Zhengchao Dong

A hybrid method for MRI brain image classification

Automated and accurate classification of MR brain images is of importance for the analysis and interpretation of these images and many methods have been proposed. In this paper, we present a neural network (NN) based method to classify a given MR brain image as normal or abnormal. This method first employs wavelet transform to extract features from images, and then applies the technique of principle component analysis (PCA) to reduce the dimensions of features. The reduced features are sent to a back propagation (BP) NN, with which scaled conjugate gradient (SCG) is adopted to find the optimal weights of the NN. We applied this method on 66 images (18 normal, 48 abnormal). The classification accuracies on both training and test images are 100%, and the computation time per image is only 0.0451s.

Preclinical Diagnosis of Magnetic Resonance (MR) Brain Images via Discrete Wavelet Packet Transform with Tsallis Entropy and Generalized Eigenvalue Proximal Support Vector Machine (GEPSVM)

Background: Developing an accurate computer-aided diagnosis (CAD) system of MR brain images is essential for medical interpretation and analysis. In this study, we propose a novel automatic CAD system to distinguish abnormal brains from normal brains in MRI scanning. Methods: The proposed method simplifies the task to a binary classification problem. We used discrete wavelet packet transform (DWPT) to extract wavelet packet coefficients from MR brain images. Next, Shannon entropy (SE) and Tsallis entropy (TE) were harnessed to obtain entropy features from DWPT coefficients. Finally, generalized eigenvalue proximate support vector machine (GEPSVM), and GEPSVM with radial basis function (RBF) kernel, were employed as classifier. We tested the four proposed diagnosis methods (DWPT + SE + GEPSVM, DWPT + TE + GEPSVM, DWPT + SE + GEPSVM + RBF, and DWPT + TE + GEPSVM + RBF) on three benchmark datasets of Dataset-66, Dataset-160, and Dataset-255. Results: The 10 repetition of K-fold stratified cross validation results showed the proposed DWPT + TE + GEPSVM + RBF method excelled not only other three proposed classifiers but also existing state-of-the-art methods in terms of classification accuracy. In addition, the DWPT + TE + GEPSVM + RBF method achieved accuracy of 100%, 100%, and 99.53% on Dataset-66, Dataset-160, and Dataset-255, respectively. For Dataset-255, the offline learning cost 8.4430s and online prediction cost merely 0.1059s. Conclusions: We have proved the effectiveness of the proposed method, which achieved nearly 100% accuracy over three benchmark datasets.

Classification of Alzheimer Disease Based on Structural Magnetic Resonance Imaging by Kernel Support Vector Machine Decision Tree

In this paper we proposed a novel classiflcation system to distinguish among elderly subjects with Alzheimers disease (AD), mild cognitive impairment (MCI), and normal controls (NC). The method employed the magnetic resonance imaging (MRI) data of 178 subjects consisting of 97NCs, 57MCIs, and 24ADs. First, all these three dimensional (3D) MRI images were preprocessed with atlas-registered normalization. Then, gray matter images were extracted and the 3D images were under-sampled. Afterwards, principle component analysis was applied for feature extraction. In total, 20 principal components (PC) were extracted from 3D MRI data using singular value decomposition (SVD) algorithm, and 2 PCs were extracted from additional information (consisting of demographics, clinical examination, and derived anatomic volumes) using alternating least squares (ALS). On the basic of the 22 features, we constructed a kernel support vector machine decision tree (kSVM-DT). The error penalty parameter C and kernel parameter ae were determined by Particle Swarm Optimization (PSO). The weights ! and biases b were still obtained by quadratic programming method. 5-fold cross validation was employed to obtain the out-of-sample estimate. The results show that the proposed kSVM-DT achieves 80% classiflcation accuracy, better than 74% of the method without kernel. Besides, the PSO exceeds the random selection method in choosing the parameters of the classifler. The computation time to predict a new patient is only 0.022s.

Detection of subjects and brain regions related to Alzheimer's disease using 3D MRI scans based on eigenbrain and machine learning

