Automated three-dimensional detection and shape classification of dendritic spines from fluorescence microscopy images.
Rodriguez A, Ehlenberger DB, Dickstein DL, Hof PR, Wearne SL. · 2008
Read Automated three-dimensional detection and shape classification of dendritic spines from fluorescence microscopy images. by Rodriguez A, Ehlenberger DB, Dickstein DL, Hof PR, Wearne SL. (2008) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc2292261.txt.
Introduction
Rodriguez, Ehlenberger, Dickstein, Hof and Wearne built an automated three-dimensional system in 2008 for finding and classifying dendritic spines in microscope images. Spines are tiny protrusions where excitatory synapses form, and their shapes relate to learning, memory and brain disease. Measuring them precisely in 3D was slow and low-throughput. Their software module, part of NeuronStudio, processes laser scanning microscopy volumes automatically, using adaptive thresholding, voxel clustering and a sampling technique to estimate spine diameters.
The system proved more accurate than existing tools and at least ten times faster. Working fully in three dimensions, it resolves spines that two-dimensional tools miss entirely. Diameter profiles classify spines into shape types while limiting optical blur and measurement artefacts. The technique enables objective studies of spine change during learning, development, ageing and neurodegenerative disease. Large volumetric datasets can finally support robust statistics on spine morphology.
Catalog record
| CREATOR | Rodriguez A, Ehlenberger DB, Dickstein DL, Hof PR, Wearne SL. |
|---|---|
| YEAR | 2008 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 58 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Automated Three-Dimensional Detection and Shape Classification of Dendritic Spines from Fluorescence Microscopy Images By Alfredo Rodriguez, Douglas B Ehlenberger, Dara L Dickstein, Patrick R Hof, Susan L Wearne, Bernado Sabatini Published in: PLoS ONE DOI: 10.1371/journal.pone.0001997 ABSTRACT Abstract A fundamental challenge in understanding how dendritic spine morphology controls learning and memory has been quantifying three-dimensional (3D) spine shapes with sufficient precision to distinguish morphologic types, and sufficient throughput for robust statistical analysis. The necessity to analyze large volumetric data sets accurately, efficiently, and in true 3D has been a major bottleneck in deriving reliable relationships between altered neuronal function and changes in spine morphology. We introduce a novel system for automated detection, shape analysis and classification of dendritic spines from laser scanning microscopy (LSM) images that directly addresses these limitations. The system is more accurate, and at least an order of magnitude faster, than existing technologies. By operating fully in 3D the algorithm resolves spines that are undetectable with standard two-dimensional (2D) tools. Adaptive local thresholding, voxel clustering and Rayburst Sampling generate a profile of diameter estimates used to classify spines into morphologic types, while minimizing optical smear and quantization artifacts. The technique opens new horizons on the objective evaluation of spine changes with synaptic plasticity, normal development and aging, and with neurodegenerative disorders that impair cognitive function. INTRODUCTION Because of their capacity for plasticity and their key role as the location of excitatory synapses in the cerebral cortex, accurately characterizing the structure of dendritic spines is of profound biological significance. Spine morphology determines the strength, stability and function of excitatory synaptic connections that subserve the neuronal networks underlying cognitive function. Precise quantification of spine morphology in three dimensions (3D) is essential to understanding the structural determinants of normal neuronal function, its development, plasticity, and its dysfunction in neurodegenerative disorders. To date however, this level of precision has been restricted to time-intensive electron microscopy (EM) approaches with limited throughput, that are impractical for comparative population studies. In this paper we introduce a novel computational approach for detection and shape analysis of neuronal dendritic spines from confocal and multiphoton laser scanning microscopy (CLSM and MPLSM) images, that operates fully in 3D, and is faster and more accurate than existing semi-automated technologies. The algorithm is a module of our NeuronStudio software application [1], an integrated system for semi-automated digitization, morphometry and analysis of complex neuronal morphology at high resolution. Spines come in multiple shapes and sizes, which subserve a diversity of function [2]–[6]. Although spine shapes in fixed tissue form a continuum rather than distinct categories, broad subclasses (e.g., thin, stubby, mushroom) have traditionally been distinguished on the basis of morphology (see refs [3], [7] for review). Spine morphology also varies