- Dr. Jac Fredo received his doctorate in Electrical Engineering from Anna University (India) and continued his researc... moreDr. Jac Fredo received his doctorate in Electrical Engineering from Anna University (India) and continued his research as a Research Associate at the Non-Invasive Imaging and Diagnostics Laboratory, Indian Institute of Technology Madras (India). He received a DST-MANF fellowship to pursue his Ph.D and a SERB-IUSSTF fellowship to pursue his post-doctoral research at Brain Development Imaging Lab (BDIL), San Diego State University (SDSU), USA. He continually seeks to broaden his scientific experiences, visiting labs in the USA, Germany, and The Netherlands through a series of travel awards. His main interests are neuroimage processing, computer vision techniques and machine learning.edit
BACKGROUND Autism Spectrum Disorder (ASD) is a neurodevelopmental disability with altered connectivity in brain networks. NEW METHOD In this study, brain connections in Resting-state functional Magnetic Resonance Imaging (Rs-fMRI) of ASD... more
BACKGROUND Autism Spectrum Disorder (ASD) is a neurodevelopmental disability with altered connectivity in brain networks. NEW METHOD In this study, brain connections in Resting-state functional Magnetic Resonance Imaging (Rs-fMRI) of ASD and Typical Developing (TD) are analyzed by partial and full correlation methods such as Gaussian Graphical Least Absolute Shrinkage and Selection Operator (GLASSO), Max-Det Matrix Completion (MDMC), and Pearson Correlation Co-Efficient (PCCE). We investigated Functional Connectivity (FC) of ASD and TD brain from 238 functionally defined regions of interest. Furthermore, we constructed a series of feature sets by applying conditional random forests and conditional permutation importance. We built classifier models by Random Forest (RF), Oblique RF (ORF), Support Vector Machine (SVM), and Convolutional Neural Network (CNN) for each feature set. FC features are ranked based on p-value and we analyzed the top 20 FC features. RESULTS We achieved a single-trial test accuracy of 72.5%, though MDMC-SVM and PCCE-CNN pipelines. Further, PCCE-CNN pipeline gives better average test accuracy (70.31%) and area under the curve (0.73) compared to other pipelines. We found that top-20 PCCE based FC features are from networks such as Dorsal Attention (DA), Cingulo-Opercular Task Control (COTC), somatosensory motor hand and subcortical. In addition, among top 20 PCCE features, many FC links are found between COTC and DA (4 connections) which helped to discriminate the ASD and TD. COMPARISON WITH EXISTING METHODS AND CONCLUSIONS The generalized classifier models built in our study for highly heterogeneous participants perform better than previous studies with similar data sets and diagnostic groups.
Research Interests:
Autism spectrum disorders (ASDs) are heterogeneous neurodevelopmental conditions. In fMRI studies, including most machine learning studies seeking to distinguish ASD from typical developing (TD) samples, cohorts differing in gender and... more
Autism spectrum disorders (ASDs) are heterogeneous neurodevelopmental conditions. In fMRI studies, including most machine learning studies seeking to distinguish ASD from typical developing (TD) samples, cohorts differing in gender and symptom severity composition are often treated statistically as one ‘ASD group.’ Using resting-state functional connectivity (FC) data, we implemented random forest to build diagnostic classifiers in four ASD samples including a total of 656 participants (NASD = 306, NTD = 350, ages 6–18). Groups were manipulated to titrate heterogeneity of gender and symptom severity and partially overlapped. Each sample differed on inclusionary criteria: (1) all genders, unrestricted severity range; (2) only male participants, unrestricted severity; (3) all genders, higher severity only; and (4) only male participants, higher severity. Each set consisted of 200 participants per group (ASD, TD; matched on age and head motion): 160 for training and 40 for validation. FMRI time series from 237 regions of interest (ROIs) were Pearson correlated in a 237 × 237 FC matrix, and classifiers were built using random forest in training samples. Classification accuracies in validation samples were 62.5%, 65%, 70%, and 73.75%, respectively, for samples 1–4. Connectivity within cingulo-opercular task control (COTC) network, and between COTC ROIs and default mode and dorsal attention network contributed overall most informative features, but features differed across sets. Findings suggest that diagnostic classifiers vary depending on ASD sample composition. Specifically, greater homogeneity of samples regarding gender and symptom severity enhances classifier performance. However, given the true heterogeneity of ASDs, performance metrics alone may not adequately reflect classifier utility.
