Bharti Rana

Regions-of-interest based automated diagnosis of Parkinson's disease using T1-weighted MRI

First automated ROI based computer aided diagnosis of PD using T1-weighted MRI.Analysis of GM, WM and CSF from five ROIs individually, in pairs and triplets.Evaluation on acquired age and gender matched dataset (30 PD and 30 healthy).Forward feature selection based on mutual information outperforms ranking method.Best classification accuracy of 86.67% is achieved with SN for GM and SN+HP for WM. Parkinsons disease (PD) is the second most common neurodegenerative disorder of the central nervous system. For its early management and accurate prognosis, there is a need to develop automated and non-invasive computer-aided diagnosis (CAD) technique(s). The present study proposes a novel region-of-interest (ROI) based CAD technique using T1-weighted magnetic resonance imaging (MRI) to discriminate PD patients from healthy subjects. A volumetric 3D T1-weighted (1mm isovoxel) MRI of 30 PD patients and age & gender matched 30 healthy subjects is acquired on a 1.5T MRI scanner. Five well-documented regions affected in PD, namely substantia nigra (SN), thalamus, hippocampus, frontal lobe (FL) and mid-brain are analyzed individually and in combinations of pairs and triplets. Features are constructed from gray matter (GM), white matter (WM) and cerebrospinal fluid (CSF) values of voxel from these regions. A small set of discriminating and non-redundant features is selected using mutual information based approach in conjunction with unpaired two-tailed two-sample t-test based ranking. A decision model is built with the help of support vector machine (SVM) as a classifier. The performance of the decision model is evaluated in terms of sensitivity, specificity and accuracy with leave-one-out cross-validation scheme. Experimental results demonstrate that the proposed method is able to differentiate PD from healthy subjects with a maximum accuracy of 86.67% with SN for GM and combination of SN & FL for WM; and outperforms the voxel-based morphometry method. Furthermore, loss in GM and WM volume and gain in CSF volume is observed in PD patients in comparison to healthy subjects. The excellent performance of the proposed method is beneficial for clinicians as it can be used as a decision support system which requires less time and efforts in diagnosing PD. In addition, it also encourages the application of CAD in medical domain.

Graph‐theory‐based spectral feature selection for computer aided diagnosis of Parkinson's disease using T1‐weighted MRI

Parkinsons disease (PD) is a neurological disorder, which is diagnosed on the basis of clinical history and examination alone as there are no diagnostic tests available. However, the current diagnosis highly depends on the knowledge and experience of clinicians and hence subjective in nature. Thus, the focus of this study is to develop a computer‐aided diagnosis (CAD) method using T1‐weighted magnetic resonance imaging (MRI) to differentiate PD from controls.

3d discrete wavelet transform for computer aided diagnosis of Alzheimer's disease using t1-weighted brain MRI

Early and antemortem diagnosis of Alzheimers disease (AD) may help in the development of appropriate treatment and in slowing down the disease progression. In this work, a three‐phase computer aided approach is suggested for classification of AD patients and controls using T1‐weighted MRI. In the first phase, smoothed modulated gray matter (GM) probability maps are obtained from T1‐weighted MRIs. In the second phase, 3D discrete wavelet transform is applied on GM of five brain regions, which are well‐documented regions affected in AD, to construct features. In the third phase, a minimal set of relevant and nonredundant features are obtained using Fishers discriminant ratio and minimum redundancy maximum relevance feature selection methods. To check the efficacy of the proposed approach, experiments were carried out on three datasets derived from the publicly available OASIS database, using three commonly used classifiers. The performance of the proposed approach was evaluated using three performance measures namely sensitivity, specificity and classification accuracy. Further, the proposed approach was compared with the existing state‐of‐the‐art techniques in terms of three performance measures, ROC curves, scoring and computation time. Irrespective of the datasets and the classifiers, the proposed method outperformed the existing methods. In addition, the statistical test also demonstrated that the proposed method is significantly better in comparison to the other existing methods. The appreciable performance of the proposed method supports that it will assist clinicians/researchers in the classification of AD patients and controls.

A Novel Approach for Computer Aided Diagnosis of Schizophrenia using Auditory Oddball Functional MRI

Schizophrenia is a serious mental illness that requires timely and accurate diagnosis. Functional magnetic resonance imaging (fMRI) helps in identifying variations in activation patterns of schizophrenia patients and healthy subjects. But, manual diagnosis using fMRI is cumbersome and prone to subjective errors. This has drawn the attention of pattern recognition and computer vision research community towards developing a reliable and efficient decision model for computer aided diagnosis (CAD) of schizophrenia. However, high dimensionality and limited availability of fMRI samples leads to curse-of-dimensionality which may deteriorate the performance of a decision model. In this research work, a combination of feature extraction and feature selection techniques is employed to obtain a reduced set of relevant features for differentiating schizophrenia patients from healthy subjects. A general linear model approach is used for feature extraction on pre-processed fMRI data. Further t-test based feature selection is employed to determine a subset of discriminative features which are used for learning a decision model using support vector machine. Experiments are carried out on two balanced and well-age matched datasets (acquired on 1.5 Tesla and 3 Tesla scanners) of auditory oddball task derived from a publicly available multisite FBIRN dataset. The performance is evaluated in terms of sensitivity, specificity and classification accuracy, and compared with two well-known existing approaches. Experimental results demonstrate that the proposed model outperforms the two existing approaches in terms of sensitivity, specificity and classification accuracy. With the proposed approach, the classification accuracy of 80.9% and 88.0% is achieved for 1.5 Tesla and 3 Tesla datasets respectively. In addition, the brain regions containing the discriminative features are identified which may be used as biomarkers for CAD of schizophrenia using fMRI.

