- [Show abstract] [Hide abstract] ABSTRACT: Based on previous work on regional temporal mammogram registration, this study investigates the combination of image features measured from single regions (single features) and image features measured from the matched regions of temporal mammograms (temporal features) for the classification of malignant masses. Three SVM kernels, the multilayer perceptron kernel, the polynomial kernel, and the gaussian radial basis function kernel, and the combination of these kernels, the multiple kernel learning method, were applied to both single and temporal features for the mass classification. To combine the two types of features, 3 combination rules, Linear combination, Max and Min, were used to combine classification results obtained on single and temporal features. The results showed that combining the MKL classification results on single features, and MKL classification results on temporal features, with Min rule produces the best classification results. The experiment result indicates that incorporating the temporal change information in mammography mass classification can improve the performance detection.
- [Show abstract] [Hide abstract] ABSTRACT: This paper presents a method for incorporating fuzzy sets based spatial relations in registering temporal mammogram pairs. In the proposed method, four spatial relations, to the right of, to the left of, below, and above, are considered. The histogram of all possible angles between all pairs of points in a pair of regions of interest (ROIs) is treated as a fuzzy set and spatial relations between the pair of ROIs are characterized by measuring to what degree this fuzzy set approaches the four spatial relations. Based on the spatial relations, association of ROIs of temporal mammogram pairs is then treated as a graph matching problem and registration of temporal mammogram pairs is realized by finding the common subgraph between two graphs representing a pair of temporal mammograms. 95 pairs of real temporal mammograms are used to test the proposed method. 70.8% of matched ROI pairs are visually identified as “good” matches. When the registration results are incorporated in a cancer detection scheme, the score (area under the ROC curve) is improved from 0.846 to 0.852. The results demonstrated that registration of temporal mammograms based on fuzzy spatial relations improves the overall detection efficiency.
- [Show abstract] [Hide abstract] ABSTRACT: Higher-order texture features from 100 mammographic images with known cancer were compared to texture features from 100 images from women with no known cancer. Texture features from images of the same breasts from screening rounds two and four years previously were also compared. The A z score for classifying cancer images from non-cancer images was 0.749. The A z score for classification two years previous to detection of cancer was 0.674 and the score for four years previous was 0.601. There was no signicant difference between classifying images from the round in which cancer was actually detected and the screening rounds two and four years previous. Similar results were obtained if the breast with no known cancer (contralateral breast) was used instead the breast with cancer, leading to the conclusion that texture alone has moderate predictive power regarding breast cancer risk and that this predictive value is roughly constant in the four years prior to mammographically apparent cancer.
- [Show abstract] [Hide abstract] ABSTRACT: Texture analysis based on textons is extended by introducing a method for computing textons of arbitrary order. First-, second- and third-order textons are applied to classify screening mammograms as to indicate a low or high risk of breast cancer. First-order textons are found to provide better estimates of breast cancer risk than other orders on their own but the combination of first- and second-order textons outperforms first-order textons alone and other combinations of two orders. Combining all three orders of textons does not improve classification. This example indicates that including higher-order textons has the potential to improve classification performance.
- [Show abstract] [Hide abstract] ABSTRACT: Breast density is a known risk factor for breast cancer. Here two classes of texture features, one based on textons derived from local pixel intensity variation and one based on oriented tissue structure characteristics are measured on different regions of the breast in an effort to clarify the potential contribution of texture independent of local tissue density to estimate breast cancer risk. The region just behind the nipple is found to be the most significant local region for estimating risk, but estimates based on the entire breast perform better. Texton features are found to perform better than features based on oriented tissue structure.
- [Show abstract] [Hide abstract] ABSTRACT: Image intensity and texture in screening mammograms are thought to be associated with the risk of breast cancer. Studies on developing automatic breast cancer risk assessment schemes tend to employ texture measures which are correlated to local background intensity. Accordingly, the contribution of texture alone to risk assessment is not known. Here background intensity independent texture measures are used to assess cancer risk. Moreover risk assessment based on background intensity independent texture outperforms intensity dependent texture suggesting that local image background intensity may confound risk assessment. Performance seems to depend on the view of the breast and so suggests that optimizing schemes for different views may improve risk assessment.
- [Show abstract] [Hide abstract] ABSTRACT: A model is presented for characterizing the process by which cancellous bone changes in volume and structure over time. The model comprises simulations of local changes resulting from individual remodelling events, known as bone multicellular units (BMU), and an ordinary differential equation for connecting the number of remodelling events to real time. The model is validated on micro-CT scans of tibiae of normal rats, estrogen deprived rats and estrogen deprived rats treated with bisphosphonates. The model explains the asymptotic trends seen in changes of bone volume over time resulting from estrogen deprivation as well as trends seen subsequent to treatment. The model demonstrates that both bone volume and structure changes can be explained in terms of resetting remodelling parameters. The model also shows that either current understanding of the effects of bisphosphonates is not correct or that the simplest description of remodelling does not suffice to explain both the change in bone volume and structure of rats treated with bisphosphonates.
