As populations age worldwide, dementia is becoming an increasingly pressing healthcare challenge. Patients suffering from dementia often experience difficulties in communication, memory, and behavior, making everyday interactions and care coordination more complex. At the same time, family members, caregivers, and medical professionals must collaborate effectively to ensure high-quality care, often without timely access to crucial patient information.
AI-based voice assistants - already commonplace in our homes - offer untapped potential to support dementia care in meaningful ways. If equipped with intelligent language capabilities, such systems could support daily health routines, adapt to patients’ cognitive needs, and assist caregivers with real-time insights.
Research Focus and Goals
This research project builds upon the ADELE health assistant platform to explore AI-powered voice assistants tailored for dementia care. It focuses on three core areas of innovation:
(1) Detecting Dementia Progression via Speech
We explore machine learning models that can estimate the degree of cognitive impairment based on spoken language patterns and behavioral cues. Both direct responses to targeted questions and subtle, implicit language changes are analyzed over time to recognize signs of disease progression.
(2) Adapting Conversational AI for Dementia Patients
Leveraging techniques from the field of generative large language models, we aim to make voice assistants more adaptive. The goal is to personalize conversations to the patient's current cognitive state, adjusting speech complexity, repetition, and conversational strategies to match their capabilities and comfort.
(3) Improving Access to Patient-Centered Health Information for Families, Caregivers, and Doctors
We develop a smart information system that enables seamless communication between patients, family members, caregivers, and healthcare providers. By providing context-aware, voice- and text-based summaries of relevant health data, we aim to reduce misunderstandings and prevent information loss during care transitions.
We are actively seeking motivated bachelor and master students interested in exploring this innovative intersection of AI and healthcare through guided research or thesis projects. Working students are also encouraged to apply.
Sponsored by:
Bavarian Ministry of Economic Affairs, Regional Development and Energy