Salient features selection for multiclass texture classification

Texture classification is one of the important components in texture analysis which has drawn the attention of research community during the past few decades. Various texture feature extraction techniques have been proposed in the literature. However, combining texture methods from different families has demonstrated to produce better classification at the cost of complexity of the learning model. In this paper, we have investigated three parametric test statistics (ANOVA F statistic, Welch test statistic, Adjusted Welch test statistic) to determine salient features for multiclass texture classification. The salient features are obtained from a pool of features obtained using five textural feature extraction methods. Experiments are performed on a widely used publicly available Brodatz dataset. Experimental results show that the classification error decreases significantly with the use of all the three feature selection methods with all classifiers. The reduced set of features will also lead to significant decrease in computation time of the learning model.

Relevant 3D local binary pattern based features from fused feature descriptor for differential diagnosis of Parkinson’s disease using structural MRI

Computer-aided diagnosis (CAD) of Parkinson’s disease (PD) using structural magnetic resonance imaging is an emerging research field for the pattern recognition community. The existing research works have utilized gray matter, white matter and cerebrospinal fluid tissues individually for diagnosis of PD and have ignored the intercorrelation among the three tissues. Thus, there is a need to define a fused feature descriptor (FFD) which can capture information and intercorrelation of all the three tissues simultaneously, and to further enhance the performance of CAD. The present study proposes a simple and efficient FFD, in terms of all the three tissues, for CAD of PD. Each brain volume is represented in terms of the FFD. Then each fused volume is segmented into 118 brain regions. Thereafter, features extraction is carried out from each brain region using 3D local binary pattern. Then, a set of discriminating and uncorrelated features are identified using t-test in conjunction with minimum redundancy maximum relevance feature selection method. Finally, support vector machine is utilized to build a decision model. Volumetric 3D T1-weighted magnetic resonance imaging dataset (age & gender matched 30 PD and 30 healthy subjects) is acquired using 1.5T machine and is utilized to investigate the efficacy of the proposed method. The classification accuracy of 95% is achieved using leave-one-out cross-validation scheme which is superior to the existing methods. Regions namely Hippocampus_R, Cingulum_Mid_L, Frontal_Inf_Tri_L, Precentral_R, Precentral_L, Frontal_Mid_L, Frontal_Mid_Orb_L, Cingulum_Ant_L and Hippocampus_L, are observed to be the most discriminative regions for diagnosis of PD. The notable performance of the proposed method suggests that instead of studying the three tissues independently, their intercorrelation should also be considered. Further, the proposed method may be employed as a diagnostic tool for diagnosis of PD.

Voxel-based morphometry and minimum redundancy maximum relevance method for classification of Parkinson's disease and controls from T1-weighted MRI

Parkinsons disease (PD) is a neurodegenerative disorder, which needs to be accurately diagnosed in early stage. Voxel-based morphometry (VBM) has been extensively utilized to determine focal changes between PD patients and controls. However, it is not much utilized in differential diagnosis of an individual subject. Thus, in this study, VBM findings in conjunction with minimum redundancy maximum relevance (mRMR) method are utilized to obtain a set of relevant and non-redundant features for computer-aided diagnosis (CAD) of PD using T1-weighted MRI. In the proposed method, firstly, statistical features are extracted from the clusters obtained from statistical maps, generated using VBM, of gray matter (GM), white matter (WM) and cerebrospinal fluid (CSF) independently and their different combinations. Then mRMR, a multivariate feature selection method, is utilized to find a minimal set of relevant and non-redundant features. Finally, support vector machine is utilized to learn a decision model using the selected features. Experiments are performed on newly acquired T1-weighted MRI of 30 PD patients and 30 age & gender matched controls. The performance is evaluated using leave-one out cross-validation scheme in terms of sensitivity, specificity and classification accuracy. The maximum accuracy of 88.33% is achieved for GM+WM and GM+WM+CSF. In addition, the proposed method outperforms the existing methods. It is also observed that the selected clusters belong to regions namely middle and superior frontal gyrus for GM, inferior, middle frontal gyrus and insula for WM and lateral ventricle for CSF. Further, correlation of UPDRS/H&Y staging scale with GM/WM/CSF volume is observed to be not significant. Appreciable classification performance of the proposed method highlights the potential of the proposed method in CAD support system for the clinicians in PD diagnosis.

