Skip to main content

Trinity College Dublin, The University of Dublin

Trinity Menu Trinity Search

You are here FRAILMatics


FRAILMatics is hiring! Click here to see vacancies

FRAILMatics’ quest is to research and develop more accurate frailty tools that without expert input automatically identify subtle dysregulated responses to stressors across physiological systems. For this quest, FRAILMatics will have access to population and clinical cohorts that contain a vast amount of suitable cross-sectional and longitudinal data. The population cohort will be the vast and rich data resource contained in TILDA. Later on, signals will be validated in a small (n=100) clinical cohort recruited from ambulatory care clinics at the Mercer’s Institute for Successful Ageing (MISA) in St James’s Hospital, Dublin.

TILDA is one of the most comprehensive research platforms in existence, nationally and internationally. With FRAILMatics, TILDA’s vast data resource will be leveraged for the first time through an SFI award. TILDA is a study based on a population-representative sample of >8,000 community-dwelling individuals aged 50 or over. Each participant undergoes an interview in their home, fills in a self-completion questionnaire and is invited to undergo a detailed health assessment. Five waves of data collection are complete and the study will collect its sixth wave (i.e. its third health assessment wave) in 2020 ( A key strength of TILDA is its longitudinal design, and detailed longitudinal health data spanning 10 years will be available to FRAILMatics’ researchers.

In TILDA health assessments, computer-based neurocognitive tests are used to measure response times and errors in tests of sustained attention (SART: sustained attention to response task) and choice reaction (CRT: choice reaction time), where participants experience the stress of time under specific test instructions involving a fine motor response. TILDA also has a vast cross-sectional and longitudinal dataset of automatically recorded gait parameters at the person’s preferred speed, under cognitive challenges, and at maximum speed (GAITRite® system). In the clinical cohort we will also collect three-dimensional gait analysis with the Codamotion® system. In both cohorts, multiple cardiovascular parameters are non-invasively recorded continuously (with the Finometer®) during an active stand test, which challenges the ability of our body systems to compensate for a rapid change in position. We simultaneously record brain oxygenation parameters during the active stand with the PortaLite® NIRS system.

Real time physiological measurement generates vast amounts of data. FRAILMatics will have access to big cross-sectional and longitudinal datasets of over 1 billion data points, containing detailed dynamic physiological data across the systems and also a wide range of outcome data that can be used to train models if desired. FRAILMatics will tackle this big data challenge by embracing two complementary approaches:

On the one hand, we will use machine learning that incorporates manually extracted features. For example, for the active stand we have pilot data using information entropy, which directly correlates with the complexity of a signal and can be measured before and after stress. We can use multiscale entropy methods to integrate cardiovascular signals, and signals across systems.

On the other hand, we will use deep learning methods where models will not be constrained by given features. For example, for the active stand we have pilot data using Bayesian functional clustering. These models are computationally expensive, especially if we want to cluster a large number of functional variables and variables across systems. This is why FRAILMatics requires a new start-up HPC infrastructure.

The outputs of FRAILMatics will be in the form of new models of multiple physiological dysregulation to stressors. FRAILMatics’ ambition is to produce software packages that when incorporated into a brief frailty assessment battery based on existing technologies can detect and quantify patterns of vulnerability to stressors across the frailty spectrum, and therefore increase the detection of those who are at higher risk of complications with higher specificity than current frailty identification methods. FRAILMatics will advance the science of frailty and align with the Research Priority Areas in the areas of Diagnostics and Medical Devices, and also with the Sláintecare Action plan, in that savings from reducing the cost of avoidable complications could be invested in expanding new models of care based on e-health solutions. Over 5 years, FRAILMatics will generate a high volume of scientific publications and invention disclosures. At the end of the programme, FRAILMatics’ tools will be at an early stage of technology readiness, requiring new collaborations in the field of medical device trials.

To facilitate FRAILMatics’ research programme, SFI granted the PI an infrastructure budget of circa €250,000 to purchase a new High Performance Computing (HPC) system based in TILDA. This new system is made up of a Storage Component, a HPC Cluster (mixed CPU and GPU cluster), and a high speed network between the storage component and the HPC cluster to facilitate rapid ingress of data and output of results. The central equipment is housed in a secure and environmentally controlled College Data Centre backed by an uninterruptable power supply, and managed by Research IT (formerly the Trinity Centre for High Performance Computing). The system is also composed of a number of new High Performance Workstations and peripheries located in TILDA offices (in Trinity College Dublin and MISA) to facilitate development of workflows and methodologies. These computing peripheries are managed by TILDA IT staff.