Interpretable Machine Learning for Assessing Frailty Syndrome (intFrail)

Project no.: PP2022/58/2

Project description:

Frailty syndrome is becoming a major challenge of the aging society. The prevalence of frailty is 17% whereas prefrailty may affect up to 60% of those over 65 years old. Frailty is a progressive impairment of physiological systems leading to reduced physical reserves which cause susceptibility to internal (e.g., surgery, comorbidities) and external (e.g., physical activity) stressors and increase the risk of various adverse health effects. The early stages of frailty can be clinically reversible hence it is possible to improve the frailty status by timely identification of frailty features and prescription of an appropriate intervention (e.g., physical training protein-enriched diet, etc.). Frailty is a multicomponent syndrome involving physiological reserve, physical and cognitive functions, so it is important to accurately identify the frailest components for an individual patient and strengthen them using personalized exercise programs. Training programs are often continued at the home environment, making the existing methods that rely on indexes and questionnaires and/or require supervision and special equipment unsuitable for frailty assessment outside the clinical setting. Accordingly, the intFrail project proposes an objective approach to frailty assessment based on the features extracted from biosignals recorded using wearable devices. This is expected to be achieved through interpretable machine learning which purpose will be to identify clinically informative features that would provide information on the frailest components of an individual patient. The use of interpretable machine training will allow to associate the frailty stage with the frailty features and thus will provide additional information to rehabilitation clinicians for the development of personalized training programs.

Project funding:

KTU Research and Innovation Fund

Project results:

During the project, an interpretable machine learning algorithm was developed to assess the weakened body functions of a patient with frailty syndrome.

Period of project implementation: 2022-04-01 - 2022-12-31

Project partners: Lithuanian University of Health Sciences

Andrius Rapalis

2022 - 2022

Biomedical Engineering Institute, Laboratory of Multimodal Biosignal Streams