Research Interests:
Abstract In this work, the strength of the composite material is tested and the damages are classified using supervised method. The image is obtained from the front and rear sides of the composite material after applying 5 mm, 6 mm and... more
Abstract In this work, the strength of the composite material is tested and the damages are classified using supervised method. The image is obtained from the front and rear sides of the composite material after applying 5 mm, 6 mm and 7 mm impingement. Initially, the images are filtered using anisotropic diffusion filter. Global and local damages in the structures are segmented using Fuzzy C-Means (FCM) clustering method. Geometrical features and Zernike Moments (ZM) are calculated from the delineated regions. The performance of the features is tested using Support Vector Machine classifier. Results show that the FCM with three and four cluster centres is able to segment the global and local damages respectively. The global damages due to different impinges are classified better compared to the local damages. The global damages in the rear side are able to classify better compared to the front side in both geometrical and ZM features. In the case of local damages, the rear side is able to classify better in 5 mm–6 mm and front side in 6 mm–7 mm. It is concluded that the features obtained from the ZM gives better accuracy in both global and local damages compared to the geometrical features. The image based analysis carried out on this work is able to classify the impairment in composite materials; this framework can be used in the industrial applications for the quantification of damages.
Research Interests:
ABSTRACT In this work, subcortical regions of autism spectrum disorder are analysed using fuzzy Gaussian distribution model-based distance regularised multi-phase level set method in autistic MR brain images. The fuzzy Gaussian... more
ABSTRACT In this work, subcortical regions of autism spectrum disorder are analysed using fuzzy Gaussian distribution model-based distance regularised multi-phase level set method in autistic MR brain images. The fuzzy Gaussian distribution model is used as the intensity discriminator. The segmented images are validated with the ground truth using geometrical measure area. The results show that the fuzzy Gaussian distribution model-based multi-phase level set method is able to extract the subcortical tissue boundaries. The subcortical regions segmented using this method gives high correlation with ground truth. The corpus callosum area gives very high (R = 0.94) correlation. The brain stem and cerebellum present high correlations of 0.89 and 0.84, respectively. Also, it is found the segmented autistic subcortical regions have reduced area and are statistically significant (p < 0.0001). The ratio metric analysis proves the relation in reduction of the area in subcortical regions with total brain area.
Research Interests:
Research Interests:
ABSTRACT In this work, subcortical regions of autistic magnetic resonance brain images are analyzed using multiphase level set method. The images considered in this work are obtained from autism brain image data exchange database. The... more
ABSTRACT In this work, subcortical regions of autistic magnetic resonance brain images are analyzed using multiphase level set method. The images considered in this work are obtained from autism brain image data exchange database. The subcortical regions such as corpus callosum, cerebellum, and brain stem are segmented from the cortical region using Fuzzy c-means (FCM)-based multiphase level set method. FCM with three cluster center is used as the intensity discriminator and the evolution of the level set curve is regularized by a distance function. The results show that the multiphase level set method is able to segment the desired subcortical regions. The results are validated with the ground truth images. The average similarity values are found to be 0.85. The segmented subcortical regions of autistic have reduced tissue area and are distinct from the controls (p < 0.0001). Further, it is observed that the subcortical area gives comparable results with clinical intelligent quotient values and is able to discriminate the controls and autistic subjects. As the feature area extracted from brain subcortical regions are significant, this study seems to be clinically helpful in mass screening of autistic subjects. © 2014 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 24, 256–262, 2014
Research Interests:
Research Interests:
In this work, the segmentation and analysis of the corpus callosum (CC) in autistic MR brain images is carried out using a fuzzy c-means (FCM)-based level set method. Initially, the images are skull-stripped using the geodesic active... more
In this work, the segmentation and analysis of the corpus callosum (CC) in autistic MR brain images is carried out using a fuzzy c-means (FCM)-based level set method. Initially, the images are skull-stripped using the geodesic active contour method. The CC is extracted from the skull-stripped images using the FCM-based level set method. FCM clustering forms the initial contour. The evolution of the curve is then regularized using a distance function in the level sets. The segmented CC is divided into five segments, whose areas are measured. The subjective results show that the proposed method is able to extract the CC from skull-stripped images. It is demonstrated that the level set with the FCM as the initial contour gives better results than those obtained with a manual initial contour. It is found that the autistic subjects have a reduced CC area compared to that of control subjects. The total CC area of autistic subjects gives a correlation of R = 0.39 with the verbal intelligence quotient (IQ) values. Further analysis shows that the anterior third region of the CC gives significant discrimination of the control and autistic subjects compared to the other segments. Its correlation (R) with verbal IQ is found to be 0.27 in autistic subjects. The feature area extracted from the CC and its segments are significant, hence the results may be clinically helpful in the mass screening of autistic subjects.
