Analyzing biosignals using Artificial Intelligence

What if artificial intelligence could analyze an electro-cardiogram (ECG) or a medical image scan and highlight results that physicians should be looking at in priority? Thanks to signal and image processing, and machine learning techniques, that is the goal of the new Biosignal Processing Group at the Idiap Research Institute.

Biological data extractible from the human body are an amazing opportunity. However, it is often hard to make sense from these data, especially when it comes to diagnose or to understand a biological process in the human body. The newly created Biosignal Processing Group of André Anjos will use machine learning tools – the automated statistical learning process by a software – to be able to analyze such data. The long-term aim is to have tools that one can apply to medical data and that can help clinicians and health-care practitioners in their tasks, as well as improve treatments and patients well-being.

A great potential for practical applications

For example, the detection of pathologies in ECG (electro-cardiograms) require measuring variations that make sense from a medical point of view, as there is a lot of ordinary variations in the natural heartbeat. In general, pattern recognition in challenging conditions, such as with regard to ECG signals, is a broader issue in many research fields. Health care is not an exception. An X-ray scan or an eye-fundus image, as well as the evolutionary curve of the sugar level in blood are not that different from an ECG: they are images signals from which one needs to extract a meaningful information.

A toolkit for Artificial Intelligence

Building on the Idiap expertise in image, signal processing and pattern recognition, the tools developed by the institute can be useful for the analysis of many biological signals from the human body. The toolkit for reproducible research called “Bob” already contains various baseline algorithms in signal, image analysis and pattern recognition. The goal is to continue to extend this framework to accommodate tools that are also useful for biosignal data processing and understanding.

Website: Biosignal Processing