Biomarker Engineering and Analytics

What our Biomarker Engineering and Analytics (BEA) group offers

Methods & biomarkers

BEA is engaged in the development of new measurement methods. The raw data collected by these methods are processed and funnelled into informative biomarkers. By combining biomarkers (features) with machine learning we can also yield new insights and new (compound) biomarkers.

Analytics & machine learning

BEA’s analysts aim to distil biomarkers from the vast amount of (raw) data collected. We work with a range of data sources, including data collected using a single method — such as responses to stimuli and questionnaires, or electrophysiological data — as well as data collected from multiple methods or even multiple studies. Data from multiple methods or studies are often used to build predictive and generalisable machine learning models.

Validation

We perform validation studies to ensure the validity of our methods and biomarkers. The first step is to validate the system and/or method based on GAMP5 guidelines. Then, by means of a pilot study, the range, repeatability, and minimal detectable effect of the method is explored. To finalise the validation process, the method is deployed in a clinical trial to assess its power to detect effects of healthcare interventions. 

Analytics project examples

From raw data to biomarkers

• EEG analysis

  • E(R)P analysis in time, frequency, and time-frequency domains

  • TMS evoked potential (TEP) analysis in temporal-spatial domain, including statistical analysis of multidimensional TMS-EEG data

  • Quantification of high-gamma activity (between 50–170Hz)

• Trial@home remote monitoring

  • Estimation of Timed Up and Go (TUG) times using smartphones and wearable devices

  • Automated detection of coughing and crying of paediatric patients using a smartphone microphone (paper)

  • Monitoring of asthma and cystic fibrosis severity among paediatric patients using a smartphone-connected spirometry device (paper)

• Time-series analysis

  • Development of temporal-spatial-frequency biomarkers for finger tapping behaviour

  • Development of biomarkers for spiral tracing

  • Development of biomarkers for drift in eye focus

  • Development of biomarkers for muscle strength and fatigue

Applied machine learning

• EEG analysis

  • Identification of features predictive of treatment effects (example)

  • Development and validation of a model estimating changes in brain age

• Trial@home remote monitoring

  • Identification of features that can distinguish behaviours of patients with major depressive disorder from behaviours of healthy controls

  • Estimation of major depressive disorder severity using smartphones and wearable devices

  • Exploration of smartphones and wearables to characterise facioscapulohumeral muscular dystrophy (FSHD) patients (paper)

  • Estimation of post-discharge recovery after acute paediatric lung disease using a smartwatch (paper)

• Microbiome analysis

  • Identification of key proteins and bacteria for eczema in both lesioned and non-lesioned skin

• Driving performance

  • Assessment of drug-induced impaired driving behaviour

  • Characterisation of sleep-deprived driving behaviour