Spotlight on data analysis and biosignatures

We spoke to colleagues from the work stream on data analysis and biosignatures to find out more about their contribution to the study.

The data analysis and biosignatures stream of work is overseen by industry lead Nikolay (better known as ‘Kolya’) Manyakov, and academic lead Björn Schuller. Kolya is based in Beerse, Belgium and leads the digital phenotyping group at Janssen Pharmaceutica NV. Björn is a professor at the University of Augsburg in Germany, where Nick Cummins, who supports him on this work stream, is also based.

Björn and Nick talked to us about their work on the project.

 

What exactly is the purpose of the data analysis and biosignatures work stream?
Björn: Our aim is to enable the analysis of data collected in RADAR-CNS trials, to explore the associations between remote measuring technology metrics and the remission, relapse, and recurrence of the disorders.

We are primarily responsible for extracting meaningful information from raw signals recorded during clinical trials and aggregating them in such a way that our colleagues in the clinical work streams can perform their analysis. We also develop modelling methodologies for analysing data recorded during clinical trials.

What sort of backgrounds do you and your colleagues have?
Nick: We come from both academic and industry backgrounds. The team has very diverse expertise in the processing of different streams of data collected during trials, including audio and speech data, heart monitoring data, data from accelerometers, electrodermal activity, electromyography and electroencephalography, and data collected during clinical tests to assess a person’s mobility.

We also have expertise in the analysis of ecological momentary assessment (the completion of brief self-reports at random occasions during normal waking hours, which are designed to capture in-the-moment assessments of current experiences). All partners have extensive expertise in machine learning, data mining and statistical analysis for longitudinal data.

What are your major achievements so far – and what are you next steps?
Björn: We have developed an analysis plan, identified features to be extracted from biosignals and other data streams, started programming algorithms for feature extraction, developed a methodology to model data and designed a schema for a features database (a structured and organised set of participant information extracted from the data collected via the RADAR-BASE apps).

Our task now is to finalise the build of the feature database and feature extractors (the processing of the raw patient data into more manageable groups for future processing), and conduct the automatic processing of data and consequent population of the database.

What has been the biggest challenge in your work so far?
Björn: That would be dealing with everything connected to data which is collected in real life - so missing data, noisy data, inaccurate data and so on.

What is the most fun part about working for RADAR-CNS?
Björn: It is great to work in such an interdisciplinary setting, with us all united to potentially make a huge change in life quality at scale.

Nick: I really enjoy working towards solving real-world problems in health tech as part of a cooperative, encouraging and friendly interdisciplinary team. It is a very motivating environment to be working in.

What is the most unexpected thing you have discovered as part of your work?
Björn: We expected a much higher rate of data completion by participants and the reasons for the missing data are still unclear. This is an interesting challenge: coping with missing data at a ratio of significantly more missing than existing data requires powerful new algorithms and ideas from our team.

Nick: Just how challenging the collection and analysis of real-world data has been. However, at the same time, this is exciting as it gives us the opportunity to take what we have learnt in more lab-based studies and learn how to apply it in the real world.

What do you predict for the future of wearable tech for health?
Björn: Wearable health tech will eventually become as ‘normal’ as simple hygiene in a hospital. Today, it seems unimaginable that Ignaz Semmelweiss and others had to fight for decades and almost in vain to introduce hand-washing in the hospitals of their time. The day will come when it is hard to imagine a time before wearable healthcare technology, as it will become increasingly powerful and reliable in diagnosing earlier than ever, thanks to personalised real-time and in-situ monitoring.

 Nick: I believe the combination of health tech and Artificial Intelligence (AI) provides a myriad of opportunities to enhance medicine and healthcare, especially AI-based systems for patient-centred, personalised and preventative treatments.