Filtering and decoding for invasive neural interfaces


As of 1999, more than 3 million people suffered from physical disabilities in Japan, according to Annual Reports on Health and Welfare of 1998-1999 [1]. In Brazil, numbers are even worse: more than 10 million people had declared to suffer from some kind of motor difficulty [2]. Different approaches from several fields have been developed along the years in order to attend to these people’s necessities. Among them, one in particular has shown promising results – the usage of brain signals to aid patients in the task of interacting with the world.

Even though this notion has matured enough to breed all sorts of researches, it has become clearer and clearer that the usage of invasive electrodes guarantee better signal quality and, therefore, results. Meanwhile, works involving multiple channels for measurement have become mandatory in order to acquire more information. One of the approaches that combined both qualities the best is in the spotlight and is the main topic of this proposed research: Electrocorticograms (ECoGs) combined with Utah Slanted Electrode Arrays (USEAs).

Overall, the “Utah electrodes” stand above other electrodes when considering its three-dimensional configuration. While other electrodes deal very well with longitudinal acquisition in nerves (such as “LIFE”) or even with it transversally (e.g. “TIME”), USEA takes advantage of its slanted array-shaped setup to acquire it in both planes. In our case, however, we will be implanting USEAs in the cortical area, where we will benefit of its three-dimensionality when acquiring signals from multiple neurons within the same functional area and, simultaneously, sensing different depths of the cortex, which are directly related to signal transmission to other brain structures.

Replace this text with your caption

In order to use USEAs electrodes to its fullest, several mathematical and signal-processing tools are usually implemented. Firstly, it is interesting to have the signal filtered in order to avoid unwanted interferences, such as electromagnetic, communication, etc. The signal, then, is interpreted in order to acquire when exactly an “action potential” (or “spike”) occurs; further on, spike frequency is measured. Followingly, signal redundancy and correlation analysis are done in order to deal with the great number of electrodes within the array. This reduced-complexity information is lastly used in decoding algorithms, which can be both discrete or continuous. The main goal of this step, however, is to understand what different spike rates perceived from different group of neurons mean in terms of functionality. For example, specific neurons may increase their firing rate when the patient moves an object from inside out in a specific angle. For different angles, different groups of neurons are more frequently activated; thus, the algorithm must understand which angle is being used for a specific array reading.