MiiCare secures patent for groundbreaking machine learning model predicting falls & dementia through footstep sound analysis
MiiCare has received a patent that uses the sound of the footsteps to analyse abnormal gait patterns linked to fall risks and dementia.
MiiCare harnesses the power of Artificial Intelligence (AI) to support individuals aged 50 and above in managing chronic conditions, thereby contributing to their overall health and longevity, empowering caregivers with personalised health and wellbeing plans, along with tools to facilitate self-care within the home environment. MiiCare now serves 4 ICBs, 18 Local Authorities (City Governments) and 12 Private Care providers in the UK and are expanding to the US, EU and India.
At the heart of this cutting-edge solution is a home hub housing a Language Model (LLM) powered Virtual Companion (VC) named Monica. Co-designed with input from older adults, including a lady named Monica, this digital carer guides individuals in adhering to their care plans, including medication compliance. Monica utilises a network of wireless infra-red sensors and smart wearables to analyse behaviour and health indicators, providing caregivers with "meaningful insights." Caregivers affectionately refer to Monica as their co-pilot, supporting them in delivering optimal care and achieving the best possible health outcomes.
One notable application of MiiCare's technology is MiiGait, a tool employed to analyse the gait parameters of an older adult, which was implemented to users over a 15-month period spanning 2021-2022.
MiiCare has been developing a novel Machine Learning (ML) algorithm that utilises the sound of footsteps to analyse the gait patterns of older adults. This model, incorporating a Bidirectional LSTM paired with the Attention mechanism, extracts "features" from audio data to detect gait. Importantly, the model operates independently at the edge on the MiiCube, eliminating the need for internet connectivity.
Clinicians rely on gait analysis to predict physiological and cognitive abnormalities. For instance, the "cadence" of gait, i.e., the average number of steps taken per minute, has proven indicative of a high risk of falls, increased frailty, and the onset of neurodegenerative conditions such as Dementia and Alzheimer’s.
MiiCare's AI team has contributed to four peer-reviewed publications on Audio-based Gait Analysis technology. Whenever an older adult walks past the strategically placed home hub, the ML model captures and processes the sound of footsteps to extract gait parameters. MiiCare's Health AI then analyses trends in these parameters, predicting instances when an older adult is at a heightened risk of falls or exhibiting signs of cognitive decline.
The MiiCare team studied the anatomy of the human feet and how it produces the sound of footsteps every time the older adult walks past the MiiCube. MiiGait was then trained to analyse ambient sound to extract these footstep sounds as the first step towards computing gait parameters.
In contrast to existing fall detection and prediction technologies that heavily rely on manual inputs from caregivers and older adults, MiiCare is at the forefront of revolutionising the industry. By automating the collection of "digital biomarkers" and providing caregivers with actionable intelligence, MiiCare enables them to dedicate more time to meaningful interactions with the older adults under their care.
Patent no. GB2607561
