Alexei Manso Correa Machado

MammoSys: A content-based image retrieval system using breast density patterns

In this paper, we present a content-based image retrieval system designed to retrieve mammographies from large medical image database. The system is developed based on breast density, according to the four categories defined by the American College of Radiology, and is integrated to the database of the Image Retrieval in Medical Applications (IRMA) project, that provides images with classification ground truth. Two-dimensional principal component analysis is used in breast density texture characterization, in order to effectively represent texture and allow for dimensionality reduction. A support vector machine is used to perform the retrieval process. Average precision rates are in the range from 83% to 97% considering a data set of 5024 images. The results indicate the potential of the system as the first stage of a computer-aided diagnosis framework.

Corpus callosum morphology and ventricular size in chromosome 22q11.2 deletion syndrome

In this paper, novel methods were used to map the corpus callosum morphology of children with chromosome 22q11.2 deletion syndrome in order to further investigate changes to that structure and to examine their possible effects on cognitive function. The callosal profiles were extracted from the centermost MRI midsagittal slice by supervised thresholding and the structures boundary and midline were computed automatically. Difference analysis was based on non-rigid registration, in which a template image is warped to conform to the shape of each corpus callosum in the sample. Boundaries and midlines were registered to a template and the results used to determine the average callosal shapes for children with the deletion and for controls. Pointwise registration also enabled the detailed evaluation of callosal curvature, width, area and length. Significant differences between the two groups were found in shape, size and bending angle. Results showed group differences that were concentrated in the anterior part of the structure, more specifically in the rostrum, which was larger and longer in the group with the syndrome. Correlation analyses showed that ventricular enlargement does not fully account for callosal morphology differences in children with the deletion. However, areal measurements did reveal important relationships between changes in callosal morphology and cognitive function. These novel findings reveal intricate relationships between genetic and disease-specific factors in the callosal anatomy and the potential impact of those changes on cognitive functions.

Atlas warping for brain morphometry

In this work, we describe an automated approach to morphometry based on spatial normalizations of the data, and demonstrate its application to the analysis of gender differences in the human corpus callosum. The purpose is to describe a population by a reduced and representative set of variables, from which a prior model can be constructed. Our approach is rooted in the assumption that individual anatomies can be considered as quantitative variations on a common underlying qualitative plane. We can therefore imagine that a given individuals anatomy is a warped version of some referential anatomy, also known as an atlas. The spatial warps which transform a labeled atlas into anatomic alignment with a population yield immediate knowledge about organ size and shape in the group. Furthermore, variation within the set of spatial warps is directly related to the anatomic variation among the subjects. Specifically, the shape statistics--mean and variance of the mappings--for the population can be calculated in a special basis, and an eigendecomposition of the variance performed to identify the most significant modes of shape variation. The results obtained with the corpus callosum study confirm the existence of substantial anatomical differences between males and females, as reported in previous experimental work.

Visual data mining for modeling prior distributions in morphometry

This article presents a novel method for visual data mining based on exploratory factor analysis. Modern imaging modalities provide an overwhelming amount of information that cannot be effectively handled without computerized tools. Data mining techniques aim to discover new knowledge from the collected data and to statistically represent this knowledge in the form of prior distributions that may be used to validate new hypotheses. When applied to morphometric studies, factor analysis is able to minimize data redundancy and reveal subtle or hidden patterns. The characterization of structural shape is performed in a new lower-dimensional basis in which the variables account for the correlation among regions of interest and provide morphological meaning. Data analysis is based on a set of vector variables obtained from image registration. The method is applied to a magnetic resonance imaging (MRI) study of the human corpus callosum and is able to reveal differences in the callosal morphology between male and female samples, based on unsupervised analysis.

Multiple Testing Correction in Medical Image Analysis

Abstract We present a novel method for correcting the significance level of hypothesis testing that requires multiple comparisons. It is based on the spectral graph theory, in which the variables are seen as the vertices of a complete undirected graph and the correlation matrix as the adjacency matrix that weights its edges. The method increases the statistical power of the analysis by refuting the assumption of independence among variables, while keeping the probability of false positives low. By computing the eigenvalues of the correlation matrix, it is possible to obtain valuable information about the dependence levels among the variables of the problem, so that the effective number of independent variables can be estimated. The method is compared to other available models and its effectiveness illustrated in case studies involving high-dimensional sets of variables.

