Limits...
A New GLLD Operator for Mass Detection in Digital Mammograms.

Gargouri N, Dammak Masmoudi A, Sellami Masmoudi D, Abid R - Int J Biomed Imaging (2012)

Bottom Line: We have used 1000 regions of interest (ROIs) obtained from the Digital Database for Screening Mammography (DDSM).The area under the curve of the corresponding approach has been found to be A(z) = 0.95 for the mass detection step.A comparative study with previous approaches proves that our approach offers the best performances.

View Article: PubMed Central - PubMed

Affiliation: Computer Imaging and Electronic System Group, CEM Laboratory, Department of Electrical Engineering, Sfax Engineering School, University of Sfax, P.O. Box 1169, 3038 Sfax, Tunisia.

ABSTRACT
During the last decade, several works have dealt with computer automatic diagnosis (CAD) of masses in digital mammograms. Generally, the main difficulty remains the detection of masses. This work proposes an efficient methodology for mass detection based on a new local feature extraction. Local binary pattern (LBP) operator and its variants proposed by Ojala are a powerful tool for textures classification. However, it has been proved that such operators are not able to model at their own texture masses. We propose in this paper a new local pattern model named gray level and local difference (GLLD) where we take into consideration absolute gray level values as well as local difference as local binary features. Artificial neural networks (ANNs), support vector machine (SVM), and k-nearest neighbors (kNNs) are, then, used for classifying masses from nonmasses, illustrating better performance of ANN classifier. We have used 1000 regions of interest (ROIs) obtained from the Digital Database for Screening Mammography (DDSM). The area under the curve of the corresponding approach has been found to be A(z) = 0.95 for the mass detection step. A comparative study with previous approaches proves that our approach offers the best performances.

No MeSH data available.


Related in: MedlinePlus

The central pixel gc and its P circularly symmetric neighbor with radius R.
© Copyright Policy - open-access
Related In: Results  -  Collection


getmorefigures.php?uid=PMC3539378&req=5

fig2: The central pixel gc and its P circularly symmetric neighbor with radius R.

Mentions: Given the new value of the central pixel gc mean and its P circularly symmetric neighbor (see Figure 2), the substraction of the value of gc mean is presented as follows:(6)T=t(gc mean,g0−gc mean,…,gP−1−gc mean).Thus, the difference between gc mean and gp may be represented as diffp = gp − gc mean and the local difference may be represented with a vector noted diffp knowing that diffp = [diff0,…, diffP−1], diffp describes the local image structure around the gc mean. Because of its robustness and efficiency, the obtained vector diffp is decomposed of sign and modulus components in order to achieve much better performance in texture classification. In our proposal, sp corresponds to the sign of the differences, and it is obtained by thresholding with respect to the value of gc mean as expressed in (8). However, mp corresponds to the absolute value of diffp as expressed in (9). We obtain, also, two vectors, the sign vector [s0,…, sP−1] and the modulus vector [m0,…, mP−1], with(7)diffp=mp·sp,knowing that(8)sp={1,diffp≥0−1,diffp<0,(9)mp=/diffp/.


A New GLLD Operator for Mass Detection in Digital Mammograms.

Gargouri N, Dammak Masmoudi A, Sellami Masmoudi D, Abid R - Int J Biomed Imaging (2012)

The central pixel gc and its P circularly symmetric neighbor with radius R.
© Copyright Policy - open-access
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC3539378&req=5

fig2: The central pixel gc and its P circularly symmetric neighbor with radius R.
Mentions: Given the new value of the central pixel gc mean and its P circularly symmetric neighbor (see Figure 2), the substraction of the value of gc mean is presented as follows:(6)T=t(gc mean,g0−gc mean,…,gP−1−gc mean).Thus, the difference between gc mean and gp may be represented as diffp = gp − gc mean and the local difference may be represented with a vector noted diffp knowing that diffp = [diff0,…, diffP−1], diffp describes the local image structure around the gc mean. Because of its robustness and efficiency, the obtained vector diffp is decomposed of sign and modulus components in order to achieve much better performance in texture classification. In our proposal, sp corresponds to the sign of the differences, and it is obtained by thresholding with respect to the value of gc mean as expressed in (8). However, mp corresponds to the absolute value of diffp as expressed in (9). We obtain, also, two vectors, the sign vector [s0,…, sP−1] and the modulus vector [m0,…, mP−1], with(7)diffp=mp·sp,knowing that(8)sp={1,diffp≥0−1,diffp<0,(9)mp=/diffp/.

Bottom Line: We have used 1000 regions of interest (ROIs) obtained from the Digital Database for Screening Mammography (DDSM).The area under the curve of the corresponding approach has been found to be A(z) = 0.95 for the mass detection step.A comparative study with previous approaches proves that our approach offers the best performances.

View Article: PubMed Central - PubMed

Affiliation: Computer Imaging and Electronic System Group, CEM Laboratory, Department of Electrical Engineering, Sfax Engineering School, University of Sfax, P.O. Box 1169, 3038 Sfax, Tunisia.

ABSTRACT
During the last decade, several works have dealt with computer automatic diagnosis (CAD) of masses in digital mammograms. Generally, the main difficulty remains the detection of masses. This work proposes an efficient methodology for mass detection based on a new local feature extraction. Local binary pattern (LBP) operator and its variants proposed by Ojala are a powerful tool for textures classification. However, it has been proved that such operators are not able to model at their own texture masses. We propose in this paper a new local pattern model named gray level and local difference (GLLD) where we take into consideration absolute gray level values as well as local difference as local binary features. Artificial neural networks (ANNs), support vector machine (SVM), and k-nearest neighbors (kNNs) are, then, used for classifying masses from nonmasses, illustrating better performance of ANN classifier. We have used 1000 regions of interest (ROIs) obtained from the Digital Database for Screening Mammography (DDSM). The area under the curve of the corresponding approach has been found to be A(z) = 0.95 for the mass detection step. A comparative study with previous approaches proves that our approach offers the best performances.

No MeSH data available.


Related in: MedlinePlus