Limits...
Automated reconstruction of neuronal morphology based on local geometrical and global structural models.

Zhao T, Xie J, Amat F, Clack N, Ahammad P, Peng H, Long F, Myers E - Neuroinformatics (2011)

Bottom Line: Digital reconstruction of neurons from microscope images is an important and challenging problem in neuroscience.We first formulate a model structure, then develop an algorithm for computing it by carefully taking into account morphological characteristics of neurons, as well as the image properties under typical imaging protocols.The method has been tested on the data sets used in the DIADEM competition and produced promising results for four out of the five data sets.

View Article: PubMed Central - PubMed

Affiliation: Qiushi Academy for Advanced Studies, Zhejiang University, 38 ZheDa Road, Hangzhou 310027, China. tingzhao@gmail.com

ABSTRACT
Digital reconstruction of neurons from microscope images is an important and challenging problem in neuroscience. In this paper, we propose a model-based method to tackle this problem. We first formulate a model structure, then develop an algorithm for computing it by carefully taking into account morphological characteristics of neurons, as well as the image properties under typical imaging protocols. The method has been tested on the data sets used in the DIADEM competition and produced promising results for four out of the five data sets.

Show MeSH
Conjugate gradient descent reaches the optimal score faster than steepest gradient descent
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC3104133&req=5

Fig3: Conjugate gradient descent reaches the optimal score faster than steepest gradient descent

Mentions: One could use a greedy search to optimize the four free parameters (two orientation parameters and two scale parameters) of the template when it is placed at a given point in space. But this is less appropriate for such a large number of parameters because search time increases exponentially in the number of parameters. If the score function is smooth and has a single peak, gradient descent is certainly preferred. Fortunately, our scoring function does have this property in a fairly large zone of the parameter space around the true optimum. Moreover, the symmetry of the function makes it well fit by a quadratic function within this zone, so as long as the template is placed in a reasonable initial position (which could for example be determined by a coarse discrete sampling of the parameter space), one can use the conjugate gradient descent method, which has been shown to be the best for optimizing general quadratic functions. In our implementation, we used the Polak–Ribière conjugate gradient descent method (Wright and Nocedal 2006). Results showed that it is significantly better than steepest gradient descent (e.g. Fig. 3).Fig. 3


Automated reconstruction of neuronal morphology based on local geometrical and global structural models.

Zhao T, Xie J, Amat F, Clack N, Ahammad P, Peng H, Long F, Myers E - Neuroinformatics (2011)

Conjugate gradient descent reaches the optimal score faster than steepest gradient descent
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC3104133&req=5

Fig3: Conjugate gradient descent reaches the optimal score faster than steepest gradient descent
Mentions: One could use a greedy search to optimize the four free parameters (two orientation parameters and two scale parameters) of the template when it is placed at a given point in space. But this is less appropriate for such a large number of parameters because search time increases exponentially in the number of parameters. If the score function is smooth and has a single peak, gradient descent is certainly preferred. Fortunately, our scoring function does have this property in a fairly large zone of the parameter space around the true optimum. Moreover, the symmetry of the function makes it well fit by a quadratic function within this zone, so as long as the template is placed in a reasonable initial position (which could for example be determined by a coarse discrete sampling of the parameter space), one can use the conjugate gradient descent method, which has been shown to be the best for optimizing general quadratic functions. In our implementation, we used the Polak–Ribière conjugate gradient descent method (Wright and Nocedal 2006). Results showed that it is significantly better than steepest gradient descent (e.g. Fig. 3).Fig. 3

Bottom Line: Digital reconstruction of neurons from microscope images is an important and challenging problem in neuroscience.We first formulate a model structure, then develop an algorithm for computing it by carefully taking into account morphological characteristics of neurons, as well as the image properties under typical imaging protocols.The method has been tested on the data sets used in the DIADEM competition and produced promising results for four out of the five data sets.

View Article: PubMed Central - PubMed

Affiliation: Qiushi Academy for Advanced Studies, Zhejiang University, 38 ZheDa Road, Hangzhou 310027, China. tingzhao@gmail.com

ABSTRACT
Digital reconstruction of neurons from microscope images is an important and challenging problem in neuroscience. In this paper, we propose a model-based method to tackle this problem. We first formulate a model structure, then develop an algorithm for computing it by carefully taking into account morphological characteristics of neurons, as well as the image properties under typical imaging protocols. The method has been tested on the data sets used in the DIADEM competition and produced promising results for four out of the five data sets.

Show MeSH