Brain activity relies on complex interactions among large populations of neurons located in different brain regions. Nowadays, studying the role of intricate neuronal networks in a living organism cannot be separated from the use of optical methods. These methodologies, which include fluorescence microscopy and optogenetic intervention, by exploiting the use of light, enable the long-term noninvasive investigation of large populations of neurons in the intact vertebrate brain with high spatio-temporal resolution. As a counterpart, this has dramatically increased the data throughput, stressing the need for automated decision support methods for scientists.
In the “3D Autonomous 2-PhoTon Optogenetics MICroscopy (DAPTOMIC)” project, we devised a microscope system able to autonomously “read” and “write” neuronal cell activity by means of a machine learning-based software module. In detail, we integrated a custom light-sheet microscope with a precise custom light-targeting system, both separately developed by our group during the previous ERC Advanced Grant “BrainBIT”. While the light-sheet microscope allows to image the neuronal activity of the entire larval zebrafish brain at high spatio-temporal resolution, the innovative light targeting system employs two couples of acousto-optic deflectors (AODs) to optogenetically activate multiple neurons dislocated in different positions inside the brain volume with extreme spatial accuracy and consistent targeting rates close to the megahertz range. Owing to a fine tuning of the AODs control system, we significantly improved the homogeneity of the light stimulation energy over the addressed volume, thus solving an inherent drawback which so far limited the use of these devices in optogenetic applications. In order to control the microscope we devised a new dedicated software module for closing the loop between live brain imaging and optogenetic actuation. Through an unsupervised machine learning method, within a few seconds following image acquisition, the software identifies the neuronal coordinates to be stimulated according to the experimental goals previously set by the researcher. The control software can both provide the novel optogenetic targets to the experimenter or directly proceed with the experiment. With DAPTOMIC we moved a decisive step towards automation in microscopy which will help scientists save time and resources, and thus increase the efficacy and efficiency of their mission.
