pH range 6.5–8.5, varying bacterial loads). Ahmed: manufacturing requirements specification, flexible substrate material sourcing, PCB-to-disposable coupling mechanism design.

Months 4–6: Train and compress wound classification model. Validate compressed model against full-precision baseline. Begin microfluidic drug delivery channel design with volumetric flow sensing. Ahmed: first prototype disposable sensor strip using screen printing (target: 10 units for benchtop testing). Prepare FDA Pre-Sub meeting request with CDRH.

9. Team Fit

The team does not include an electrochemical sensor specialist or a regulatory affairs professional with Class III PMA experience. Grant funding would support: (1) a postdoctoral researcher with nanoengineered biosensor expertise, (2) a regulatory consulting subcontract for FDA Pre-Sub and IDE preparation, and (3) a clinical wound care specialist (dermatologist or vascular surgeon) as clinical PI for the human validation study.

10. Recommended Next Steps

Target funder programs

Estimated total: $2–5M over 24 months for Phase 1 (sensor validation + ML compression + first prototype + FDA Pre-Sub).

24-month milestone timeline

• M1–3 R&D: Wound sensor dataset acquisition, TinyML architecture selection, electrochemical sensor characterization in simulated wound fluid. Manufacturing: Requirements specification, flexible substrate material sourcing, screen-printed sensor strip prototyping. Regulatory: Literature review for combination product classification, draft Pre-Sub briefing document.
• M4–8 R&D: ML model training and compression (target: <500KB, AUC >0.90). Microfluidic drug delivery channel design with flow sensing. Benchtop closed-loop validation (sensor+ML+actuator on simulated wound). Manufacturing: Prototype v1 — integrated disposable strip (sensors + drug reservoir + e-stim electrodes). Go/no-go M8: compressed ML model accuracy confirmed against clinician ground truth.
• M9–14 R&D: Large-animal preclinical study (porcine chronic wound model). Closed-loop drug delivery + e-stim vs. monitoring-only vs. standard care. Manufacturing: Prototype v2 with wireless PCB integration, battery management, Bluetooth connectivity. DFM analysis — COGS target <$50/disposable, <$200/reusable PCB. Regulatory: FDA Pre-Sub meeting with CDRH. Go/no-go M12: sensor drift <15% after 7 days in wound exudate.
• M15–20 R&D: Large-animal study completion. Data analysis. IDE application preparation. Manufacturing: Process validation for screen-printed sensor production. Supplier qualification for medical-grade polymers. Quality system documentation (21 CFR 820). Regulatory: IDE submission for first-in-human pilot study.
• M21–24 R&D: First-in-human pilot study design (target: 20–30 patients, DFU and VLU). IRB approval at partner wound care center. Begin enrollment if IDE cleared. Manufacturing: Production pilot run (50–100 disposable units for clinical trial). Regulatory: FDA IDE review and response. Phase 2 funding application. Go/no-go M24: large-animal closed-loop treatment demonstrates statistically significant healing acceleration (p<0.05) vs. standard care.