Purpose: Early diagnosis or detection of Alzheimers disease (AD) from the normal elder control (NC) is very important. However, the computer-aided diagnosis (CAD) was not widely used, and the classification performance did not reach the standard of practical use. We proposed a novel CAD system for MR brain images based on eigenbrains and machine learning with two goals: accurate detection of both AD subjects and AD-related brain regions. Method: First, we used maximum inter-class variance (ICV) to select key slices from 3D volumetric data. Second, we generated an eigenbrain set for each subject. Third, the most important eigenbrain (MIE) was obtained by Welchs t-test (WTT). Finally, kernel support-vector-machines with different kernels that were trained by particle swarm optimization, were used to make an accurate prediction of AD subjects. Coefficients of MIE with values higher than 0.98 quantile were highlighted to obtain the discriminant regions that distinguish AD from NC. Results: The experiments showed that the proposed method can predict AD subjects with a competitive performance with existing methods, especially the accuracy of the polynomial kernel (92.36 ± 0.94) was better than the linear kernel of 91.47 ± 1.02 and the radial basis function (RBF) kernel of 86.71 ± 1.93. The proposed eigenbrain-based CAD system detected 30 AD-related brain regions (Anterior Cingulate, Caudate Nucleus, Cerebellum, Cingulate Gyrus, Claustrum, Inferior Frontal Gyrus, Inferior Parietal Lobule, Insula, Lateral Ventricle, Lentiform Nucleus, Lingual Gyrus, Medial Frontal Gyrus, Middle Frontal Gyrus, Middle Occipital Gyrus, Middle Temporal Gyrus, Paracentral Lobule, Parahippocampal Gyrus, Postcentral Gyrus, Posterial Cingulate, Precentral Gyrus, Precuneus, Subcallosal Gyrus, Sub-Gyral, Superior Frontal Gyrus, Superior Parietal Lobule, Superior Temporal Gyrus, Supramarginal Gyrus, Thalamus, Transverse Temporal Gyrus, and Uncus). The results were coherent with existing literatures. Conclusion: The eigenbrain method was effective in AD subject prediction and discriminant brain-region detection in MRI scanning.

Mitochondrial Dysfunction as a Neurobiological Subtype of Autism Spectrum Disorder: Evidence From Brain Imaging

IMPORTANCE Impaired mitochondrial function impacts many biological processes that depend heavily on energy and metabolism and can lead to a wide range of neurodevelopmental disorders, including autism spectrum disorder (ASD). Although evidence that mitochondrial dysfunction is a biological subtype of ASD has grown in recent years, no study, to our knowledge, has demonstrated evidence of mitochondrial dysfunction in brain tissue in vivo in a large, well-defined sample of individuals with ASD. OBJECTIVES To assess brain lactate in individuals with ASD and typically developing controls using high-resolution, multiplanar spectroscopic imaging; to map the distribution of lactate in the brains of individuals with ASD; and to assess correlations of elevated brain lactate with age, autism subtype, and intellectual ability. DESIGN, SETTING, AND PARTICIPANTS Case-control study at Columbia University Medical Center and New York State Psychiatric Institute involving 75 children and adults with ASD and 96 age- and sex-matched, typically developing controls. MAIN OUTCOMES AND MEASURES Lactate doublets (present or absent) on brain magnetic resonance spectroscopic imaging. RESULTS Lactate doublets were present at a significantly higher rate in participants with ASD (13%) than controls (1%) (P = .001). In the ASD group, the presence of lactate doublets correlated significantly with age (P = .004) and was detected more often in adults (20%) than in children (6%), though it did not correlate with sex, ASD subtype, intellectual ability, or the Autism Diagnostic Observation Schedule total score or subscores. In those with ASD, lactate was detected most frequently within the cingulate gyrus but it was also present in the subcortical gray matter nuclei, corpus callosum, superior temporal gyrus, and pre- and postcentral gyri. CONCLUSIONS AND RELEVANCE In vivo brain findings provide evidence for a possible neurobiological subtype of mitochondrial dysfunction in ASD.

Automated classification of brain images using wavelet-energy and biogeography-based optimization

It is very important to early detect abnormal brains, in order to save social and hospital resources. The wavelet-energy was a successful feature descriptor that achieved excellent performances in various applications; hence, we proposed a novel wavelet-energy based approach for automated classification of MR brain images as normal or abnormal. SVM was used as the classifier, and biogeography-based optimization (BBO) was introduced to optimize the weights of the SVM. The results based on a 5 × 5-fold cross validation showed the performance of the proposed BBO-KSVM was superior to BP-NN, KSVM, and PSO-KSVM in terms of sensitivity and accuracy. The study offered a new means to detect abnormal brains with excellent performance.

Pathological Brain Detection in Magnetic Resonance Imaging Scanning by Wavelet Entropy and Hybridization of Biogeography-Based Optimization and Particle Swarm Optimization

Background) We proposed a novel computer-aided diagnosis (CAD) system based on the hybridization of biogeography-based optimization (BBO) and particle swarm optimization (PSO), with the goal of detecting pathological brains in MRI scanning. (Method) The proposed method used wavelet entropy (WE) to extract features from MR brain images, followed by feed-forward neural network (FNN) with training method of a Hybridization of BBO and PSO (HBP), which combined the exploration ability of BBO and exploitation ability of PSO. (Results) The 10 repetition of k-fold cross validation result showed that the proposed HBP outperformed existing FNN training methods and that the proposed WE + HBP-FNN outperformed fourteen state-of-the-art CAD systems of MR brain classification in terms of classification accuracy. The proposed method achieved accuracy of 100%, 100%, and 99.49% over Dataset-66, Dataset-160, and Dataset-255, respectively. The offline learning cost 208.2510 s for Dataset-255, and merely 0.053s for online prediction. (Conclusion) The proposed WE + HBP-FNN method achieves nearly perfect detection pathological brains in MRI scanning.

