Intrasubject subcortical quantitative referencing to boost MRI sensitivity to Parkinson's disease.

Several postmortem studies have shown iron accumulation in the substantia nigra of Parkinson's disease patients. Iron concentration can be estimated via MRI-R2∗ mapping. To assess the changes in R2∗ occurring in Parkinson's disease patients compared to controls, a multicentre transversal study was carried out on a large cohort of Parkinson's disease patients (n = 163) with matched controls (n = 82). In this study, 44 patients and 11 controls were removed due to motion artefacts, 21 patient and 6 controls to preserve matching. Thus, 98 patients and 65 age and sex-matched healthy subjects were selected with enough image quality. The study was conducted on patients with early to late stage Parkinson's disease. The images were acquired at 3Tesla in 12 clinical centres. R2∗ values were measured in subcortical regions of interest (substantia nigra, red nucleus, striatum, globus pallidus externus and globus pallidus internus) contralateral (dominant side) and ipsilateral (non dominant side) to the most clinically affected hemibody. As the observed inter-subject R2∗ variability was significantly higher than the disease effect, an original strategy (intrasubject subcortical quantitative referencing, ISQR) was developed using the measurement of R2∗ in the red nucleus as an intra-subject reference. R2∗ values significantly increased in Parkinson's disease patients when compared with controls; in the substantia nigra (SN) in the dominant side (D) and in the non dominant side (ND), respectively (PSN_D and PSN_ND < 0.0001). After stratification into four subgroups according to the disease duration, no significant R2∗ difference was found in all regions of interest when comparing Parkinson's disease subgroups. By applying our ISQR strategy, R2(ISQR)∗ values significantly increased in the substantia nigra (PSN_D and PSN_ND < 0.0001) when comparing all Parkinson's disease patients to controls. R2(ISQR)∗ values in the substantia nigra significantly increased with the disease duration (PSN_D = 0.01; PSN_ND = 0.03) as well as the severity of the disease (Hoehn & Yahr scale <2 and ≥ 2, PSN_D = 0.02). Additionally, correlations between R2(ISQR)∗ and clinical features, mainly related to the severity of the disease, were found. Our results support the use of ISQR to reduce variations not directly related to Parkinson's disease, supporting the concept that ISQR strategy is useful for the evaluation of Parkinson's disease.

A decision support tool for early detection of knee OsteoArthritis using X-ray imaging and machine learning: Data from the OsteoArthritis Initiative.

This paper presents a fully developed computer aided diagnosis (CAD) system for early knee OsteoArthritis (OA) detection using knee X-ray imaging and machine learning algorithms. The X-ray images are first preprocessed in the Fourier domain using a circular Fourier filter. Then, a novel normalization method based on predictive modeling using multivariate linear regression (MLR) is applied to the data in order to reduce the variability between OA and healthy subjects. At the feature selection/extraction stage, an independent component analysis (ICA) approach is used in order to reduce the dimensionality. Finally, Naive Bayes and random forest classifiers are used for the classification task. This novel image-based approach is applied on 1024 knee X-ray images from the public database OsteoArthritis Initiative (OAI). The results show that the proposed system has a good predictive classification rate for OA detection (82.98% for accuracy, 87.15% for sensitivity and up to 80.65% for specificity).

Independent Component Analysis-Support Vector Machine-Based Computer-Aided Diagnosis System for Alzheimer's with Visual Support.

Computer-aided diagnosis (CAD) systems constitute a powerful tool for early diagnosis of Alzheimer's disease (AD), but limitations on interpretability and performance exist. In this work, a fully automatic CAD system based on supervised learning methods is proposed to be applied on segmented brain magnetic resonance imaging (MRI) from Alzheimer's disease neuroimaging initiative (ADNI) participants for automatic classification. The proposed CAD system possesses two relevant characteristics: optimal performance and visual support for decision making. The CAD is built in two stages: a first feature extraction based on independent component analysis (ICA) on class mean images and, secondly, a support vector machine (SVM) training and classification. The obtained features for classification offer a full graphical representation of the images, giving an understandable logic in the CAD output, that can increase confidence in the CAD support. The proposed method yields classification results up to 89% of accuracy (with 92% of sensitivity and 86% of specificity) for normal controls (NC) and AD patients, 79% of accuracy (with 82% of sensitivity and 76% of specificity) for NC and mild cognitive impairment (MCI), and 85% of accuracy (with 85% of sensitivity and 86% of specificity) for MCI and AD patients.

Linear intensity normalization of DaTSCAN images using Mean Square Error and a model-based clustering approach.

The analysis of 3D SPECT brain images requires several pre-processing steps such as intensity normalization and brain feature extraction. In this sense, a new method for intensity normalization of 123I-ioflupane-SPECT (DaTSCAN) brain images based on minimization of the Mean Square Error (MSE) between the Gaussian Mixture Model (GMM)-based extracted features from each subject image and a template in the so-defined non-specific region is derived. Our approach to feature extraction consists of using the set of parameters that define the template features, such as weights, covariance matrices and mean vectors to model the remaining images by reducing, consequently their dimensionality. The proposed method is compared to a widely used approach such as specific-to-non-specific binding ratio normalization. This comparison is performed on a DaTSCAN image database comprising analysis and classification stages for the development of a computer aided diagnosis (CAD) system for Parkinsonian syndrome (PS) detection.

Automatic classification of segmented MRI data combining Independent Component Analysis and Support Vector Machines.

This paper proposes a novel method for automatic classification of magnetic resonance images (MRI) based on independent component analysis (ICA). Our methodology consists of three processing steps. First, all the MRI scans are normalized and segmented into gray matter, white matter and cerebrospinal fluid. Then, ICA is applied to the preprocessed images for extracting relevant features which will be used as inputs to a support vector machine (SVM) classifier in order to reduce the feature space dimensionality. The system discriminates between Alzheimer's disease (AD) patients, mild cognitive impairment (MCI), and normal control (NC) subjects. All MRI data used in this work were obtained from the Alzheimers Disease Neuroimaging Initiative (ADNI). The experimental results showed that our methodology can successfully discriminate AD and MCI patients from NC subjects.