Interdisciplinary Initiatives Program Round 8 - 2016
Daniel L. Rubin, Radiology, Biomedical Data Science, and Medicine (Biomedical Informatics Research)
Carla J. Shatz, Biology and Neurobiology
Synapses are key elements of neurons that enable neuronal communication within the brain, and they play a central role in learning and memory, as well as brain-related disorders. Most excitatory synapses have structures called dendritic spines—protuberances from the dendritic branches of a neuron. Dendritic spines are dynamic, changing in numbers with experience. It is thought that novel experience influences a shape of the spines, increasing the size of the spine head when synaptic activity strengthens, or decreasing the head size when synaptic activity weakens. In pathological conditions, excessive synaptic weakening can result in a permanent removal of a spine, a process thought to underlie the massive loss of spines observed in Alzheimer’s, Huntington’s and Parkinson’s diseases, and could explain the cognitive decline seen in these devastating conditions.
The goal of this proposal is to develop computerized methods to automatically detect dendritic spines and to measure their structural properties in an automated fashion, and to leverage machine learning methods in large collections of neuronal images to use this information to discover features of dendritic spines that are associated with disease. We propose a unique interdisciplinary collaboration between the labs of D.L. Rubin and C.J. Shatz to accomplish this goal. These two labs bring together complimentary expertise in quantitative image analysis with the neuroscience of dendritic spines in normal cognition and neurodegenerative disorders. Recent work from the Shatz lab using modern high resolution imaging techniques showed that the numbers and stability of dendritic spines increase in the brain with a novel experience, thus representing a structural correlate of memory. New dendritic spines are also needed for the recovery from cortical blindness induced by prolonged visual deprivation, further underscoring the role of dendritic spines in diseased states of the brain. Tremendous amounts of imaging data are generated in this lab as well as other labs internationally that need to be comprehensively analyzed to thoroughly characterize all the dendritic spines, including their density, shape, and dimensions. Most of these analyses are being done by hand (outlining and counting each dendritic spine), which is very laborious, error-prone, and not practical given the explosion in the number of studies and by the volumetric character of these data. Advances in medical image processing techniques being developed in Dr. Rubin’s lab can be applied to volumetric neuronal images being obtained in Dr. Shatz’s lab to overcome the challenge of fully automated extraction and characterization of dendritic spines. In this proposed collaboration, we will (1) Develop automated methods for detection, segmentation, and characterization of dendritic spines in 3D microscopic images of neurons, (2) Validate the reliability and accuracy of our automated methods, and (3) Apply our automated methods to a neuroscience problem: to undertake in-depth characterization of dendritic spine shape and size in the context of ocular dominance plasticity in the visual system. Our methods will enable acquiring and studying many novel characteristics of dendritic spines in large scale, and they will promote “data driven discovery” to more accurately characterize neurological diseases. Our work strongly fits the Bio-X mission of stimulating novel and impactful interdisciplinary collaborations and catalyzing discovery in neuroscience. Our Bio-X supported work will also lead to NIH grants on new scientific questions and produce new tools valuable to the community for studies of important neurobiological questions.