dynamically in response to synaptic activity [8]–[13]. Smaller spines are less stable and more motile [14]–[16], and as a result, more plastic than large spines [17]. Recent data indicate that the size and morphology of the spine head are correlated with numbers of docked presynaptic vesicles [18], numbers of postsynaptic receptors [19], and hence with synaptic strength. From a biophysical viewpoint, these effects give rise to increased synaptic currents and reduced time constants for calcium compartmentalization in larger spine heads (see refs [5], [20] for review), modulating postsynaptic mechanisms that determine functions such as learning and memory [21]–[24]. Spine neck length and diameter also affects diffusional coupling between dendrite and spine [25]–[28], and spine density and shape regulate the degree of anomalous diffusion of chemical signals within the dendrite [29]. As more precise data from these studies emerges, the need for accurate spine morphometry in true 3D, and on large enough scales for robust statistical analyses becomes increasingly critical. Digital representation of neuronal morphology using light microscopy has traditionally relied on manual tracing from a computer screen [30], [31], and is prone to subjective errors. Despite the recent introduction of semi-automated tracing methods (e.g., AutoNeuron, MBF Bioscience, Williston, VT), the problem of detecting and characterizing spine shapes automatically, in 3D, remains unsolved. In existing image analysis-based tools, precision at the finest scales is limited by the skeletonization methods used and by quantization errors. Even when imaged at the limits of CLSM resolution, single spines may span as few as 3–10 voxels, and neck diameters may be subvoxel resolution. Accurate recovery of spine geometry thus requires new analysis algorithms capable of using subvoxel information, and the full three dimensions of structural information from a LSM image stack. The spine detection and analysis method presented here is compatible with many methods for tracing the dendritic tree and thus can be implemented as an add-on module to existing manual or automated tracing packages. Spine voxels are clustered based on connectivity, analyzed with subvoxel precision using Rayburst-based shape analysis routines [1], and classified into three morphologic types, mushroom, stubby and thin. Procedures for declumping of merged spines and for combining detached spine heads with their stems are implemented to ensure accurate spine counts. The method is at least an order of magnitude faster than previous algorithms, and as a 64-bit application, can handle multi-gigabyte datasets. These algorithms have been tested and validated in our freely distributed software application, NeuronStudio [1], [32] (http://www.mssm.edu/cnic), which provides visual verification of spine classification and the ability for manual spine editing through interactive 2D and 3D displays. RESULTS Image Segmentation and Voxel Processing The spine analysis module described here utilizes a previously computed model of the dendritic tree, comprising a series of nodes of specified diameter forming conical frusta (Fig. 1A,B), which could be obtained by any of several existing methods [33]–[37], and which is compatible with many existing neuronal morphometry applications [32], [38]. Because fluorescence intensity can vary with adequacy of filling, imaging depth and XY spatial extent in CLSM and MPLSM image stacks, segmentation based on a globally selected intensity threshold is in general not feasible, and requires a dynamically adjusted, local threshold. [Figure: Processing of Candidate Voxels for Spine Detection.A) Dendritic model rendered as green balls superimposed on volume-rendered data (light grey voxels), augmented by the maximum spine height (MSH, dark grey envelope surrounding dendrites that contains all spine candidate voxels). Four corners of the axis-aligned bounding box (AABB) are shown, with the length of each axis segment representing 10 µm to depict the scale in 3D. B) Close-up view of a spiny dendritic section, volume rendered in grayscale, with superimposed 3D model formed by sequential green frustra. A single capsule formed by a cylindrical dendritic segment capped by two hemispheres at consecutive nodes is outlined in blue. Spine candidate voxels V1 and V2 are shown with their corresponding distance to surface (DTS) values (red arrows). The dark grey envelope around the dendrite represents the MSH, measured from the surface of the 3D model, as indicated by the white dashed line. C) Octree calculated for the top fork of the dendritic section shown in (A). The root node is outlined in red. Recursive subdivision of the 3D space results in increasingly smaller cubes surrounding the model. The inset on the right illustrates bulk rejection of voxels in a single leaf (red shaded cube) that does not intersect the MSH zone (grey envelope surrounding model) created by a small dendritic section. D) Volume-rendered spiny dendrite with voxels representing