M. Termenon

Computer Aided Diagnosis system for Alzheimer Disease using brain Diffusion Tensor Imaging features selected by Pearson's correlation

The aim of this paper is to obtain discriminant features from two scalar measures of Diffusion Tensor Imaging (DTI) data, Fractional Anisotropy (FA) and Mean Diffusivity (MD), and to train and test classifiers able to discriminate Alzheimers Disease (AD) patients from controls on the basis of features extracted from the FA or MD volumes. In this study, support vector machine (SVM) classifier was trained and tested on FA and MD data. Feature selection is done computing the Pearsons correlation between FA or MD values at voxel site across subjects and the indicative variable specifying the subject class. Voxel sites with high absolute correlation are selected for feature extraction. Results are obtained over an on-going study in Hospital de Santiago Apostol collecting anatomical T1-weighted MRI volumes and DTI data from healthy control subjects and AD patients. FA features and a linear SVM classifier achieve perfect accuracy, sensitivity and specificity in several cross-validation studies, supporting the usefulness of DTI-derived features as an image-marker for AD and to the feasibility of building Computer Aided Diagnosis systems for AD based on them.

A Two Stage Sequential Ensemble Applied to the Classification of Alzheimer's Disease Based on MRI Features

We present a two stage sequential ensemble where data samples whose output from the first classifier fall in a low confidence output interval (LCOI) are processed by a second stage classifier. Training is composed of three processes: training the first classifier, determining the LCOI of the first classifier, and training the second classifier upon the data items whose output fall in the LCOI. The LCOI is determined varying a threshold on the false positive rate (FPR) and false negative rate (FNR) curves. We have tested the approach on a database of feature vectors for the classification of Alzheimer’s disease (AD) and control subjects extracted from structural magnetic resonance imaging (sMRI) data. In this paper, we focus on the combinations obtained when the first classifier is a relevance vector machine (RVM). Obtained results improve over previous results for this database.

Discovering Alzheimer's disease and bipolar disorder white matter effects building computer aided diagnostic systems on brain diffusion tensor imaging features.

The aim of this study is to look for differential effects in white matter (WM) of bipolar disorder (BD) and Alzheimers disease (AD) patients. We proceed by investigating the feasibility of discriminating between BD and AD patients, and from healthy controls (HC), using multivariate data analysis based on diffusion tensor imaging (DTI) data features. Specifically, support vector machine (SVM) classifiers were trained and tested on fractional anisotropy (FA). Voxel sites are selected as features for classification if their Pearsons correlation between FA values at voxel site across subjects and the indicative variable specifying the subject class is above the threshold set by a percentile of its empirical distribution. To avoid double dipping, selection was performed only on training data in a leave one out cross-validation study. Classification results show that FA features and a linear SVM classifier achieve perfect accuracy, sensitivity and specificity in AD vs. HC, BD vs. HC, and AD vs. BD leave-one-out cross-validation studies. The localization of the discriminant voxel sites on a probabilistic tractography atlas shows effects on seven major WM tracts in each hemisphere and two commissural tracts.

Extreme Learning Machines for Feature Selection and Classification of Cocaine Dependent Patients on Structural MRI Data

In this paper, we present a Computer Aided Diagnosis and image biomarker identification system for cocaine dependence, which selects relevant regions from a set of brain structural magnetic resonance images (sMRI). After sMRI volume preprocessing for spatial normalization, we compute Pearson’s correlation between pixel values across volumes and the indicative variable, obtaining a volume of correlation values (VCV). We calculate the gradient of the VCV which is used to perform a watershed segmentation of the brain volume into regions. A region selection stage finds the most relevant watershed regions. We propose two different approaches to characterize region relevance: (a) a wrapper procedure using extreme learning machines (ELM), and (b) apply correlation distribution percentiles to select most discriminant regions. Next, we consider three different procedures to extract the image features corresponding to selected regions: (1) collecting the sMRI intensity values of all the voxels that compose each region, compute (2) the mean or (3) the median of the sMRI intensity value of the voxels contained in each selected region. Extracted feature vectors are used to build a classifier aiming to discriminate between cocaine dependent patients and healthy controls. We compare results of several classifiers: ELM, OP-ELM, SVM and 1NN. Also, we visualize the brain locations of selected regions, checking if these locations are in accordance with previous findings in the medical literature.

