SynAPSeg: A novel dataset and image analysis framework for deep learning-based synapse detection and quantification

Area of Science:

• Neuroscience
• Computational Neuroscience
• Biophysics

Background:

• Quantifying synaptic organization at circuit scales is a major challenge in neuroscience.
• Current deep learning tools often fail to segment synaptic puncta in dense tissues.
• There is a need for robust, scalable analysis methods for synaptic architecture.

Purpose of the Study:

• To introduce SynAPSeg, an open-source deep learning framework for synaptic puncta segmentation and analysis.
• To develop and validate deep learning models for accurate synaptic puncta quantification.
• To enable large-scale mapping of synaptic organization and study synaptic changes in aging.

Main Methods:

• Developed SynAPSeg, an open-source deep learning framework.
• Created the first large-scale, publicly available instance segmentation dataset for synaptic puncta.
• Trained deep learning models achieving expert-level performance on benchmark datasets.
• Integrated models into an interactive interface supporting multi-dimensional data and automated pipelines.
• Applied SynAPSeg to map millions of PSD95 puncta in the hippocampus and analyze aging-related synaptic changes in CA1 parvalbumin neurons.

Main Results:

• SynAPSeg models achieved expert-level performance in synaptic puncta segmentation.
• A comprehensive mapping of nearly 4 million PSD95 puncta in dorsal hippocampus inhibitory interneurons revealed regional differences.
• Analysis of aged CA1 parvalbumin neurons showed a reduction in PSD95 density, indicating impaired glutamatergic recruitment.
• The framework demonstrated scalability for large-scale synaptic architecture analysis.

Conclusions:

• SynAPSeg provides a scalable, automated solution for synaptic puncta segmentation and quantification.
• The framework facilitates large-scale synaptic mapping and the study of synaptic plasticity in health and disease.
• Findings suggest that reduced PSD95 density in aged CA1 PV neurons may contribute to cognitive decline.