A Novel Approach for Classification of Schizophrenia Patients and Healthy Subjects Using Auditory Oddball Functional MRI

Schizophrenia is a serious psychiatric illness which needs early and accurate diagnosis. Difference in activation patterns of schizophrenia patients and healthy subjects can be identified with the help of functional magnetic resonance imaging (fMRI). However, manual diagnosis using fMRI depends on subjective observation and may be erroneous. This has motivated the pattern recognition and machine learning research community to develop a reliable and efficient decision model for classification of schizophrenia patients and healthy subjects. However, high dimensionality and low availability of fMRI data leads to the curse-of-dimensionality problem which may degrade the performance of decision model. In the present research work, a combination of feature extraction and feature selection techniques is utilised to obtain a reduced set of relevant and non-redundant features for classification of schizophrenia patients and healthy subjects. Features are extracted from pre-processed fMRI data using a general linear model approach. Next Fishers discriminant ratio, a univariate method, is employed for feature selection i.e. To determine a subset of discriminative features. Further, minimum Redundancy Maximum Relevance (mRMR) feature selection, a multivariate method, is employed to obtain a set of relevant and non-redundant features which are used for learning a decision model using support vector machine. Two balanced and well-age matched datasets of auditory oddball task derived from a publicly available multisite FBIRN dataset are used for experiments. First dataset consists of fMRI scans of 34 schizophrenia patients and 34 healthy subjects acquired through 1.5 Tesla scanners while second dataset consists of 25 schizophrenia patients and 25 healthy subjects acquired through 3 Tesla scanners. The performance is evaluated in terms of sensitivity, specificity and classification accuracy, and compared with two well-known existing approaches for classification of schizophrenia patients and healthy subjects using fMRI. Experimental results demonstrate that the proposed model outperforms the existing approaches in terms of sensitivity, specificity and classification accuracy. The proposed approach achieves classification accuracy of 88.2% and 78.0% for 1.5 Tesla and 3 Tesla datasets respectively. In addition, the brain regions containing the discriminative features are identified which may be potential biomarkers for diagnosis of schizophrenia using fMRI.

A novel fuzzy rough selection of non-linearly extracted features for schizophrenia diagnosis using fMRI.

BACKGROUND AND OBJECTIVES Schizophrenia is a severe brain disorder primarily diagnosed through externally observed behavioural symptoms due to the dearth of established clinical tests. Functional magnetic resonance imaging (fMRI) can capture the distortions caused by schizophrenia in the brain activation. Hence, it can be useful for developing a decision model that performs computer-aided diagnosis of schizophrenia. But, fMRI data is huge in dimension. Therefore dimension reduction is indispensable. It is additionally required to identify the discriminative brain regions. Hence, we aim to build an effective decision model that incorporates suitable dimension reduction and also identifies discriminative brain regions. METHODS We propose a three-phase dimension reduction. First phase involves spatially-constrained fuzzy clustering of 3-dimensional spatial maps (obtained from general linear model and independent component analysis). In the second phase, non-linear features are extracted from each cluster using a generalized discriminant analysis. In the third phase, a novel fuzzy rough feature selection is proposed. The features obtained after the third phase are used for learning a decision model by the help of support vector machine classifier. This complete method is implemented within leave-one-out cross-validation on two balanced datasets (respectively acquired on 1.5Tesla and 3Tesla scanners). Both these datasets are created using Function Biomedical Informatics Research Network multisite data and contain fMRI data acquired during auditory oddball task performed by age-matched schizophrenia patients and healthy subjects. A permutation test is also carried out to ensure that no bias is involved in the learning. RESULTS The results indicate that the proposed method achieves maximum classification accuracy of 97.1% and 98.0% for the two datasets respectively. The proposed method outperforms the state-of-the-art methods. The results of the permutation test show that p-values are lesser than the significance level i.e. 0.05. Therefore, the classifier has found a significant class structure and does not involve any bias. Further, discriminative brain regions are identified and are in agreement with the findings in related literature. CONCLUSION The proposed method is able to derive suitable non-linear features and the related brain regions for effective computer-aided diagnosis. The fuzzy and rough set based approaches help in handling uncertainty and ambiguity in real data.

An enhanced texture-based image retrieval approach with features selected from integration of feature extraction techniques

Texture is vital in characterising images for effective content-based image retrieval. Integrating features from various feature extraction techniques improves the performance of decision system in comparison to individual techniques as it provides complimentary information as a whole. However, this integration creates a large feature vector which may contain irrelevant and redundant features and hence degrade the performance. Therefore, we propose a three-phase texture-based image retrieval approach for enhanced performance. In the first phase, pool of texture features from seven feature extraction techniques is created. In the second phase, some popular feature selection techniques are applied to this pool to obtain a reduced set of relevant and non-redundant features. In the third phase, three well-known distance measures are utilised to retrieve images based on the reduced features set. The performance of the proposed approach is evaluated on Brodatz dataset. The proposed approach outperforms individual feature extraction techniques.