AI-supported wound reporting
KickstartAI x LUMC
Through a collaborative effort, the Leiden University Medical Center (LUMC) and KickstartAI leveraged AI to deliver two key outcomes: minimized administrative work for nurses and a major uplift in the consistency and quality of electronic wound documentation. This initiative ensures healthcare professionals can focus more energy and time on patient-centered care.
Here you can learn about the project, from the approach to ultimately developing TAIME, an AI assistant that drafts parts of a wound report from a photo. From problem to first pilot:
The problem
Goals
Approach
Model development: TAIME, an AI assistant for wound reporting
First pilot results
Key takeaways
Next steps
The problem
Within the LUMC, countless wounds are treated annually. The care provided for those wounds relies heavily on accurate, timely administration. Nurses record the status of wounds in the Electronic Health Record (EHR) to ensure that the information needed to provide appropriate wound care is available across the entire care team.
In practice, this reporting process is manual and time-consuming, and often takes place daily. Nurses first assess the wound, take a photo, upload it to the electronic patient record (EPR), and then complete the TIME form – a structured framework used to document key aspects of wound healing, including tissue condition, infection, moisture balance, and wound edges. At a later stage, a wound care specialist reviews the information and provides an assessment.
With increasing pressure on healthcare specialists in the Netherlands, this essential administrative task is widely experienced as a significant burden among nurses in the LUMC. Therefore, LUMC and KickstartAI partnered to develop an AI solution to decrease this administrative workload, standardize documentation, and ultimately allow nurses to dedicate more time to direct patient care.
Goals
The goal of the project is to develop an AI solution using image recognition that can automatically generate and populate structured wound administration reports from a single photograph of a wound.
Approach
The approach consists of three phases:
1. Use case identification (completed): Understanding the current process and defining the specific problem to be solved.
2. Model development (first pilot completed): Building and testing the AI models. Watch how TAIME works and the first pilot results.
3. Implementation and adoption (next step): Integrating TAIME into LUMC's IT landscape and clinical workflows, starting with a pilot with real patients. Go to the next steps.
In the use case identification phase, LUMC and KickstartAI completed a crucial on-site observation through various shadowing sessions with nurses in surgical wards and the outpatient wound clinic. By combining these practical insights with deep-dive interviews with clinical experts, the team gained a clear understanding of the key pain points.
This process ultimately led to a well-defined use case for the next phase: developing a solution for automatic photo-to-wound reporting.
Key features
The AI system we are developing consists of the following components:
Segmentation: Detecting the precise wound area within the photograph, including the identification of different tissue types or colors present in the wound bed.
Object size measurement: Accurately determining the real-life size of the wound and its segmented areas (colors).
Translation to wound report: Translating the segmented and measured data into a standardized, structured digital wound report in the TIME format suitable for the EHR.
Upcoming technical blog by machine learning engineer, Julian von der Goltz, will follow shortly.
Expected results
The expected results of the AI solution are twofold:
Reduced Workload: The time required for nurses to complete a wound report will decrease dramatically from the current 5–10 minutes to an estimated 1 minute per wound care moment, significantly lowering the administrative burden and freeing up valuable time for direct patient care.
Improved Quality: Creating wound reports in a uniform structure and manner during each assessment can improve the consistency and quality of reporting, allowing nurses to document more efficiently while reducing variation in interpretation, level of detail, and photo quality. This standardization will support clearer information exchange between caregivers and contribute to more consistent and effective treatment decisions.
Model development: TAIME, an AI assistant for wound reporting
Together with LUMC, we developed TAIME, an AI assistant that helps nurses document wounds faster.
First pilot results
During the pilot, three experienced wound care nurses assessed 150 existing wound cases. For wounds measuring five centimetres or more, the average deviation of the automated measurements was 6.8% for wound length and 11.4% for wound width.
The test shows that AI can already prepare several parts of the wound documentation process. Detecting precise wound boundaries still needs improvement.
Read the full press release: AI helps nurses document wounds faster
Read the press release in Dutch
Key takeaways
“We all need to learn from one another if we want to successfully bring AI into clinical practice and achieve real adoption in such a challenging environment. And therefore, collaboration is key,” says AI project lead Anouk Wolters.
From the use case identification to model development, here is what we learned working alongside LUMC:
Shadowing nurses on the wards and in the outpatient wound clinic helped us define a use case that fits their daily work.
The pilot shows that AI can already prepare parts of a wound report. Detecting the exact edges of a wound still needs improvement.
Every correction a nurse makes creates new, expert-validated data that we can use to improve the models.
Next steps
TAIME originated from the 2025 National AI Challenge, an initiative by KickstartAI and the AI Coalition for the Netherlands (AIC4NL), and is developed together with Cairelab, LUMC’s AI centre of excellence.
The next step is a pilot with real patients to test how much time is actually saved on a day-to-day basis and whether it helps nurses complete wound reports more frequently.
Related reading about AI in healthcare: Hallucination detection for small on-prem medical LLMs: what works, what doesn’t, and why