Developing EZspine for Advanced Dendritic Spine Analysis

In Fall 2024, the Neuroscience team made significant strides in advancing dendritic spine analysis through the development of EZspine, a novel computer vision pipeline designed to automate the classification of dendritic spines. With neurodegenerative diseases such as Alzheimer’s, Parkinson’s, and Huntington’s affecting millions worldwide, understanding dendritic spine morphology and its association with neural communication is critical. Traditional methods for quantifying spine morphology are labor-intensive, prone to human error, and limited in their applicability to specialized imaging modalities like confocal microscopy. The EZspine software addresses these challenges by providing a fully automated solution that reliably processes diverse image types, including Golgi-stained and low-quality images. The MVP of EZspine, completed this semester, employs two primary approaches. The first constructs 3D models of neurons to isolate spines, capturing rich structural data. The second, computationally less expensive approach, bypasses 3D modeling by isolating spines from flattened, 2D Z-stacks. The current MVP is based on the second approach and is now undergoing evaluation by researchers at the Miami Project to Cure Paralysis. Key innovations include algorithms for automatically generating 3D and 2D neuron representations, a GUI for visualizing and modifying neuron models, and methods for summarizing and classifying dendritic spine morphology. These tools are set to enhance the accuracy and efficiency of spine analysis, with a manuscript detailing the methods and findings expected to be submitted for publication in early 2025.