Scientists at the Max Planck Florida Institute for Neuroscience (MPFI), in collaboration with ZEISS and MetaCell, have developed a new imaging technique that maps brain activity at an unprecedented level of detail. The method, called Neuroplex, makes it possible to simultaneously track the activity of up to nine different types of brain cells in freely moving mice.
This innovation accelerates research into how the brain controls behavior and offers new opportunities to better understand complex neural networks.
Limitations of existing techniques
Until now, neuroscientists have faced a major limitation. So-called miniscopes, tiny microscopes placed on the heads of laboratory animals, can measure brain activity during behavior, but they barely distinguish between different cell types.
According to lead researcher Mary Phillips, this was a major obstacle: “To understand behavior, we need to know which specific neurons are active. But with existing techniques, we could usually distinguish only two cell types at a time.”
As a result, researchers were forced to repeat experiments multiple times, each time with a different labeled cell type. This made studies time-consuming, costly, and less reliable due to variations between individual animals. Alternative methods, such as analyzing brain tissue post-mortem, did offer more detail, but made it impossible to track changes in brain activity over time.
Neuroplex
Neuroplex overcomes these limitations by combining two complementary imaging techniques in the same living animal. Researchers first label different neuronal populations with unique fluorescent color codes. Next, brain activity is recorded using a miniscope while the animal moves freely. The same brain region is then imaged again using a confocal microscope, such as the ZEISS LSM 980, which can distinguish the different color codes.
Using a specially developed analysis tool, both datasets are precisely aligned. This allows researchers to determine exactly which type of neuron is responsible for specific activity patterns. According to co-researcher Zhe Dong, this combination makes it possible to reliably analyze complex data and ensure reproducibility. The results of the study have been published in eLife.
Nine neural circuits
As a proof-of-principle, the team focused on the medial prefrontal cortex, a brain region involved in decision-making. They labeled nine different neural circuits and tracked their activity during social behaviors, such as sniffing and interacting with other mice.
The results are impressive: approximately 75% of the active neurons could be assigned to a specific cell type, with an accuracy of about 90%. This enables direct comparison between different neural circuits, something that was previously hardly feasible.
Senior researcher Ryohei Yasuda emphasizes that this approach significantly increases the efficiency and reliability of data collection.
Longitudinal research
A key advantage of Neuroplex is that it is applied entirely in live animals. This allows researchers to track the same neurons over extended periods and study changes in activity. This opens the door to longitudinal studies of learning, aging, and disease progression. This can yield valuable insights, particularly regarding neurodegenerative disorders and developmental disorders.
The researchers are now working on further improvements, including making the technique more accessible to laboratories without advanced equipment. By also using standard microscopes, they hope to make Neuroplex widely available.
This innovation brings a better understanding of brain functions and diseases one step closer, with the potential for a major impact on both basic research and future medical applications.
BrainAC
In March, researchers from Mass General Brigham and Harvard Medical School developed BrainIAC, a new AI foundation model that can predict multiple brain disorders using MRI scans. Unlike traditional AI systems designed for a single condition, BrainIAC was trained on nearly 49,000 brain MRI scans and can be adapted for a wide range of clinical applications. Using self-supervised learning, the model learned brain structures from mostly unannotated data, enabling it to recognize patterns linked to diseases such as Alzheimer’s, Parkinson’s, autism spectrum disorder, stroke, dementia, and brain tumors.
Researchers found that BrainIAC often required up to ten times less training data than specialized AI models while achieving similar accuracy. Published in Nature Neuroscience, the open-source model could support broader AI use in neurological research and potentially improve early detection of brain disorders in clinical practice.
