Institute of Cognitive Science,
49090 Osnabrück, Germany
Improving the Signal Quality of a Mobile EEG Device with Deep Learning
For me, neuroinformatics provides the opportunity to research the human brain and to better understand it. Using this knowledge to develop and improve e.g. deep learning applications that further promote the understanding of healthy or abnormal brain activity creates a fascinating circle of research areas. Furthermore, I am interested in different ways to communicate and share our research and knowledge. This includes teaching university classes and workshops for middleschool students, as well as our recently created Youtube Channel
The goal is to develop the DreamMachine (Traumschreiber), a low budget, mobile EEG device that transmits data to Android devices via Bluetooth. The DreamMachine should be easy to use for non-expert users. Possible applications are sleep research at home or epilepsy detection in areas with low medical infrastructure. In my PhD, the focus lies on the software side and the receiving Android companion app. The recorded signal should then be analysed to expose its deficiencies. This might be artefacts or a low spatial or temporal resolution. Afterwards, Neural Networks should be trained to improve weak points of the signal. For example, Nejedly et al. trained a Convolutional Neural Network to detect different types of artefacts . Yang et al. trained an Autoencoder to simulate lost data points . This can improve the signal when many artefacts disturb the signal or data points get lost in transmission. Another idea to improve potential shortcomings of the Traumschreiber is to simulate channels by means of a Generative Adversarial Network as proposed by Corley und Huang . After collecting and producing appropriate training data and training the selected Neural Networks, they should be incorporated in the companion app to ensure the best possible EEG signal with the DreamMachine.
: Nejedly, Petr, et al. "Intracerebral EEG artifact identification using convolutional neural networks." Neuroinformatics 17.2 (2019): 225-234.
: Yang, Banghua, et al. "Automatic ocular artifacts removal in EEG using deep learning." Biomedical Signal Processing and Control 43 (2018): 148-158.
: Corley, Isaac A., and Yufei Huang. "Deep EEG super-resolution: Upsampling EEG spatial resolution with generative adversarial networks." 2018 IEEE EMBS International Conference on Biomedical & Health Informatics (BHI). IEEE, 2018.
L. Krieger, G. Heidemann and J. Schöning, "Object of Interest Segmentation in Video Sequences with Gaze Data," 2018 IEEE International Conference on Image Processing, Applications and Systems (IPAS), 2018, pp. 104-109, doi: 10.1109/IPAS.2018.8708873. (not RTG related)