An MR Brain Images Classifier System via Particle Swarm Optimization and Kernel Support Vector Machine

Automated abnormal brain detection is extremely of importance for clinical diagnosis. Over last decades numerous methods had been presented. In this paper, we proposed a novel hybrid system to classify a given MR brain image as either normal or abnormal. The proposed method first employed digital wavelet transform to extract features then used principal component analysis (PCA) to reduce the feature space. Afterwards, we constructed a kernel support vector machine (KSVM) with RBF kernel, using particle swarm optimization (PSO) to optimize the parameters C and σ. Fivefold cross-validation was utilized to avoid overfitting. In the experimental procedure, we created a 90 images dataset brain downloaded from Harvard Medical School website. The abnormal brain MR images consist of the following diseases: glioma, metastatic adenocarcinoma, metastatic bronchogenic carcinoma, meningioma, sarcoma, Alzheimer, Huntington, motor neuron disease, cerebral calcinosis, Picks disease, Alzheimer plus visual agnosia, multiple sclerosis, AIDS dementia, Lyme encephalopathy, herpes encephalitis, Creutzfeld-Jakob disease, and cerebral toxoplasmosis. The 5-folded cross-validation classification results showed that our method achieved 97.78% classification accuracy, higher than 86.22% by BP-NN and 91.33% by RBF-NN. For the parameter selection, we compared PSO with those of random selection method. The results showed that the PSO is more effective to build optimal KSVM.

Exponential Wavelet Iterative Shrinkage Thresholding Algorithm for compressed sensing magnetic resonance imaging

It is beneficial for both hospitals and patients to accelerate MRI scanning. Recently, a new fast MRI technique based on CS was proposed. However, the reconstruction quality and computation time of CS-MRI did not meet the standard of clinical use. Therefore, we proposed a novel algorithm based on three successful components: the sparsity of EWT, the rapidness of FISTA, and the excellent tuning in SISTA. The proposed method was dubbed Exponential Wavelet Iterative Shrinkage/Threshold Algorithm (EWISTA). Experiments over four kinds of MR images (brain, ankle, knee, and ADHD) indicated that the proposed EWISTA showed better reconstruction performance than the state-of-the-art algorithms such as FCSA, ISTA, FISTA, SISTA, and EWT-ISTA. Moreover, EWISTA was faster than ISTA and EWT-ISTA, but slightly slower than FCSA, FISTA and SISTA.

Effects of Davunetide on N-acetylaspartate and Choline in Dorsolateral Prefrontal Cortex in Patients with Schizophrenia

Schizophrenia is associated with extensive neurocognitive and behavioral impairments. Studies indicate that N-acetylaspartate (NAA), a marker of neuronal integrity, and choline, a marker of cell membrane turnover and white matter integrity, may be altered in schizophrenia. Davunetide is a neurotrophic peptide that can enhance cognitive function in animal models of neurodegeneration. Davunetide has recently demonstrated modest functional improvement in a study of people with schizophrenia. In a subset of these subjects, proton magnetic resonance spectroscopy (1H-MRS) was conducted to explore the effects of davunetide on change in NAA/creatine (NAA/Cr) and choline/creatine (choline/Cr) over 12 weeks of treatment. Of 63 outpatients with schizophrenia who received randomized davunetide (5 and 30 mg/day) or placebo in the parent clinical trial, 18 successfully completed 1H-MRS in dorsolateral prefrontal cortex (DLPFC) at baseline and at 12 weeks. Cognition was assessed using the MATRICS Consensus Cognitive Battery (MCCB). NAA/Cr was unchanged for combined high- and low-dose davunetide groups (N=11). NAA/Cr in the high-dose davunetide group (N=8) suggested a trend increase of 8.0% (P=0.072) over placebo (N=7). Choline/Cr for combined high- and low-dose davunetide groups suggested a 6.4% increase (P=0.069), while the high-dose group showed a 7.9% increase (P=0.040) over placebo. Baseline NAA/Cr correlated with the composite MCCB score (R=0.52, P=0.033), as did individual cognitive domains of attention/vigilance, verbal learning, and social cognition; however, neither metabolite correlated with functional capacity. In this exploratory study, 12 weeks of adjunctive davunetide appeared to produce modest increases in NAA/Cr and choline/Cr in DLPFC in people with schizophrenia. This is consistent with a potential neuroprotective mechanism for davunetide. The data also support use of MRS as a useful biomarker of baseline cognitive function in schizophrenia. Future clinical and preclinical studies are needed to fully define the mechanism of action and cognitive effects of davunetide in schizophrenia.