Skip to main content
U.S. flag

An official website of the United States government

Official websites use .gov
A .gov website belongs to an official government organization in the United States.

Secure .gov websites use HTTPS
A lock ( ) or https:// means you’ve safely connected to the .gov website. Share sensitive information only on official, secure websites.

The Detection of Neuronal Structures using a Patch-based Multi-features and Support Vector Machines Learning Algorithm



Saadia S. Iftikhar, Afzal A. Godil


In this research, we present an (semi-) automated segmentation algo-rithm using a Support Vector Machine classifier. This algorithm uses a high-dimensional feature space generated from patch-based multiple-features and a training dataset in order to detect neuronal structures in electron microscopy images of brain tissue. One such problem of deli-neating the boundaries of Endothelial cell cytoplasm has already been presented in [1], but the problem in the data given is many-folds more difficult. In this data, we have to account for the topological neuronal structures as well as the boundaries of neurons along with the small elements in the images, such as mitochondria. Sometimes, the bounda-ries are either missing or badly affected; due to noise and/or other arti-facts. To address these problems, we modified the algorithm in [1] and introduced some new features, such as distance transform, obtained from each image in the given data in the existing features vector. A set of 34 distinct features are extracted for each pixel in the image and mapped into 34 dimensional feature spaces for the segmentation of boundaries and background. Some threshold conditions are later applied to remove small regions which are below than the certain threshold criteria and morphological operations, such as the filling of the detected objects, is done to get compactness in the objects. The performance of the segmentation method is calculated on the unseen data by using three distinct error measures: pixel error, minimum splits and mergers wrapping error and rand error as explained in [2] with their respective ground-truth. The trained SVM classifier achieves the best precision level in these three distinct errors at 0.230157983, 0.016604996 and 0.153726536 respectively on the given test dataset. Although, statistically, the results produced are not highly significant but it is one step towards exploring possible ways to solve hard-problems, like segmentation, in medical image analysis.
Proceedings Title
2."Segmentation of neuronal structures in EM stacks
Conference Dates
May 31, 2012
Conference Location
Barcelona, ES
Conference Title
The IEEE International Symposium on Biomedical Imaging (ISBI)


Neuronal segmentation, machine learning, image morphology, feature selection, image processing, medical imaging


Iftikhar, S. and Godil, A. (2013), The Detection of Neuronal Structures using a Patch-based Multi-features and Support Vector Machines Learning Algorithm, 2."Segmentation of neuronal structures in EM stacks, Barcelona, ES, [online], (Accessed July 16, 2024)


If you have any questions about this publication or are having problems accessing it, please contact

Created December 23, 2013, Updated October 12, 2021