Are you an EPFL student looking for a semester project?
Work with us on data science and visualisation projects, and deploy your project as an app on top of Graph Search.
Neural decoding of the visual system is a subject of research interest, both to understand how the visual system works and to be able to use this knowledge in areas, such as computer vision or brain-computer interfaces. Spike-based decoding is often used, but it is difficult to record data from the whole visual cortex, and it requires proper preprocessing. We here propose a decoding method that combines wide-field calcium brain imaging, which allows us to obtain large-scale visualization of cortical activity with a high signal-to-noise ratio (SNR), and convolutional neural networks (CNNs). A mouse was presented with ten different visual stimuli, and the activity from its primary visual cortex (V1) was recorded. A CNN we designed was then compared with other existing commonly used CNNs, that were trained to classify the visual stimuli from wide-field calcium imaging images, obtaining a weighted F1 score of more than 0.70 on the test set, showing it is possible to automatically detect what is present in the visual field of the animal.
Silvestro Micera, Simone Romeni, Laura Toni, Fiorenzo Artoni
Pascal Frossard, Chenglin Li, Li Wei, Qin Yang, Yuelei Li, Hao Wang