Universiti Teknologi Malaysia Institutional Repository

A hybrid approach for segmenting and validating T1-weighted normal brain mr images by employing ACM and ANN

Ahmed, M. Masroor and Mohamad, Dzulkifli and S. Khalil, Mohammad (2009) A hybrid approach for segmenting and validating T1-weighted normal brain mr images by employing ACM and ANN. In: International Conference on Soft Computing and Pattern Recognition (SoCPaR 2009), 2009, Melaka.

Full text not available from this repository.

Official URL: http://dx.doi.org/10.1109/SoCPaR.2009.56

Abstract

This study focuses on segmentation and validation of brain MR images. Artificial Neural Network (ANN) has been applied to obtain the targeted segments from these images. In preprocessing step for avoiding the chances of misclassification during training of ANN, the unwanted skull tissues were removed by employing active contour modeling (ACM). The removal of these tissues leaves an image containing various regions of interest. For training ANN these distinctive regions of interest were clustered into their respective regions by employing KMeans algorithm. Then a neural net work is trained on this classified data which eventually facilitated in obtaining the desired segments. The boundaries of these segments were detected and the pixels constituting these boundaries were counted. For validating the segments produced by ANN, ground truth segments were taken under consideration. The boundaries of these ground truth segments were also detected and pixels forming the edges of these segments were counted. Finally a comparison was drawn between the pixel counts of ANN produced segments and ground truth segments. On the basis of this comparison, accuracy of ANN is calculated.

Item Type:Conference or Workshop Item (Paper)
Subjects:Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions:Computer Science and Information System (Formerly known)
ID Code:14669
Deposited By: Liza Porijo
Deposited On:12 Sep 2011 02:11
Last Modified:12 Sep 2011 02:11

Repository Staff Only: item control page