Using neural networks to select wavelet features for breast cancer diagnosis

Kocur, Catherine M.; Rogers, Steven K.; Myers, Lemuel R.; Burns, Thomas J.; Kabrisky, Matthew; Hoffmeister, Jeffrey W.; Bauer, Kenneth W.; Steppe, Jean M.

doi:10.1109/51.499766

Cited by 39 publications

(13 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As compared to other texture-based CADx schemes analyzing ROIs containing the cluster (Dhawan et al, 1996;Kocur et al, 1996;Chan et al, 1997;Chan et al, 1998;Kramer & Aghdasi 1999;Soltanian-Zadeh et al, 2004), the performance achieved by the wavelet texture signatures, employing the MCs surrounding tissue approach, is also comparable. However, heterogeneity of the datasets analyzed renders direct comparison not feasible.…”

Section: Discussionmentioning

confidence: 64%

“…Features extracted from GLCMs provide information concerning image texture heterogeneity and coarseness, which is not necessarily visually perceived. The discriminating ability of GLCMs features, as extracted from original image ROIs containing MCs, has been demonstrated by most studies (Dhawan et al, 1996;Kocur et al, 1996;Kramer & Aghdasi 1999;Chan et al, 1997;Chan et al, 1998;Soltanian-Zadeh et al, 2004), with specific GLCMs feature combinations achieving an (Chan et al, 1997). In addition, GLCMs feature have shown to be more effective than morphology-based features (Chan et al, 1998), while their combination can provide an even higher classification performance.…”

Section: Texture-based Cadx Schemesmentioning

confidence: 99%

“…energy, entropy and square root of the coefficients norm) extracted from wavelet (Dhawan et al, 1996;Kocur et al, 1996;Soltanian-Zadeh et al, 2004) (Kocur et al, 1996;Soltanian-Zadeh et al, 2004) and shape features (Soltanian-Zadeh et al, 2004) suggesting the advantages offered by the multiscale representation of the tissue analyzed. Dhawan et al (1996) demonstrated that the combination of GLCMs with FOS wavelet features, representing global and local texture respectively, is superior to cluster features; however, best performance was achieved by a selected feature set including GLCMs, wavelet and cluster features.…”

Section: Texture-based Cadx Schemesmentioning

confidence: 99%

See 2 more Smart Citations

Computerized Image Analysis of Mammographic Microcalcifications: Diagnosis and Prognosis

Karahaliou¹,

Arikidis²,

Skiadopoulos³

et al. 2012

Mammography - Recent Advances

View full text Add to dashboard Cite

Section: Discussionmentioning

confidence: 64%

Section: Texture-based Cadx Schemesmentioning

confidence: 99%

Section: Texture-based Cadx Schemesmentioning

confidence: 99%

See 1 more Smart Citation

Computerized Image Analysis of Mammographic Microcalcifications: Diagnosis and Prognosis

Karahaliou¹,

Arikidis²,

Skiadopoulos³

et al. 2012

Mammography - Recent Advances

View full text Add to dashboard Cite

“…It should be noted that number of points on a ROC depends on number of possible answers from a method. In the case of using radiologist the curve is comprised of 5 points because there are only 5 possible ratings (1)(2)(3)(4)(5). The computational algorithm of calculating the area under the ROC curve is based on the method of Beck and Schultz 16 .…”

Section: Resultsmentioning

confidence: 99%

Diagnostic abilities of three CAD methods for assessing microcalcifications in mammograms and an aspect of equivocal cases decisions by radiologists

Hung

Nguyen

Lee

et al. 2003

Australas. Phys. Eng. Sci. Med.

View full text Add to dashboard Cite

Radiologists use an "Overall impression" rating to assess a suspicious region on a mammogram. The value ranges from 1 to 5. They will definitely send a patient for biopsy if the rating is 4 or 5. They will send the patient for core biopsy when a rating of 3 (indeterminate) is given. We have developed three methods to aid diagnosis of cases with microcalcifications. The first two methods, namely, Bayesian and multiple logistic regression (with a special "cutting score" technique), utilise six parameter ratings which minimise subjectivity in characterising the microcalcifications. The third method uses three parameters (age of patient, uniformity of size of microcalcification and their distribution) in a multiple stepwise regression. For both training set and test set, all three methods are as good as the two radiologists in terms of percentages of correct classification. Therefore, all three proposed methods potentially can be used as second readers.

show abstract

“…Practically the approximation and detail projection coefficients associated with V j and W j are computed from the approximation and detail coefficients at the next higher scale V j −1 , using a Quadrature Mirror Filter (QMF) pair and a pyramidal subband coding scheme [12,13].…”

Section: The Wavelet Transformationmentioning

confidence: 99%

Supervised and unsupervised neural network methods applied to textile quality control based on improved wavelet feature extraction techniques

Karras

Karkanis²,

Mertzios

1998

International Journal of Computer Mathematics

View full text Add to dashboard Cite

This paper aims at investigating novel solutions to the problem of textile defect detection from images, that can find applications in building robust quality control vision based systems in textile production. The proposed solutions focus on detecting defects from the textural properties of their corresponding wavelet transformed images. More specifically a novel methodology is investigated for discriminating defects in textile images by applying supervised and unsupervised neural classification techniques, employing multilayer perceptrons (MLP) -trained with the on-line backpropagation algorithm -and Kohonen's Self-Organizing Feature Maps (SOFM) respectively. These parallel techniques are applied to innovative wavelet based feature vectors. These vectors are extracted from the wavelet transformed original images using the cooccurrence matrices framework and SVD analysis. The results of the proposed methodology are illustrated in defective textile images where the defective areas are recognized with about 98.5% accuracy.

show abstract

Using neural networks to select wavelet features for breast cancer diagnosis

Cited by 39 publications

References 18 publications

Computerized Image Analysis of Mammographic Microcalcifications: Diagnosis and Prognosis

Computerized Image Analysis of Mammographic Microcalcifications: Diagnosis and Prognosis

Diagnostic abilities of three CAD methods for assessing microcalcifications in mammograms and an aspect of equivocal cases decisions by radiologists

Supervised and unsupervised neural network methods applied to textile quality control based on improved wavelet feature extraction techniques

Contact Info

Product

Resources

About