Extracting information from a noisy simple electrical signal

This flyer shows feature extraction from a type of field potential (FP) of human induced pluripotent stem cell derived cardiomyocytes (hiPSC-CMs) obtained by multi-electrode array (MEA) technology [1]. The challenge dealing with FP data is the high dimension curse which causes problems related to costs of data processing and computation. The reason is that any simple operation on the high dimensional FP will cost tremendous computing resources compared to low dimensional data. Today, we would like to present a few matters about signal processing to overcome the high dimension curse.  

So, we introduce the feature extraction technique which is a common signal processing technique that can reduce the dimension of the raw signal data and extract the important information in the FP. One example in the flyer is to compute Depolarization Amplitude (DA), DA is computed by using the maximum amplitude minus the minimum amplitude. By defining similar methods as DA computation to compute meaningful features, we could observe changes in the FP signals induced by drug toxicity through specific features instead of dealing with high dimensional signals directly. Because of the manageable data size of the extracted features, we could also speed up the training phase and reduce the computation costs of further modelling.  

Reference:

1. F. Raphel, T. De Korte, D. Lombardi, S. Braam, J.F. Gerbeau, “A greedy classifier optimization strategy to assess ion channel blocking activity and pro-arrhythmia in hiPSC-cardiomyocytes,” PLoS Computational Biology, vol. 16, no. 9, pp. e1008203, Sep, 2020. DOI: 10.1371/jour- nal.pcbi.1008203