Xuerong Xiao1, Maja Djurisic2, Assaf Hoogi3, Richard W Sapp2, Carla J Shatz2, Daniel L Rubin3. 1. Department of Electrical Engineering, Stanford University, David Packard Building, 350 Serra Mall, Stanford, CA 94305, USA. Electronic address: xuerong@stanford.edu. 2. Departments of Biology and Neurobiology, and Bio-X, Stanford University, James H. Clark Center, 318 Campus Dr, Stanford, CA 94305, USA. 3. Department of Biomedical Data Science, Stanford University, Medical School Office Building, 1265 Welch Rd, Stanford, CA 94305, USA.
Abstract
BACKGROUND: Dendritic spines are structural correlates of excitatory synapses in the brain. Their density and structure are shaped by experience, pointing to their role in memory encoding. Dendritic spine imaging, followed by manual analysis, is a primary way to study spines. However, an approach that analyses dendritic spines images in an automated and unbiased manner is needed to fully capture how spines change with normal experience, as well as in disease. NEW METHOD: We propose an approach based on fully convolutional neural networks (FCNs) to detect dendritic spines in two-dimensional maximum-intensity projected images from confocal fluorescent micrographs. We experiment on both fractionally strided convolution and efficient sub-pixel convolutions. Dendritic spines far from the dendritic shaft are pruned by extraction of the shaft to reduce false positives. Performance of the proposed method is evaluated by comparing predicted spine positions to those manually marked by experts. RESULTS: The averaged distance between predicted and manually annotated spines is 2.81 ± 2.63 pixels (0.082 ± 0.076 microns) and 2.87 ± 2.33 pixels (0.084 ± 0.068 microns) based on two different experts. FCN-based detection achieves F scores > 0.80 for both sets of expert annotations. COMPARISON WITH EXISTING METHODS: Our method significantly outperforms two well-known software, NeuronStudio and Neurolucida (p-value < 0.02). CONCLUSIONS: FCN architectures used in this work allow for automated dendritic spine detection. Superior outcomes are possible even with small training data-sets. The proposed method may generalize to other datasets on larger scales.
BACKGROUND:Dendritic spines are structural correlates of excitatory synapses in the brain. Their density and structure are shaped by experience, pointing to their role in memory encoding. Dendritic spine imaging, followed by manual analysis, is a primary way to study spines. However, an approach that analyses dendritic spines images in an automated and unbiased manner is needed to fully capture how spines change with normal experience, as well as in disease. NEW METHOD: We propose an approach based on fully convolutional neural networks (FCNs) to detect dendritic spines in two-dimensional maximum-intensity projected images from confocal fluorescent micrographs. We experiment on both fractionally strided convolution and efficient sub-pixel convolutions. Dendritic spines far from the dendritic shaft are pruned by extraction of the shaft to reduce false positives. Performance of the proposed method is evaluated by comparing predicted spine positions to those manually marked by experts. RESULTS: The averaged distance between predicted and manually annotated spines is 2.81 ± 2.63 pixels (0.082 ± 0.076 microns) and 2.87 ± 2.33 pixels (0.084 ± 0.068 microns) based on two different experts. FCN-based detection achieves F scores > 0.80 for both sets of expert annotations. COMPARISON WITH EXISTING METHODS: Our method significantly outperforms two well-known software, NeuronStudio and Neurolucida (p-value < 0.02). CONCLUSIONS:FCN architectures used in this work allow for automated dendritic spine detection. Superior outcomes are possible even with small training data-sets. The proposed method may generalize to other datasets on larger scales.
Authors: Michelle Day; Zhongfeng Wang; Jun Ding; Xinhai An; Cali A Ingham; Andrew F Shering; David Wokosin; Ema Ilijic; Zhuoxin Sun; Allan R Sampson; Enrico Mugnaini; Ariel Y Deutch; Susan R Sesack; Gordon W Arbuthnott; D James Surmeier Journal: Nat Neurosci Date: 2006-01-15 Impact factor: 24.884
Authors: Firdaus Janoos; Kishore Mosaliganti; Xiaoyin Xu; Raghu Machiraju; Kun Huang; Stephen T C Wong Journal: Med Image Anal Date: 2008-07-24 Impact factor: 8.545
Authors: Ganesh M Shankar; Brenda L Bloodgood; Matthew Townsend; Dominic M Walsh; Dennis J Selkoe; Bernardo L Sabatini Journal: J Neurosci Date: 2007-03-14 Impact factor: 6.167
Authors: Maja Djurisic; George S Vidal; Miriam Mann; Adam Aharon; Taeho Kim; Alexandre Ferrao Santos; Yi Zuo; Mark Hübener; Carla J Shatz Journal: Proc Natl Acad Sci U S A Date: 2013-12-03 Impact factor: 11.205
Authors: Yong Zhang; Kun Chen; Matthew Baron; Merilee A Teylan; Yong Kim; Zhihuan Song; Paul Greengard; Stephen T C Wong Journal: Neuroimage Date: 2010-01-25 Impact factor: 6.556
Authors: Konstantinos Kamnitsas; Christian Ledig; Virginia F J Newcombe; Joanna P Simpson; Andrew D Kane; David K Menon; Daniel Rueckert; Ben Glocker Journal: Med Image Anal Date: 2016-10-29 Impact factor: 8.545
Authors: Alfredo Rodriguez; Douglas B Ehlenberger; Dara L Dickstein; Patrick R Hof; Susan L Wearne Journal: PLoS One Date: 2008-04-23 Impact factor: 3.240