Epifanio Bagarinao

Estimation of general linear model coefficients for real-time application

An algorithm using an orthogonalization procedure to estimate the coefficients of general linear models (GLM) for functional magnetic resonance imaging (fMRI) calculations is described. The idea is to convert the basis functions or explanatory variables of a GLM into orthogonal functions using the usual Gram-Schmidt orthogonalization procedure. The coefficients associated with the orthogonal functions, henceforth referred to as auxiliary coefficients, are then easily estimated by applying the orthogonality condition. The original GLM coefficients are computed from these estimates. With this formulation, the estimates can be updated when new image data become available, making the approach applicable for real-time estimation. Since the contribution of each image data is immediately incorporated into the estimated values, storing the data in memory during the estimation process becomes unnecessary, minimizing the memory requirements of the estimation process. By employing Cholesky decomposition, the algorithm is a factor of two faster than the standard recursive least-squares approach. Results of the analysis of an fMRI study using this approach showed the algorithms potential for real-time application.

Application of independent component analysis to magnetic resonance imaging for enhancing the contrast of gray and white matter.

An application of independent component analysis (ICA) was attempted to develop a method of processing magnetic resonance (MR) images to extract physiologically independent components representing tissue relaxation times and achieve improved visualization of normal and pathologic structures. Anatomical T1-weighted, T2-weighted and proton density images were obtained from 10 normal subjects, 3 patients with brain tumors and 1 patient with multiple sclerosis. The data sets were analyzed using ICA based on the learning rule of Bell and Sejnowski after prewhitening operations. The three independent components obtained from the three original data sets corresponded to (1) short T1 components representing myelin of white matter and lipids, (2) relatively short T1 components representing gray matter and (3) long T2 components representing free water. The involvement of gray or white matter in brain tumor cases and the demyelination in the case of multiple sclerosis were enhanced and visualized in independent component images. ICA can potentially achieve separation of tissues with different relaxation characteristics and generate new contrast images of gray and white matter. With the proper choice of contrast for the original images, ICA may be useful not only for extracting subtle or hidden changes but also for preprocessing transformation before clustering and segmenting the structure of the human brain.

Dynamic monitoring of brain activation under visual stimulation using fMRI : The advantage of real-time fMRI with sliding window GLM analysis

An fMRI technique based on real-time analysis was applied to evaluate the advantages of dynamic monitoring of the t-statistics based on a general linear model. The temporal change of the t-statistics in V1 and V4 under four conditions of visual stimuli covering different visual fields with or without coloring was estimated using an incremental analysis and a sliding window analysis (SWA). The SWA not only visualized the dynamic change of the activation in response to the task conditions and switching, but also enabled us to evaluate the temporal correlation of the t-statistics among the four visual areas. It was suggested that the activity in the V4 was bilaterally organized, and the altering color stimuli gave stronger stimulation to the V1 than did the black and white stimuli. Although the activation map at each time point represents the brain activity during several task and rest blocks, a SWA will be useful to evaluate the transition of neuronal activation in response to several sequential task conditions. An incremental analysis will be useful to monitor the ongoing activation in real-time during the scan, since it gives a higher t-value according to the accumulation of volume data. These two methods will be complementary.

Evaluation of Resting State Networks in Patients with Gliomas: Connectivity Changes in the Unaffected Side and Its Relation to Cognitive Function

In this study, we investigated changes in resting state networks (RSNs) in patients with gliomas located in the left hemisphere and its relation to cognitive function. We hypothesized that long distance connection, especially between hemispheres, would be affected by the presence of the tumor. We further hypothesized that these changes would correlate with, or reflect cognitive changes observed in patients with gliomas. Resting state functional MRI datasets from 12 patients and 12 healthy controls were used in the analysis. The tumor’s effect on three well-known RSNs including the default mode network (DMN), executive control network (ECN), and salience network (SN) identified using independent component analysis were investigated using dual regression analysis. Scores of neuropsychometric testing (WAIS-III and WMS-R) were also compared. Compared to the healthy control group, the patient group showed significant decrease in functional connectivity in the right angular gyrus/inferior parietal lobe of the ventral DMN and in the dorsolateral prefrontal cortex of the left ECN, whereas a significant increase in connectivity in the right ECN was observed in the right parietal lobe. Changes in connectivity in the right ECN correlated with spatial memory, while that on the left ECN correlated with attention. Connectivity changes in the ventral DMN correlated with attention, working memory, full IQ, and verbal IQ measures. Although the tumors were localized in the left side of the brain, changes in connectivity were observed in the contralateral side. Moreover, these changes correlated with some aspects of cognitive function indicating that patients with gliomas may undergo cognitive changes even in the absence of or before the onset of major symptoms. Evaluation of resting state networks could be helpful in advancing our hodological understanding of brain function in glioma cases.

Activation of the precuneus is related to reduced reaction time in serial reaction time tasks

Multiple brain areas are activated during serial reaction time (RT) tasks (SRTTs), but the part of the brain that facilitates reductions in RT remains unclear. The present study attempted to determine the brain region contributing most to improved RTs during explicit SRTTs. Subjects comprised 18 healthy volunteers who were instructed to press one of four buttons corresponding to visual stimuli as quickly as possible and with minimal errors during functional MRI. Stimuli were presented either in random order (control condition) or in a repeated six-item sequence (learning condition). Conventional analysis contrasting learning and control conditions revealed activation in the prefrontal-parietal area, which shifted to motor area. Subjects with high RT reduction showed more prominent activation in the precuneus than subjects with low RT reduction. Intra-subject correlation analysis revealed that time course of precuneus activation was unrelated to time-course of RT reduction. However, inter-subject correlation analysis revealed that RT changes correlate only with precuneus activation, meaning that subjects showing more prominent RT reduction revealed more prominent activation of the precuneus, which is known to play critical roles in controlling finger movements with reference to buffered memory.

