Data Mining Practical Machine Learning Tools and Techniques

Measuring the Strangeness of Strange Attractors

Acts in what Hutchinson (1965) has called the 'ecological theatre' are played out on various scales of space and time. To understand the drama, we must view it on the appropriate scale. Plant ecologists long ago recognized the importance of sampling scale in their descriptions of the dispersion or distribution of species (e.g. Greig-Smith, 1952). However, many ecologists have behaved as if patterns and the processes that produce them are insensitive to differences in scale and have designed their studies with little explicit attention to scale. Kareiva & Andersen (1988) surveyed nearly 100 field experiments in community ecology and found that half were conducted on plots no larger than 1 m in diameter, despite considerable differences in the sizes and types of organisms studied. Investigators addressing the same questions have often conducted their studies on quite different scales. Not surprisingly, their findings have not always matched, and arguments have ensued. The disagreements among conservation biologists over the optimal design of nature reserves (see Simberloff, 1988) are at least partly due to a failure to appreciate scaling differences among organisms. Controversies about the role of competition in structuring animal communities (Schoener, 1982; Wiens, 1983, 1989) or about the degree of coevolution in communities (Connell, 1980; Roughgarden, 1983) may reflect the

https://www.jstor.org/stable/info/2389612

Spatial Scaling in Ecology

Thirty-seven programs that summarize data with histograms and other graphics, calculate measures of spatial continuity, provide smooth least-squares-type maps, and perform stochastic spatial simulation.

GSLIB: Geostatistical Software Library and User's Guide

https://www.thelancet.com/pdfs/journals/lancet/PIIS0140-6736(06)68880-6.pdf

Mild cognitive impairment

Blood flow and tissue velocity can be measured using phase-contrast MRI. In this work, the statistical properties of 4D phase-contrast images are derived, and a novel probabilistic blood flow mapping method based on sequential Monte Carlo sampling is presented. The resulting flow maps visualize and quantify the uncertainty in conventional flow visualization techniques such as streamlines and particle traces.

/pdf/probabilistic-4d-blood-flow-mapping-4o3iyuylzy.pdf

Probabilistic 4D blood flow mapping

The scaling properties of the portal vein and the hepatic vein are examined, based on three-dimensional computed tomography images of casts of human livers. In particular, a quantitative analysis of the branching ratio based on the Strahler ordering scheme as well as the diameter and length ratios are performed. To quantify the segment anatomy of the liver, the volume of supplied liver parenchyma is measured and related to the corresponding vessel radius. The implications of these findings for segment-oriented liver surgery are discussed. We also investigate the 3D interdependence of the intertwined portal and hepatic veins based on a concept of tree distance that will be introduced.

Fractal properties, segment anatomy, and interdependence of the human portal vein and the hepatic vein in 3d

Many medical applications require a registration of different images of the same organ. In many cases, such a registration is accomplished by manually placing landmarks in the images. In this paper we propose a method which is able to find reasonable landmarks automatically. To achieve this, nodes of the vessel systems, which have been extracted from the images by a segmentation algorithm, will be assigned by the so-called association graph method and the coordinates of these matched nodes can be used as landmarks for a non-rigid registration algorithm.

Matching of tree structures for registration of medical images

We introduce novel data structures and algorithms for clustering white matter fiber tracts to improve accuracy
and robustness of existing techniques. Our novel fiber grid combined with a new randomized soft-division
algorithm allows for defining the fiber similarity more precisely and efficiently than a feature space. A fine-tuning
of several parameters to a particular fiber set - as it is often required if using a feature space - becomes obsolete.
The idea is to utilize a 3D grid where each fiber point is assigned to cells with a certain weight. From this grid, an
affinity matrix representing the fiber similarity can be calculated very efficiently in time O( n ) in the average case,
where n denotes the number of fibers. This is superior to feature space methods which need O( n 2 ) time. Our novel
eigenvalue regression is capable of determining a reasonable number of clusters as it accounts for inter-cluster
connectivity. It performs a linear regression of the eigenvalues of the affinity matrix to find the point of maximum
curvature in a list of descending order. This allows for identifying inner clusters within coarse structures, which
automatically and drastically reduces the a-priori knowledge required for achieving plausible clustering results.
Our extended multiple eigenvector clustering exhibits a drastically improved robustness compared to the well-known
elongated clustering, which also includes an automatic detection of the number of clusters. We present
several examples of artificial and real fiber sets clustered by our approach to support the clinical suitability and
robustness of the proposed techniques.

Grid-based spectral fiber clustering

In follow-up CT examinations of cancer patients, therapy success is evaluated by estimating the change in tumor size from diameter or volume comparison between corresponding lesions. We present an algorithm that automatizes the detection of matching lesions, given a baseline segmentation mask. It is generally applicable and does not need an organ mask or CAD findings, only a coarse registration of the datasets is required. In the first step, lesion candidates are identified in a local area based on gray value filtering and detection of circular structures using a Hough transform. On all candidate voxels, a template matching is performed minimizing normalized cross-correlation. The method was evaluated on clinical follow-up data comprising 94 lung nodules, 107 liver metastases, and 137 lymph nodes. The ratio of correctly detected lesions was 96%, 84% and 85%, respectively, at an average computation time of 0.9 s per lesion on a standard PC.

Heinz-Otto Peitgen

Papers

Probabilistic 4D blood flow mapping

Fractal properties, segment anatomy, and interdependence of the human portal vein and the hepatic vein in 3d

Matching of tree structures for registration of medical images

Grid-based spectral fiber clustering

A general framework for automatic detection of matching lesions in follow-up CT