We’ve been saying it for years: a healed diabetic foot ulcer isn’t cured. It’s in remission. And like any remission, the period right after you declare it is when the risk of relapse is highest — roughly 40% within a year.
The problem is that remission care mostly happens at home, between clinic visits, in the gap where patients are left to interpret complex instructions on their own. Wear these shoes. Watch your skin. Don’t do too much. But also don’t do too little. It’s a lot to ask of someone who just spent months healing a wound.
Our new protocol paper, just published today in Sensors, describes a conversational agent — a chatbot — designed to sit in that gap. Led by Lucian Feraru with David Klonoff, Bijan Najafi, Magdalena Antoszewska, and me, it lays out a system that treats walking like a drug: dosed carefully, titrated based on tissue response, and held or reduced when warning signs appear.
Here’s how it works. After a patient’s wound heals or they undergo limb reconstruction, they enter a 12-week chatbot-guided remission protocol. The system pulls in step counts from their phone or wearable, paired foot skin temperatures, shoe wear time, and symptom checks. Every day, it runs through a decision loop: ingest data, check quality, assign a risk tier (green, amber, or red), and issue the next day’s activity and footwear recommendation. If temperatures spike — a side-to-side difference of 2.2°C or more, the threshold validated in multiple RCTs — the chatbot pulls back on activity and, if it persists, escalates to the clinical team.
The footwear module is just as structured. New therapeutic shoes get broken in over a week: 30 to 60 minutes on Day 1, adding one to two hours daily, with post-wear skin checks at each step. Redness that doesn’t resolve in 30 minutes pauses the whole progression. Only after full-day wear is tolerated does the activity dosing begin — starting at baseline step counts and advancing by roughly 500 steps per day each week, gated by thermometry.
This builds directly on our SmartBoot R01 work, where we embedded sensors into offloading boots and used chatbot-driven nudges to improve adherence in a nested sample of patients. The results were encouraging enough that we’re now designing a formal large-scale study to confirm or refute what we saw. The current protocol — funded by the NIH and the PhRMA Foundation — is the bridge: a single-arm feasibility pilot of 30 patients to test engagement, safety, and implementation fidelity before we go bigger.
What I find most compelling about this work is the framing. We don’t tell patients to “be careful.” We give them a number. A step target. A shoe schedule. And then we watch what happens — in near real time — and adjust. It’s the same logic we use when titrating insulin or adjusting blood pressure medications. The tissue tells you what it can tolerate. You listen.
The paper also looks ahead. The trajectory from conversational AI to remotely operated physical systems is already visible — CMS recently highlighted Alabama’s plan to deploy telerobotic ultrasound in maternity care deserts, and early evidence from rural Norway shows that robot-assisted obstetric ultrasound achieves excellent reliability. The chatbot we describe here may eventually serve as the conversational front end for a broader ecosystem of remotely operated diagnostic and therapeutic devices extending into the home.
This is a protocol paper. No clinical outcomes yet. But the architecture is built, the decision logic is prespecified, and the pilot is ready to enroll. We’ll report back. Our current data from the SmartBoot study using a nested sample of patients interacting with chatbots has been very exciting indeed.
The paper is open access — you can read the full thing here.
Citation: Feraru LM, Klonoff DC, Najafi B, Antoszewska M, Armstrong DG. “AI-Guided Remission: Protocol for a Conversational Agent (Chatbot) for Dosing Activity and Footwear Progression After Diabetic Limb Reconstruction.” Sensors. 2026;26(8):2299. doi: 10.3390/s26082299
#DiabeticFoot #DFU #Remission #Chatbot #AI #ArtificialIntelligence #DigitalHealth #Thermometry #Footwear #Offloading #Wearables #Sensors #LimbPreservation #ActAgainstAmputation #SmartBoot #ActivityDosing #WoundPrevention #Telerobotics
Leave a Reply