AAU logo

Ph.D. Defence by Sofyan Hammad Himaidan Hammad

Sofyan Hammad Himaidan Hammad will defend his Ph.D. thesis: "Intra-cortical Brain Computer interface systems based on advanced digital signal processing techniques"

Time

02.03.2018 kl. 13.00 - 16.00

Description

PROGRAM

Download Program

ABSTRACT

Intra-cortical brain-computer interface (BCI) is a relatively new technology that aims at improving the quality of life of severely disabled persons. The BCI systems work by interpreting acquired brain activity to drive the external devices. Intra-cortical recording methods have the advantage that they acquire highly specific information about a movement with very low latency. The earlier intra-cortical BCI systems traditionally rely on spike-sorting (sorting of single unit action potentials) with a high computational demand to extract information, and have shown to work mainly in restricted and well-controlled environments. This thesis hypothesized that movement intention can be decoded from non-spike sorted, intracortical recordings obtained from freely moving animals.                    
To address this hypothesis four studies were conducted. Rats were instrumented with intra-cortical electrodes, and brain signals were acquired while the animals performed an on/off type hitting task with the forelimb. Study 1 investigated the need for preprocessing and the effect of automatic spike-thresholding on the detection of the movement intention. The need for preprocessing emerges due to the noise present with the multi-unit recordings from freely moving subjects. It was found that a preprocessing step before movement detection was needed, but there was no difference was found between automatic and fixed thresholding. Study 2 evaluated the effect of denoising on the detection accuracy of the movement intention. Wavelet denoising was found to significantly improve the accuracy of movement detection. The aim of study 3 was to compare different detection methods. The results showed that the detection method had no effect on the detection accuracy, which allows the utilization of different detection methods. In Study 1-3, only one feature (the firing rate) was used as a basis for the detection. In Study four, it investigated if combining more features could improve the detection accuracy. Thus, combinations of seven features evaluated over three window lengths were investigated in a simulated real-time experimental setup. The combination of features and short window lengths was found to yield more stable and higher detection accuracies.         
In this thesis movement intention could be detected with an acceptable accuracy from non-spike sorted intracortical recordings obtained in freely moving paradigm. The movement investigated was a simple ‘on-off’ hitting task, but the work indicates that with a careful choice of signal processing methods it is possible to overcome noisy signals and unpredictable environments and thereby design more autonomous intra-cortical BCI systems in the future.