A Quantitative Analysis of the Nonlinearity of the BOLD Response in Different Cortices D. A. Soltysik 1 , K. K. Peck 2 , K. Gopinath 1 , K. D. White 3 , R. W. Briggs 4 1 Nuclear and Radiological Engineering, University of Florida, Gainesville, FL, United States, 2 Radiology, University of Florida, Gainesville, FL, United States, 3 Psychology, University of Florida, Gainesville, FL, 4 Radiology, VA RR&D Brain Rehabilitation Center, University of Florida, Gainesville, FL, United States Abstract We investigated the nonlinearity of the BOLD response due to changes in stimulus duration in three different primary cortices (auditory, motor, and visual). A quantitative analysis was performed on the hemodynamic response function (HRF) that examined the change in maximum percent change in signal (%∆S), time-to-peak (TTP), full width at half maximum (FWHM), and area under the curve. Our results show that changes in stimulus duration produce linear changes in TTP, FWHM, and area but produce nonlinear changes in %∆S. Introduction The nonlinearity of the BOLD response complicates analysis of functional brain activation in paradigms where the brain may be stimulated for differing amounts of time. Researchers have previously shown that nonlinearity exists for auditory 1 , motor 1 , and visual 2,3,4 stimuli. Differences in opinion exist, however, over the range of stimulus duration periods where the BOLD response is linear. The linear range has been reported for duration periods greater than 4 seconds 3,4 and for duration periods less than 4 seconds 1 . Methods Four right-handed males (19±1.9 years old) were each scanned on two separate occasions. The first experiment activated the primary auditory and motor cortices. A binaural stimulation using 440 Hz tones was delivered at a rate of 4 Hz to the subjects who were asked to bring the fingers of their right hand together with their thumb in unison with the tones. The baseline was an absence of tones and no finger motion. The second experiment activated the primary visual cortex using a b/w checkerboard flashing at 7.5 Hz and subtending 10° of the subject’s field of view. A gray screen of equal luminance was used as the baseline. The visual field was projected onto a screen viewable to the subject via a mirror. Functional MRI data were acquired via a 3T GE LX scanner and a dome-shaped quadrature RF head coil. Twenty 3-mm thick coronal slices were obtained with a 1-shot spiral scan (TR/TE/FA = 1000ms/18ms/60°) with a 64×64 matrix (FOV = 200mm). The first two functional scans consisted of a 12-second on/12-second off periodic stimulation with ten cycles each. Eight functional scans followed, each with two trials of each of five different duration periods (1, 2, 4, 8, and 16 seconds) followed by a 20-second baseline and arranged in a pseudorandom order. A 3D- spoiled GRASS sequence was used to provide anatomical reference. Analysis A cross correlation analysis 5 was performed on the first two periodic runs using a reference vector with the positive lobes of a sine curve. Phase-shifting allowed selection of the best reference vector. The time series of the highest correlated voxel (r > 0.6) was then averaged across stimulus cycles and smoothed to produce an ideal HRF curve. A time-averaged response vector 1 was made from this ideal HRF curve and a final cross correlation analysis was performed. All voxels located in the appropriate primary cortical region with a value of r > 0.5 were chosen as activated voxels (min=32, max=159). The HRF of each of the five stimulus duration periods was then averaged across all 16 trials and all activated voxels. Resulting HRF curves were then averaged over the four subjects. Results Changes in three of the parameters of the HRF (TTP, FWHM, and area) appear to display a fairly linear relationship (Fig. 1). None of the extrapolated ordinate- intercepts appear to be zero, however. Consequently, increments in these parameters may be linear but the ratio of the ordinate to the abscissa is not constant over changes in stimulus duration. The %∆S curves exhibit a rise with exponential decay, with the visual cortex reaching a plateau after 10-15 seconds and the auditory and motor cortices reaching a plateau after four seconds. For all three cortices, one can still describe the change between one and four seconds as roughly linear. Discussion Understanding the nonlinearity of the BOLD response is helpful in developing fMRI analysis techniques. Our results indicate that TTP, FWHM, and area of the HRF are linear over but not proportional to the stimulus duration used while the %∆S is an exponential function of stimulus duration. References 1. Glover GH, NeuroImage 9: 416-429 (1999). 2. Liu HL & Gao JH, MRI 19: 931-938 (2000). 3. Miller KL et al., HBM 13: 1-12 (2001). 4. Vazquez AL & Noll DC, NeuroImage 7: 108-118 (1998) 5. Bandettini PA et al., MRM 30: 161-173 (1993). Figure 1. Parameters of the BOLD HRF for five different stimuli duration periods. 1790 Proc. Intl. Soc. Mag. Reson. Med. 11 (2003) View publication stats View publication stats