AVIRON – Amortized Inference and Virtual Reality to segment large volumes of neuron data

There is currently a worldwide effort to understand the nervous system and the brain’s “algorithmic“ functioning. Progress in microscopy now offers the possibility of imaging large sections of neuronal data in behaving animals. A necessary task is the extraction of local and global neuron connectomes to decipher the emergence of computation and cognition capacities inside these architectures. To address this challenge, the project introduces an innovative software approach, AVIRON, that combines virtual reality (VR), human cognition and inference procedures to allow fast tracing and segmentation of large volumes of neuronal images. We will develop a software based on our current DIVA platform to: (i) integrate any stack of images in VR, (ii) develop an intuitive interface to manipulate data in space and in properties (e.g. contrast, transparency), (iii) create a motion-based segmentation tool combined with an amortized inference procedure allowing live segmenting of the data, and (iv) a post-processing interface to correct anomalies and provide various outputs.

Tribute to Maxime Dahan