ADVANCING STROKE RECOVERY: TECHNOLOGY BEHIND THE VIRTUAL ASSISTANT FOR PATIENTS

As part of the RES-Q+ project, the development of a virtual assistant for stroke survivors is progressing rapidly. Designed to collect patient-reported outcomes and provide reliable medical information, the assistant is intended to support individuals throughout their recovery journey. While still in development, the team is refining its functionality to ensure it meets the needs of both patients and healthcare professionals.

A user-centered process has been applied to inform the design of a mobile chat interface that, despite limited screen space, supports stroke survivors in completing their patient-reported outcomes. Many stroke survivors experience cognitive and motor impairments, so the design prioritizes a clear, simple, and intuitive interface to ensure accessibility. Stroke survivors have been involved at every stage—from initial research to testing—to ensure the assistant is practical, easy to use, and truly meets their needs. By incorporating their feedback, the design and functionality can be refined to better support recovery. However, recruiting stroke survivors for user testing has been challenging, as they can be hard to reach and sometimes hesitant to participate. Nonetheless, their input remains essential.

The user-centred design methodology at the core of the development process has consisted of interviews, workshops, and design studies with both patients and clinicians. Through these activities, key functionalities that directly benefit stroke survivors have been identified. One of the most important is the question clarification feature, which helps patients interpret medical questionnaires more easily and reduces confusion caused by complex or vague wording. Another crucial feature is the integrated Q&A system, which allows patients to ask stroke-related questions and receive answers based on clinically validated sources, thereby helping stroke survivors who, after being discharged from the hospital, struggle to find trustworthy information and reliable guidance.

To build an intelligent and responsive system, the team is using RASA, an open-source framework for conversational AI, in combination with LLAMA, a powerful language model. This setup enables the assistant to process natural language efficiently, making interactions feel intuitive and human-like. Developing a medical virtual assistant presents several challenges, the most critical being data security. Handling sensitive medical data requires stringent safeguards, and legal experts are consulted to ensure that patient information is securely stored and accessed only by authorised parties.

Looking ahead, preparations are underway for a large-scale study with 200+ stroke patients across two countries. This study will help evaluate the assistant’s effectiveness, usability, and impact on patient care. The collected data will also contribute to predicting patient outcomes and improving stroke rehabilitation. In the long term, the assistant’s capabilities are planned to expand with the integration of health tracking features, motivational data visualizations, and medication support. By continuously refining the technology and integrating patient feedback, the goal is to create a powerful tool that enhances stroke recovery and improves healthcare efficiency worldwide.

14. August 2025 | Veronika Němcová and Milo M. Skovfoged