Neurofeedback lunch seminar: Maja Djurisic

Maja Djurisic is a senior research scientist in the Shatz and Rubin laboratories. She will presents her work on how deep learning enables discoveries of dendritic spine biology

Dendritic spines are structural correlates of excitatory synapses that are modulated by experience and act as an anatomical substrate of memory. As such, they are also affected in different neurological conditions. In spite of their significance, analysis of dendritic spines remains laborious, time consuming, and mostly limited to manual counts of individual spines. Here we report a fully automated platform for segmentation and detection of dendritic spines in 3D confocal volumes of fluorescently labeled tissue. The automation entails segmentation of dendritic spine heads, spine necks, and whole spines, outputting 3D volumetric masks for the three classes, which enables measurements of spine head volumes and neck lengths as proxies of synaptic strength. Counting of individual spines is automatically downstream of 3D mask output. Our method outperforms known image analysis software used in medical or microscopy image analysis. Its performance is validated against previously published data on dendritic spine plasticity in mouse visual cortex. The same dataset is further explored to discover new plasticity- and sex-dependent properties of dendritic spines in primary visual cortex of adult mice. We hope that our automated platform will serve as a springboard for the deeper analysis of dendritic spines that will further our understanding of how these structures change with normal experience and in disease..