With the rapid development of neurotechnology in the past five years, the number of neurons we can record or modulate simultaneously reaches an unprecedented level, which provides a great opportunity for neuroscientists to understand the signal processing mechanisms at the mesoscale level spanning from individual neurons to neural circuits across multiple brain regions. Profound discoveries have been made including perceptual input processing, motion, learning, memory, and decision making. However, the orders-of-magnitude improvement of the data throughput in mesoscale neural recording also pose great challenges in existing methods of imaging, analysis, and modeling, such as complicated bulky imaging systems, low-fidelity temporal traces, high computational costs, extremely high-dimensional data with limited trials, integration of multimodality data, and so on.
These problems become increasingly critical with gradually understanding of the distributed neural networks at a systematic level. Computational imaging systems with spatiotemporal multiplexing and new fluorescence indicators can increase the limited space-bandwidth product of traditional microscopy with reduced signal crosstalk, while existing methods are still quite expensive or complicated, which cannot be generally accessible for the broad biology community. Deep-learning-based algorithms cannot only enhance the imaging performance with better resolution or signal-to-noise ratio, but also facilitate efficient large-scale data analysis such as motion registrations and calcium extractions. Larger datasets and better evaluation metrics are required to guarantee the robust and reliable performance of deep neural networks in diverse imaging modalities and environments. Unsupervised learning may be a good choice, since it’s usually hard to obtain the ground truth data for functional imaging. More is different, and more is also hard to understand. Mesoscale data at single neuron resolution obtained either by microscopes or electrophysiology usually has a much higher dimensionality than the data of clustering of neurons. With the limited number of trials for stimulus, new models or methods are required to infer the mesoscale functional network connectivity between the neurons to reconstruct the calculating strategy the neural circuit uses to accomplish specific signal processing missions.
For this research topic, we focus on both hardware and software towards the recording, analysis, and modeling of mesoscale neural activities, which may promote the systematic understanding of signal processing procedures in neural networks. We welcome submissions of Original Research, Brief Research Reports, reviews, Mini-Reviews and Opinions on the following topics:
- High-throughput imaging systems or electrophysiology
- Algorithms or datasets for large-scale high-fidelity extraction of the neural activities
- Evaluation metrics or datasets for the data fidelity
- Evaluation metrics or datasets for the data-fidelity
- Algorithms for functional mapping or functional circuit identification
- Models for brain-wide functional network at single neuron level
- Network analysis techniques to find more fundamental arithmetic units or their hierarchical structures of neural network.
With the rapid development of neurotechnology in the past five years, the number of neurons we can record or modulate simultaneously reaches an unprecedented level, which provides a great opportunity for neuroscientists to understand the signal processing mechanisms at the mesoscale level spanning from individual neurons to neural circuits across multiple brain regions. Profound discoveries have been made including perceptual input processing, motion, learning, memory, and decision making. However, the orders-of-magnitude improvement of the data throughput in mesoscale neural recording also pose great challenges in existing methods of imaging, analysis, and modeling, such as complicated bulky imaging systems, low-fidelity temporal traces, high computational costs, extremely high-dimensional data with limited trials, integration of multimodality data, and so on.
These problems become increasingly critical with gradually understanding of the distributed neural networks at a systematic level. Computational imaging systems with spatiotemporal multiplexing and new fluorescence indicators can increase the limited space-bandwidth product of traditional microscopy with reduced signal crosstalk, while existing methods are still quite expensive or complicated, which cannot be generally accessible for the broad biology community. Deep-learning-based algorithms cannot only enhance the imaging performance with better resolution or signal-to-noise ratio, but also facilitate efficient large-scale data analysis such as motion registrations and calcium extractions. Larger datasets and better evaluation metrics are required to guarantee the robust and reliable performance of deep neural networks in diverse imaging modalities and environments. Unsupervised learning may be a good choice, since it’s usually hard to obtain the ground truth data for functional imaging. More is different, and more is also hard to understand. Mesoscale data at single neuron resolution obtained either by microscopes or electrophysiology usually has a much higher dimensionality than the data of clustering of neurons. With the limited number of trials for stimulus, new models or methods are required to infer the mesoscale functional network connectivity between the neurons to reconstruct the calculating strategy the neural circuit uses to accomplish specific signal processing missions.
For this research topic, we focus on both hardware and software towards the recording, analysis, and modeling of mesoscale neural activities, which may promote the systematic understanding of signal processing procedures in neural networks. We welcome submissions of Original Research, Brief Research Reports, reviews, Mini-Reviews and Opinions on the following topics:
- High-throughput imaging systems or electrophysiology
- Algorithms or datasets for large-scale high-fidelity extraction of the neural activities
- Evaluation metrics or datasets for the data fidelity
- Evaluation metrics or datasets for the data-fidelity
- Algorithms for functional mapping or functional circuit identification
- Models for brain-wide functional network at single neuron level
- Network analysis techniques to find more fundamental arithmetic units or their hierarchical structures of neural network.