Structural shape characterization via exploratory factor analysis

UNLABELLED This article presents an exploratory factor analytic approach to morphometry in which a high-dimensional set of shape-related variables is examined with the purpose of finding clusters with strong correlation. This clustering can potentially identify regions that have anatomic significance and thus lend insight to knowledge discovery and morphometric investigations. METHODS The information about regional shape is extracted by registering a reference image to a set of test images. Based on the displacement fields obtained form image registration, the amount of pointwise volume enlargement or reduction is computed and statistically analyzed with the purpose of extracting a reduced set of common factors. EXPERIMENTS The effectiveness and robustness of the method is demonstrated in a study of gender-related differences of the human corpus callosum anatomy, based on a sample of 84 right-handed normal controls. RESULTS The method is able to automatically partition the structure into regions of interest, in which the most relevant shape differences can be observed. The confidence of results is evaluated by analyzing the statistical fit of the model and compared to previous experimental works.

Exploratory Factor Analysis in Morphometry

In this paper, we present an exploratory factor analytic approach to morphometry in which a high-dimensional set of shape-related variables is examined with the purpose of finding clusters with strong correlation. This clustering can potentially identify regions that have anatomic significance and thus lend insight to the morphometric investigation. The analysis is based on information about size difference between the differential volume about points in a template image and their corresponding volumes in a subject image, where the correspondence is established by non-rigid registration. The Jacobian determinant field of the registration transformation is modeled by a reduced set of factors, whose cardinality is determined by an algorithm that iteratively eliminates factors that are not informative. The results show the method’s ability to identify gender-related morphological differences without supervision.

An interval space reducing method for constrained problems with particle swarm optimization

In this paper, we propose a method for solving constrained optimization problems using Interval Analysis combined with Particle Swarm Optimization. A Set Inverter Via Interval Analysis algorithm is used to handle constraints in order to reduce constrained optimization to quasi unconstrained one. The algorithm is useful in the detection of empty search spaces, preventing useless executions of the optimization process. To improve computational efficiency, a Space Cleaning algorithm is used to remove solutions that are certainly not optimal. As a result, the search space becomes smaller at each step of the optimization procedure. After completing pre-processing, a modified Particle Swarm Optimization algorithm is applied to the reduced search space to find the global optimum. The efficiency of the proposed approach is demonstrated through comprehensive experimentation involving 100,000 runs on a set of well-known benchmark constrained engineering design problems. The computational efficiency of the new method is quantified by comparing its results with other PSO variants found in the literature.

Dependence aliasing and the control of family-wise error rate in multiple hypothesis testing

The validation of the results obtained by hypothesis testing is of special interest in applications that deal with high-dimensional sets of variables. The use of equivocated statistical methods may result in poor control of false positives. On the other hand, overconservative methods may prevent relevant findings. In this paper we define dependence aliasing as the spurious dependence relationship among variables that appears when the number of samples is lesser than the number of variables of a study. We present a novel method for estimating the adjusted p-values in applications that require multiple hypothesis testing. The method increases the statistical power of the results by exploring the dependence among the variables, while controlling false positives in strong sense. The method is compared to other relevant adjustment models such as the false discovery rate method and resampling. We illustrate the effectiveness of the method in medical imaging studies involving progressively larger sets of variables. The results show that the proposed method is able to compute adjusted p-values that are closer to the ones obtained by resampling, but at a much lower computational cost.

Bayesian model for intensity mapping in magnetic resonance image registration

We present a likelihood model for Bayesian nonrigid image registration that relates the distinct acquisition models of different MRI (magnetic resonance imaging) scanners. The model is derived from a Bayesian network that represents the imaging situation under consideration to construct the appropriate similarity measure for the given situation. The method is compared to the cross-correlation and mutual information measures in a set of registration experiments on different images and over different synthetically generated geometric and intensity distortions. The probability-based similarity measure yields, on average, more accurate and robust registrations than either the cross-correlation or mutual information measures.