Lattice independent component analysis feature selection on diffusion weighted imaging for Alzheimer's disease classification

Abstract Diffusion weighted imaging (DWI) provides information on the diffusion of water molecules which can be useful to determine structural properties in the brain. Specifically, fractional anisotropy (FA) is a scalar measure computed from each voxels diffusion tensor giving information about the existence of a privileged diffusion direction. The FA volume is the raw data in our classification approach. We apply lattice independent component analysis (LICA) across volumes for feature selection on FA data to perform classification of healthy control (HC) subjects and Alzheimers disease (AD) patients. Feature selection is done on the basis of Pearsons correlation between the LICA residuals at each voxel site and the data indicative variable. Voxel sites having an absolute value Pearsons correlation above a given percentile of its empirical distribution are selected as feature variables for classification. We compare the LICA based feature selection with (a) a Pearsons correlation approach on the raw FA data, and (b) a voxel based morphometry (VBM) approach. We apply relevance vector machines (RVM), nearest-neighbor (1NN) and linear support vector machines (LSVM) to build classifiers on these feature vectors. LSVM reach very high accuracy, specificity and sensitivity for some feature selection percentile parameter values. We provide results of the approach on data coming from an on-going study in Hospital de Santiago Apostol collecting anatomical T1-weighted MRI volumes and DWI data of HC and AD patients. Results point to the validity of FA data as an image-marker for AD.

Knowledge based tools to support the structural design process

In this paper we present an architecture and a system implementation for the exploitation of semantic aspects in the computer assisted design of Steel Detailing Structures (Structural Design). We support our approach in domain specific standardization efforts (CIS/2) by modeling the knowledge of the product structure and the design process as OWL ontologies in order to provide the designer with additional tools for his everyday work. As test case we present a collection of applications based in the proposed schema and developed in the frame of an R&D project together with an actual structural engineering company.

Cocaine dependent classification using brain magnetic resonance imaging

The purpose of this study is to elucidate if it is possible to discriminate between cocaine dependent patients and healthy controls applying computer aided diagnosis tools to brain magnetic resonance imaging. Feature extraction was done computing Pearsons correlation using subjects class as indicative variable. Linear support vector machines classifiers were trained and tested on the most significative voxels using leave one out cross-validation process. Results show that classifier achieve on average almost perfect accuracy, sensitivity and specificity in a group of 30 cocaine-dependent and 35 controls, supporting the usefulness of this process to discriminate between these subjects.

Alzheimer disease classification on diffusion weighted imaging features

An on-going study in Hospital de Santiago Apostol collects anatomical T1-weighted MRI volumes and Diffusion Weighted Imaging (DWI) data of control and Alzheimers Disease patients. The aim of this paper is to obtain discriminant features from scalar measures of DWI data, the Fractional Anisotropy (FA) and Mean Diffusivity (MD) volumes, and to train and test classifiers able to discriminate AD patients from controls on the basis of features selected from the FA or MD volumes. In this study, separate classifiers were trained and tested on FA and MD data. Feature selection is done according to the Pearsons correlation between voxel values across subjects and the control variable giving the subject class (1 for AD patients, 0 for controls). Some of the tested classifiers reach very high accuracy with this simple feature selection process. Those results point to the validity of DWI data as a image-marker for AD.

Ontology Supported Adaptive User Interfaces for Structural CAD Design

In the early stages of the Steel Detailing Design process (Structural Design), most of the activities are focused in the designer. Nowadays Detailing CAD packages offer a wide range of options that in some cases exceeds the ones needed to fulfil a specific task. Sometimes having such a wide range span can be self-defeating for a smooth process evolution as the designer has to browse repetitively in the user interface for a particular tool. In this paper we present a Knowledge based approach for the exploitation of semantic aspects (e.g. user intentions and tasks) for the real time automatic generation of graphical user interfaces on a Steel Detailing CAD software. We base our approach in international standards (CIS/2) for the specific domain and as test case we present a system implementation of the proposed schema.

A comparison of VBM results by SPM, ICA and LICA

Lattice Independent Component Analysis (LICA) approach consists of a detection of independent vectors in the morphological or lattice theoretic sense that are the basis for a linear decomposition of the data We apply it in this paper to a Voxel Based Morphometry (VBM) study on Alzheimers disease (AD) patients extracted from a well known public database The approach is compared to SPM and Independent Component Analysis results.