Nikolaos Arikidis

Breast Cancer Diagnosis: Analyzing Texture of Tissue Surrounding Microcalcifications

The current study investigates texture properties of the tissue surrounding microcalcification (MC) clusters on mammograms for breast cancer diagnosis. The case sample analyzed consists of 85 dense mammographic images, originating from the digital database for screening mammography. mammograms analyzed contain 100 subtle MC clusters (46 benign and 54 malignant). The tissue surrounding MCs is defined on original and wavelet decomposed images, based on a redundant discrete wavelet transform. Gray-level texture and wavelet coefficient texture features at three decomposition levels are extracted from surrounding tissue regions of interest (ST-ROIs). Specifically, gray-level first-order statistics, gray-level cooccurrence matrices features, and Lawspsila texture energy measures are extracted from original image ST-ROIs. Wavelet coefficient first-order statistics and wavelet coefficient cooccurrence matrices features are extracted from subimages ST-ROIs. The ability of each feature set in differentiating malignant from benign tissue is investigated using a probabilistic neural network. Classification outputs of most discriminating feature sets are combined using a majority voting rule. The proposed combined scheme achieved an area under receiver operating characteristic curve (Az) of 0.989. Results suggest that MCspsila ST texture analysis can contribute to computer-aided diagnosis of breast cancer.

Assessing heterogeneity of lesion enhancement kinetics in dynamic contrast-enhanced MRI for breast cancer diagnosis

The current study investigates the feasibility of using texture analysis to quantify the heterogeneity of lesion enhancement kinetics in order to discriminate malignant from benign breast lesions. A total of 82 biopsy-proven breast lesions (51 malignant, 31 benign), originating from 74 women subjected to dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) were analysed. Pixel-wise analysis of DCE-MRI lesion data was performed to generate initial enhancement, post-initial enhancement and signal enhancement ratio (SER) parametric maps; these maps were subsequently subjected to co-occurrence matrix texture analysis. The discriminating ability of texture features extracted from each parametric map was investigated using a least-squares minimum distance classifier and further compared with the discriminating ability of the same texture features extracted from the first post-contrast frame. Selected texture features extracted from the SER map achieved an area under receiver operating characteristic curve of 0.922 +/- 0.029, a performance similar to post-initial enhancement map features (0.906 +/- 0.032) and statistically significantly higher than for initial enhancement map (0.767 +/- 0.053) and first post-contrast frame (0.756 +/- 0.060) features. Quantifying the heterogeneity of parametric maps that reflect lesion washout properties could contribute to the computer-aided diagnosis of breast lesions in DCE-MRI.

Computer-aided diagnosis of mammographic masses based on a supervised content-based image retrieval approach

Abstract In this work, the incorporation of content-based image retrieval (CBIR) into computer aided diagnosis (CADx) is investigated, in order to contribute to the decision-making process of radiologists in the characterization of mammographic masses. The proposed scheme comprises two stages: A margin-specific supervised CBIR stage that retrieves images from reference cases along with a decision stage that is based on the retrieved items. The feature set utilized exploits state-of-the-art features along with a newly proposed texture descriptor, namely mHOG, targeted to capturing margin and core specific mass properties. Performance evaluation considers the CBIR and diagnosis stages separately and is addressed by using standard measures on an enhanced version of the widely adopted digital database for screening mammography (DDSM). The proposed scheme achieved improved performance of CADx of masses in X-ray mammography experimentally compared to the state-of-the-art.

Size-adapted microcalcification segmentation in mammography utilizing scale-space signatures

The purpose of this study is size-adapted segmentation of individual microcalcifications in mammography, based on microcalcification scale-space signature estimation, enabling robust scale selection for initialization of multiscale active contours. Segmentation accuracy was evaluated by the area overlap measure, by comparing the proposed method and two recently proposed ones to expert manual delineations. The method achieved area overlap of 0.61+/-0.15 outperforming statistically (p<0.001) the other two methods (0.53+/-0.18, 0.42+/-0.16). Only the proposed method performed equally for both small (< 460 microm) and large (>/= 460 microm) microcalcifications. Results indicate an accurate method, which could be utilized in computer-aided diagnosis schemes of microcalcification clusters.

Microcalcification oriented content-based mammogram retrieval for breast cancer diagnosis

Microcalcifications (MCs) provide a significant early indication of breast malignancy. This work introduces a supervised scheme for malignancy risk assessment of mammograms containing MCs. The proposed scheme employs shape and textural features as input to a support vector machine (SVM) ensemble, in order to perform content-based image retrieval (CBIR) of mammograms. The retrieval performance of the proposed scheme has been evaluated by taking into account the variation of MCs morphology as defined in BI-RADS. In our experiments, we use a set of 87 mammograms containing MCs, obtained from the widely adopted DDSM database for screening mammography. The experimental results demonstrate that the proposed supervised CBIR scheme addresses effective retrieval of MCs mammograms outperforming relevant unsupervised schemes.

Size-adapted segmentation of individual mammographic microcalcifications

Accurate Microcalcification (MC) segmentation is a crucial first step in morphology based computer aided diagnosis systems for microcalcifications in mammography. In this article we present an automated segmentation method of individual MCs adaptive to both size and shape variations. Size is estimated by active rays (polar-transformed active contours) on continuous wavelet representation while shape adaptivity is achieved by a subsequent region growing step. Following MC seed point annotation, contour point estimates are obtained by implementing active rays on an analytic scale-space representation in a coarse-to-fine strategy. Initial coarsest scale is automatically defined by analyzing MC responses across scales. A region growing method is used to delineate the final MC contour curve, with pixel aggregation constrained by the MC contour point estimates. The segmentation accuracy of the proposed method was quantitatively evaluated by means of area overlap by comparing automatically derived borders with manually traced ones provided by an expert radiologist. The proposed method achieved an area overlap of 0.68plusmn0.13 on a dataset of 67 individual microcalcifications, originating from pleomorphic clusters.

