Scientists at Helmholtz Munich and the LMU University Hospital Munich have introduced DELiVR which, they say, offers a new AI-based approach to the complex task of brain cell mapping. The deep learning tool reportedly democratizes advanced neuroscience by eliminating the need for coding expertise. DELiVR was designed to allow biologists to investigate disease-related spatial cell dynamics efficiently, with the goal of developing precision therapies for enhanced patient care.
The researchers discuss their work “Virtual reality-empowered deep-learning analysis of brain cells” in Nature Methods.
Automated detection of specific cells in three-dimensional datasets such as whole-brain light-sheet image stacks is challenging. Here, we present DELiVR, a virtual reality-trained deep-learning pipeline for detecting c-Fos+ cells as markers for neuronal activity in cleared mouse brains. Virtual reality annotation substantially accelerated training data generation, enabling DELiVR to outperform state-of-the-art cell-segmenting approaches,” write the investigators.
“Our pipeline is available in a user-friendly Docker container that runs with a standalone Fiji plugin. DELiVR features a comprehensive toolkit for data visualization and can be customized to other cell types of interest, as we did here for microglia somata, using Fiji for dataset-specific training. We applied DELiVR to investigate cancer-related brain activity, unveiling an activation pattern that distinguishes weight-stable cancer from cancers associated with weight loss. Overall, DELiVR is a robust deep-learning tool that does not require advanced coding skills to analyze whole-brain imaging data in health and disease.”
Many diseases are linked with changes in the expression of certain proteins in the brain. To study those changes, scientists examine how they change during disease progression in model organisms. Imaging entire mouse brains generates vast data sets, necessitating accurate quantification methods for its meaningful interpretation.
However, identifying labeled cells within large 3D image data is challenging, note the researchers, who add that while artificial intelligence (AI) holds promise for data analysis, it typically requires extensive data annotation and advanced coding skills, restricting its use to specialized labs. The German team focused on overcoming these barriers and democratizing 3D analysis for broader scientific access.
Virtual reality empowering researchers
To accurately quantify specific cells within brain images, the scientists team initially trained an AI algorithm to identify them in 3D microscopic images. Leveraging virtual reality (VR) for label generation, the researchers immersed themselves in the images, annotating cells directly in 3D, a faster and more precise method than traditional 2D slice-based approaches.
Subsequently, the team employed these VR-generated labels to train an AI algorithm for the automatic identification of active neurons. They integrated the processes of detecting cells, matching them to a brain atlas, and visualizing the results into their DELiVR (Deep Learning and Virtual Reality mesoscale annotation) pipeline.
The system operates seamlessly with Fiji, an open-source software for image analysis, in an end-to-end workflow. DELiVR also features a customizable function allowing researchers to train it for specific cell types, such as microglia, an essential immune cell in the brain, demonstrating its adaptability for diverse research projects.
“In essence, DELiVR offers a seamless solution for identifying and analyzing cells throughout the entire brain, providing invaluable insights into their roles and behaviors in both health and disease–all without requiring coding expertise from scientists,” explains Ali Erturk, PhD, who led the development of the tool at Helmholtz. “DELiVR represents a step towards developing new therapeutic interventions that could ultimately improve the quality of life for individuals affected by debilitating conditions.”
To demonstrate the power of DELiVR, the research team exemplified its capacity to transform our comprehension of how cancer influences our brain activity. Concentrating on the significant clinical challenge of tumor-induced weight loss, they discovered specific brain activity patterns distinguishing cancers that induce weight loss in mice from those that do not.
“Our findings using DELiVR have revealed potential therapeutic targets within brain regions. This might pave the way for promising strategies to combat cancer-related weight loss in the future,” says Doris Kaltenecker, PhD, a first author of the study.
