Fully automated region growing segmentation of brain lesion in diffusion-weighted MRI

N. Mohd Saad, S. A R Abu-Bakar, Sobri Muda, M. Mokji, A. R. Abdullah

Research output: Contribution to journalArticle

16 Citations (Scopus)

Abstract

This paper presents a fully automatic segmentation of brain lesions from diffusion-weighted magnetic resonance imaging (DW-MRI or DWI). The lesions are infarction, hemorrhage, tumor and abscess. Pre-processing stage is performed for intensity normalization, background removal and intensity enhancement. Then, split and merge algorithm is designed. Several statistical features are discussed and evaluated to select the best feature as homogeneity criteria. Lesions are segmented by merging the homogenous regions according to the selected criteria. This process produces blocky lesion region. Then, histogram thresholding is acquired to automate the seeds selection for region growing process. The region is iteratively grown by comparing all unallocated neighboring pixels to the seeds. The difference between pixel's intensity value and the region's mean is used as the similarity measure. The proposed segmentation technique has been validated by using misclassified area (MA), false positive rate (FPR), false negative rate (FNR), mean absolute percentage error (MAPE) and pixel absolute error ratio (r err), and compared with previous methods. The result shows that automatic region growing method can successfully segment the lesions and is suitable for analysis and classification of DWI.

Original languageEnglish
Pages (from-to)155-164
Number of pages10
JournalIAENG International Journal of Computer Science
Volume39
Issue number2
Publication statusPublished - Jun 2012

Fingerprint

Magnetic resonance imaging
Brain
Pixels
Seed
Magnetic resonance
Merging
Tumors
Imaging techniques
Processing

Keywords

  • Brain lesion
  • Diffusion-weighted MRI
  • Region growing
  • Segmentation
  • Split and merge

ASJC Scopus subject areas

  • Computer Science(all)

Cite this

Mohd Saad, N., Abu-Bakar, S. A. R., Muda, S., Mokji, M., & Abdullah, A. R. (2012). Fully automated region growing segmentation of brain lesion in diffusion-weighted MRI. IAENG International Journal of Computer Science, 39(2), 155-164.

Fully automated region growing segmentation of brain lesion in diffusion-weighted MRI. / Mohd Saad, N.; Abu-Bakar, S. A R; Muda, Sobri; Mokji, M.; Abdullah, A. R.

In: IAENG International Journal of Computer Science, Vol. 39, No. 2, 06.2012, p. 155-164.

Research output: Contribution to journalArticle

Mohd Saad, N, Abu-Bakar, SAR, Muda, S, Mokji, M & Abdullah, AR 2012, 'Fully automated region growing segmentation of brain lesion in diffusion-weighted MRI', IAENG International Journal of Computer Science, vol. 39, no. 2, pp. 155-164.
Mohd Saad, N. ; Abu-Bakar, S. A R ; Muda, Sobri ; Mokji, M. ; Abdullah, A. R. / Fully automated region growing segmentation of brain lesion in diffusion-weighted MRI. In: IAENG International Journal of Computer Science. 2012 ; Vol. 39, No. 2. pp. 155-164.
@article{bd7f3c26679d4cee82efaf6661fc886f,
title = "Fully automated region growing segmentation of brain lesion in diffusion-weighted MRI",
abstract = "This paper presents a fully automatic segmentation of brain lesions from diffusion-weighted magnetic resonance imaging (DW-MRI or DWI). The lesions are infarction, hemorrhage, tumor and abscess. Pre-processing stage is performed for intensity normalization, background removal and intensity enhancement. Then, split and merge algorithm is designed. Several statistical features are discussed and evaluated to select the best feature as homogeneity criteria. Lesions are segmented by merging the homogenous regions according to the selected criteria. This process produces blocky lesion region. Then, histogram thresholding is acquired to automate the seeds selection for region growing process. The region is iteratively grown by comparing all unallocated neighboring pixels to the seeds. The difference between pixel's intensity value and the region's mean is used as the similarity measure. The proposed segmentation technique has been validated by using misclassified area (MA), false positive rate (FPR), false negative rate (FNR), mean absolute percentage error (MAPE) and pixel absolute error ratio (r err), and compared with previous methods. The result shows that automatic region growing method can successfully segment the lesions and is suitable for analysis and classification of DWI.",
keywords = "Brain lesion, Diffusion-weighted MRI, Region growing, Segmentation, Split and merge",
author = "{Mohd Saad}, N. and Abu-Bakar, {S. A R} and Sobri Muda and M. Mokji and Abdullah, {A. R.}",
year = "2012",
month = "6",
language = "English",
volume = "39",
pages = "155--164",
journal = "IAENG International Journal of Computer Science",
issn = "1819-656X",
publisher = "International Association of Engineers",
number = "2",

}

TY - JOUR

T1 - Fully automated region growing segmentation of brain lesion in diffusion-weighted MRI

AU - Mohd Saad, N.

AU - Abu-Bakar, S. A R

AU - Muda, Sobri

AU - Mokji, M.

AU - Abdullah, A. R.

PY - 2012/6

Y1 - 2012/6

N2 - This paper presents a fully automatic segmentation of brain lesions from diffusion-weighted magnetic resonance imaging (DW-MRI or DWI). The lesions are infarction, hemorrhage, tumor and abscess. Pre-processing stage is performed for intensity normalization, background removal and intensity enhancement. Then, split and merge algorithm is designed. Several statistical features are discussed and evaluated to select the best feature as homogeneity criteria. Lesions are segmented by merging the homogenous regions according to the selected criteria. This process produces blocky lesion region. Then, histogram thresholding is acquired to automate the seeds selection for region growing process. The region is iteratively grown by comparing all unallocated neighboring pixels to the seeds. The difference between pixel's intensity value and the region's mean is used as the similarity measure. The proposed segmentation technique has been validated by using misclassified area (MA), false positive rate (FPR), false negative rate (FNR), mean absolute percentage error (MAPE) and pixel absolute error ratio (r err), and compared with previous methods. The result shows that automatic region growing method can successfully segment the lesions and is suitable for analysis and classification of DWI.

AB - This paper presents a fully automatic segmentation of brain lesions from diffusion-weighted magnetic resonance imaging (DW-MRI or DWI). The lesions are infarction, hemorrhage, tumor and abscess. Pre-processing stage is performed for intensity normalization, background removal and intensity enhancement. Then, split and merge algorithm is designed. Several statistical features are discussed and evaluated to select the best feature as homogeneity criteria. Lesions are segmented by merging the homogenous regions according to the selected criteria. This process produces blocky lesion region. Then, histogram thresholding is acquired to automate the seeds selection for region growing process. The region is iteratively grown by comparing all unallocated neighboring pixels to the seeds. The difference between pixel's intensity value and the region's mean is used as the similarity measure. The proposed segmentation technique has been validated by using misclassified area (MA), false positive rate (FPR), false negative rate (FNR), mean absolute percentage error (MAPE) and pixel absolute error ratio (r err), and compared with previous methods. The result shows that automatic region growing method can successfully segment the lesions and is suitable for analysis and classification of DWI.

KW - Brain lesion

KW - Diffusion-weighted MRI

KW - Region growing

KW - Segmentation

KW - Split and merge

UR - http://www.scopus.com/inward/record.url?scp=84864262592&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84864262592&partnerID=8YFLogxK

M3 - Article

VL - 39

SP - 155

EP - 164

JO - IAENG International Journal of Computer Science

JF - IAENG International Journal of Computer Science

SN - 1819-656X

IS - 2

ER -