Integrating multiscale polar active contours and region growing for microcalcifications segmentation in mammography

Morphology of individual microcalcifications is an important clinical factor in microcalcification clusters diagnosis. Accurate segmentation remains a difficult task due to microcalcifications small size, low contrast, fuzzy nature and low distinguishability from surrounding tissue. A novel application of active rays (polar transformed active contours) on B-spline wavelet representation is employed, to provide initial estimates of microcalcification boundary. Then, a region growing method is used with pixel aggregation constrained by the microcalcification boundary estimates, to obtain the final microcalcification boundary. The method was tested on dataset of 49 microcalcification clusters (30 benign, 19 malignant), originating from the DDSM database. An observer study was conducted to evaluate segmentation accuracy of the proposed method, on a 5-point rating scale (from 5:excellent to 1:very poor). The average accuracy rating was 3.98±0.81 when multiscale active rays were combined to region growing and 2.93±0.92 when combined to linear polynomial fitting, while the difference in rating of segmentation accuracy was statistically significant (p < 0.05).

Computerized Image Analysis of Mammographic Microcalcifications: Diagnosis and Prognosis

Breast cancer is the second leading cause of cancer deaths in women today (after lung cancer) and is the most frequently diagnosed cancer among women, excluding skin cancers. According to the American Cancer Society, an estimated of 230,480 new cancer cases are expected to be diagnosed in 2011; about 2,140 new cases are expected in men. In addition to invasive breast cancer, 57,650 new cases of in situ breast cancer are expected to occur among women in 2011. Of these, approximately 85% will be ductal carcinoma in situ (DCIS). An estimated 39,970 breast cancer deaths (39,520 women, 450 men) are expected in 2011. Death rates for breast cancer have steadily decreased in women since 1990, with larger decreases in women younger than 50 (a decrease of 3.2% per year) than in those 50 and older (2.0% per year), representing progress in both earlier detection and improved treatment.

Myocardial perfusion SPECT imaging de-noising: A phantom study

The statistical nature of SPECT imaging, due to Poisson noise effect, results in degradation of image quality, especially in case of lesions of small signal-to-noise (SNR) ratio (small size, reduced activity). In this paper, the performance of a platelet de-noising method applied, by means of a pre- processing step, on myocardial perfusion SPECT imaging is evaluated. A cardiac phantom, containing two different size cold lesions, was utilized to evaluate the platelet de-noising method performance and compare it with the performance of the Butterworth filtering method, applied on raw data in pre-processing fashion, as well as on reconstructed data, representing the clinical routine. Two experiments were conducted to simulate conditions with and without scatter irradiation from myocardial surrounding tissue. Noise, lesion contrast, SNR and lesion contrast-to-noise ratio (CNR) metrics for both lesions were computed for the three de-noising methods. Results demonstrate sufficient reduction of noise for platelet method yielding increased SNR and lesion CNR values as compared to Butterworth filtering method, applied on pre- and post-processed data, for both lesions. However, no statistically significant differences were demonstrated for all metrics considered (p>0.05). In conclusion, platelet de-noising prior to reconstruction has the potential to provide an efficient means of improving image quality in myocardial perfusion SPECT phantom.

Exploiting morphology and texture of 3D tumor models in DTI for differentiating glioblastoma multiforme from solitary metastasis

Abstract Ambiguous imaging appearance of Glioblastoma Multiforme (GBM) and solitary Metastasis (MET) is a challenge to conventional Magnetic Resonance Imaging (MRI) based diagnosis, leading to exploitation of advanced MRI techniques, such as Diffusion Tensor Imaging (DTI). In this study, 3D tumor models are generated by a DTI clustering segmentation technique, providing up to 16 brain tissue diffusivities, complemented by T1 post-contrast imaging, resulting in the identification of tumor core, whose surface is refined by a Morphological Morphing interpolation technique. The 3D models are analyzed in terms of their surface and internal signal variations characteristics towards identification of discriminant features for differentiation between GBMs and METs, utilizing a case sample composed of 10 GBMs and 10 METs. Morphology analysis of tumor core surface is assessed by 5 local curvature features. Texture analysis considers 11 first and 16 second order 3D textural features. From the 16 second order features, 11 are based on Gray Level Co-Occurrence Matrices (GLCM) and 5 on Gray Level Run Length Matrices (GLRLM), calculated from DTI isotropic and anisotropic parametric maps, corresponding to 3D tumor core segmented from the clustering technique. Also, 3 different image quantization levels (QL) were tested for both GLCM and GLRLM analysis, while 1–4 pixel displacements (D) in case of GLCM analysis. Case sample distributions of morphology and texture features were analyzed using the Mann-Whitney U test, with a cut-off value of 0.05 to identify discriminant features. The discriminatory performance of the derived features was analyzed with Receiver Operating Characteristic (ROC) curve analysis. Results highlight the value of all 5 local curvature descriptors to capture differences between the boundary of GBMs and METs. Histogram analysis of isotropy maps revealed statistical significant differences for median value and kurtosis, while 7 out of the 11 GLCM features were capable of discriminating heterogeneity of anisotropic diffusion properties of GBMs and METs, at QL = 6 and D = 2. Finally, all 5 GLRLM features extracted from diffusion isotropy maps seem to discriminate structural properties of GBMs and METs, at QL = 5. Results demonstrate the potential of surface morphology and texture analysis of 3D tumor imaging appearance in pre-treatment brain MRI tumor differentiation.