Research Interests:
Autism spectrum disorders (ASDs) are heterogeneous neurodevelopmental conditions. In fMRI studies, including most machine learning studies seeking to distinguish ASD from typical developing (TD) samples, cohorts differing in gender and... more
Autism spectrum disorders (ASDs) are heterogeneous neurodevelopmental conditions. In fMRI studies, including most machine learning studies seeking to distinguish ASD from typical developing (TD) samples, cohorts differing in gender and symptom severity composition are often treated statistically as one 'ASD group.' Using resting-state functional connectivity (FC) data, we implemented random forest to build diagnostic classifiers in four ASD samples including a total of 656 participants (N ASD = 306, N TD = 350, ages 6-18). Groups were manipulated to titrate heterogeneity of gender and symptom severity and partially overlapped. Each sample differed on inclusionary criteria: (1) all genders, unrestricted severity range; (2) only male participants, unrestricted severity; (3) all genders, higher severity only; and (4) only male participants, higher severity. Each set consisted of 200 participants per group (ASD, TD; matched on age and head motion): 160 for training and 40 for validation. FMRI time series from 237 regions of interest (ROIs) were Pearson correlated in a 237 9 237 FC matrix, and classifiers were built using random forest in training samples. Classification accuracies in validation samples were 62.5%, 65%, 70%, and 73.75%, respectively, for samples 1-4. Connectivity within cinguloopercular task control (COTC) network, and between COTC ROIs and default mode and dorsal attention network contributed overall most informative features, but features differed across sets. Findings suggest that diagnostic classifiers vary depending on ASD sample composition. Specifically, greater homogeneity of samples regarding gender and symptom severity enhances classifier performance. However, given the true heterogeneity of ASDs, performance metrics alone may not adequately reflect classifier utility.
Background: Autism Spectrum Disorder (ASD) is a neurodevelopmental disability with altered connectivity in brain networks. New method: In this study, brain connections in Resting-state functional Magnetic Resonance Imaging (Rs-fMRI) of... more
Background: Autism Spectrum Disorder (ASD) is a neurodevelopmental disability with altered connectivity in brain networks. New method: In this study, brain connections in Resting-state functional Magnetic Resonance Imaging (Rs-fMRI) of ASD and Typical Developing (TD) are analyzed by partial and full correlation methods such as Gaussian Graphical Least Absolute Shrinkage and Selection Operator (GLASSO), Max-Det Matrix Completion (MDMC), and Pearson Correlation CoEfficient (PCCE). We investigated Functional Connectivity (FC) of ASD and TD brain from 238 functionally defined regions of interest. Furthermore, we constructed a series of feature sets by applying conditional random forests and conditional permutation importance. We built classifier models by Random Forest (RF), Oblique RF (ORF), Support Vector Machine (SVM), and Convolutional Neural Network (CNN) for each feature set. FC features are ranked based on p-value and we analyzed the top 20 FC features. Results: We achieved a single-trial test accuracy of 72.5 %, though MDMC-SVM and PCCE-CNN pipelines. Further, PCCE-CNN pipeline gives better average test accuracy (70.31 %) and area under the curve (0.73) compared to other pipelines. We found that top-20 PCCE based FC features are from networks such as Dorsal Attention (DA), Cingulo-Opercular Task Control (COTC), somatosensory motor hand and subcortical. In addition, among top 20 PCCE features, many FC links are found between COTC and DA (4 connections) which helped