Neural substrates of phonological selection for Japanese character Kanji based on fMRI investigations

Japanese and Chinese both share the same ideographic/logographic character system. How these characters are processed, however, is inherently different for each language. We harnessed the unique property of homophone judgment in Japanese kanji to provide an analogous Chinese condition using event-related functional magnetic resonance imaging (fMRI) in 33 native Japanese speakers. We compared two types of kanji: (1) kanji that usually evokes only one pronunciation to Japanese speakers, which is representative of most Chinese characters (monophonic character); (2) kanji that evoked multiple pronunciation candidates, which is typical in Japanese kanji (heterophonic character). Results showed that character pairs with multiple sound possibilities increased activation in posterior regions of the left, middle and inferior frontal gyri (MFG and IFG), the bilateral anterior insulae, and the left anterior cingulate cortex as compared with those of kanji with only one sound. The activity seen in the MFG, dorsal IFG, and ventral IFG in the left posterior lateral prefrontal cortex, which was thought to correspond with language components of orthography, phonology, and semantics, respectively, was discussed in regards to their potentially important roles in information selection among competing sources of the components. A comparison with previous studies suggested that detailed analyses of activation in these language areas could explain differences between Japanese and Chinese, such as a greater involvement of the prefrontal language production regions for Japanese, whereas, for Chinese there is more phonological processing of inputs in the superior temporal gyrus.

Structural MRI correlates of amyotrophic lateral sclerosis progression

Purpose Amyotrophic lateral sclerosis (ALS) presents with varying degrees of brain degeneration that can extend beyond the corticospinal tract (CST). Furthermore, the clinical course and progression of ALS varies widely. Brain degeneration detected using structural MRI could reflect disease progression. Subjects and methods On study registration, 3-Tesla volumetric MRI and diffusion tensor imaging scans were obtained at baseline in 38 healthy controls and 67 patients with sporadic ALS. Patients had Amyotrophic Lateral Sclerosis Functional Rating Scale-Revised (ALSFRS-R) scores of ≥36 and did not have the chromosome 9, open reading frame 72 repeat expansion. Six months later, changes in ALSFRS-R (ΔALSFRS-R) scores were calculated and patients were grouped into three categories, namely, patients with slow progression with ΔALSFRS-R scores ≤3 (n=19), intermediate progression with ΔALSFRS-R scores =4, 5 and 6 (n=36) and rapid progression with ΔALSFRS-R scores ≥7 (n=12). We analysed voxel-based morphometry and tract-based spatial statistics among these subgroups and controls. Results In comparison with controls, patients with ALS showed grey matter atrophy and decreased fractional anisotropy beyond the motor cortex and CST, especially in the frontotemporal lobes and basal ganglia. Moreover, the degree of change was highly proportional to ΔALSFRS-R at the 6-month assessment. Conclusion A more rapid disease progression and poorer functional decline were associated with greater involvement of the extra-motor cortex and basal ganglia, suggesting that the spatial extent of brain involvement can be an indicator of the progression in ALS.

BAX: A Toolbox for the Dynamic Analysis of Functional MRI Datasets

We developed a toolbox called BAX (brain activation explorer) for the dynamic analysis of functional magnetic resonance imaging (fMRI) datasets using the general linear model. The toolbox provides a graphical user interface where several routines can be accessed to extract different sets of information from a given series of functional images. The dynamic analysis can be implemented using either an incremental approach or a sliding window approach. In particular, BAX can be used to construct dynamic activation maps that can be used to assess the contribution of newly added volumes in the final activation map, detect problematic segments in the dataset, or localize in time dynamic changes in brain activity. Consistency maps, which graphically represent the number of times voxels are consecutively detected as active in a given analysis, can also be constructed using either incremental or sliding window analysis. BAX runs under Matlab (MathWorks, Inc.) and requires some routines from SPM2 (Wellcome Department of Cognitive Neurology, London, UK) for its operation. It can be freely downloaded at http://www.medgrid.org/ website.

The application of grid computing to real-time functional MRI analysis

The analysis of brain imaging data such as functional MRI (fMRI) data often requires considerable computing resources, which in most cases are not readily available in many medical imaging facilities. This lack of computing power makes it difficult for researchers and medical practitioners alike to perform on-site analysis of the generated data. This paper proposes and demonstrates the use of Grid computing technology to provide medical imaging facilities with the capability of analyzing functional MRI data in real time with results available within seconds after data acquisition. Using PC clusters as analysis servers, and a software package that includes fMRI analysis tools, data transfer routines, and an easy-to-use graphical user interface, we are able to achieve fully real-time performance with a total processing time of 1.089 s per image volume (64 x 64 x 30 in size), much less than the per volume acquisition time set to 3.0 s. We also study the feasibility of using XML-based computational web services, and show how such web services can improve accessibility and interoperability while still making real-time analysis possible.

Adapting SVM image classifiers to changes in imaging conditions using incremental SVM: an application to car detection

In image classification problems, changes in imaging conditions such as lighting, camera position, etc can strongly affect the performance of trained support vector machine (SVM) classifiers For instance, SVMs trained using images obtained during daylight can perform poorly when used to classify images taken at night In this paper, we investigate the use of incremental learning to efficiently adapt SVMs to classify the same class of images taken under different imaging conditions A two-stage algorithm to adapt SVM classifiers was developed and applied to the car detection problem when imaging conditions changed such as changes in camera location and for the classification of car images obtained during day and night times A significant improvement in the classification performance was achieved with re-trained SVMs as compared to that of the original SVMs without adaptation.