Conference Paper: Intensity Independent Texture Analysis in Screening Mammograms[Show abstract] [Hide abstract] ABSTRACT: Image texture features for detecting malignant masses in screening mammograms are proposed that are independent of background intensity mean and variation. Subtracting local means and dividing by local standard deviation reveals linear structures of approximately 0.7 mm width in screening mammograms. A simple texture feature calculated from on this derived image is used to demonstrate that texture information associated with the location of cancer is retained in the mean and standard deviation normalized image. Such texture features have the potential to provide evidence of malignancy that better complements intensity based features for detecting breast cancer in screening mammograms.
- [Show abstract] [Hide abstract] ABSTRACT: This chapter proposes a fractal attractor model of tumor growth and metastasis. It is a 4-dimensional spatio-temporal cancer model with strong nonlinear couplings. Even the same type of tumor is different in every patient both in size and appearance, as well as in temporal behavior. This is clearly a characteristic of dynamical systems sensitive to initial conditions. The new chaotic model of tumor growth and decay is biologically motivated. It has been developed as a live Mathematica demonstration, see Wolfram Demonstrator site.
- [Show abstract] [Hide abstract] ABSTRACT: An important difference between projection images such as x-rays and natural images is that the intensity at a single pixel in a projection image comprises information from all objects between the source and detector. In order to exploit this information, a Dirichlet mixture of Gaussian distributions is used to model the intensity function forming the projection image. The model requires initial seeding of Gaussians and uses the EM (estimation maximisation) algorithm to arrive at a final model. The resulting models are shown to be robust with respect to the number and positions of the Gaussians used to seed the algorithm. As an example, a screening mammogram is modelled as the Dirichlet sum of Gaussians suggesting possible application to early detection of breast cancer.
Conference Paper: Mammographic Mass Detection with Statistical Region Merging[Show abstract] [Hide abstract] ABSTRACT: An automatic method for detection of mammographic masses is presented which utilizes statistical region merging for segmentation (SRM) and linear discriminant analysis (LDA) for classification. The performance of the scheme was evaluated on 36 images selected from the local database of mammograms and on 48 images taken from the Digital Database for Screening Mammography (DDSM). The Az value (area under the ROC curve) for classifying each region was 0.90 for the local dataset and 0.96 for the images from DDSM. Results indicate that SRM segmentation can form part of an robust and efficient basis for analysis of mammograms.
- [Show abstract] [Hide abstract] ABSTRACT: A method is presented for including information from the preceeding mammogram in a scheme for automatically detecting malignant masses in screening mammograms. The method circumvents the inherent difficulty of registering temporal mammograms by replacing image registration by graph matching. The scheme incorporates a single image mass detection algorithm and so the contribution of the temporal analysis can be measured. At a true detection rate of 80 percent, the single image scheme results in 1.02 false positive detections per image while the temporal scheme results in 0.96 false positives. At 90 percent true detection, the false positive rates per image are 1.84 and 1.63 respectively.
- [Show abstract] [Hide abstract] ABSTRACT: A system of diffusive logistic equations with fixed impulse times and contin-uous time delay is investigated. This system represents the dynamics of a multi species population. Some conditions under which the positive steady-state of the system without impulses becomes an attractor of the system with impulses are presented.
- [Show abstract] [Hide abstract] ABSTRACT: A method based on sublevel sets is presented for refining segmentation of screening mammograms. Initial segmentation is provided by an adaptive pyramid (AP) scheme which is viewed as seeding of the final segmentation by sublevel sets. Performance is tested with and without prior anisotropic smoothing and is compared to refinement based on component merging. The combination of anisotropic smoothing, AP segmentation and sublevel refinement is found to outperform other combinations.
- [Show abstract] [Hide abstract] ABSTRACT: A technique utilizing an entropy measure is developed for automatically tuning the segmentation of screening mammograms by minimum spanning trees (MST). The lack of such technique has been a major obstacle in previous work to segment mammograms for registration and applying mass detection algorithms. The proposed method is tested on two sets of mammograms: a set of 55 mammograms chosen from a publicly available Mini-MIAS database, and a set of 37 mammograms selected from a local database. The method performance is evaluated in conjunction with three different preprocessing filters: gaussian, anisotropic and neutrosophic. Results show that the automatic tuning has the potential to produce state-of-the art segmentation of mass-like objects in mammograms. The neutrosophic filtering provided the best performance.
- [Show abstract] [Hide abstract] ABSTRACT: A technique utilizing an entropy measure is developed for automatically tuning the segmentation of screening mammograms by minimum spanning trees (MST). The lack of such technique has been a major obstacle in previous work to segment mammograms for registration and applying mass detection algorithms.
- [Show abstract] [Hide abstract] ABSTRACT: We propose a strange-attractor model of tumor growth and metastasis. It is a 4-dimensional spatio-temporal cancer model with strong nonlinear couplings. Even the same type of tumor is different in every patient both in size and appearance, as well as in temporal behavior. This is clearly a characteristic of dynamical systems sensitive to initial conditions. The new chaotic model of tumor growth and decay is biologically motivated. It has been developed as a live Mathematica demonstration, see Wolfram Demonstrator site: http://demonstrations.wolfram.com/ChaoticAttractorInTumorGrowth/ Key words: Reaction-diffusion tumor growth model, chaotic attractor, sensitive dependence on initial tumor characteristics
Tarndarnya, South Australia, Australia
- • School of Computer Science, Engineering and Mathematics
- • School of Chemical and Physical Sciences