Robert L. Savoy

Functional magnetic resonance imaging (fMRI) “brain reading”: detecting and classifying distributed patterns of fMRI activity in human visual cortex

Traditional (univariate) analysis of functional MRI (fMRI) data relies exclusively on the information contained in the time course of individual voxels. Multivariate analyses can take advantage of the information contained in activity patterns across space, from multiple voxels. Such analyses have the potential to greatly expand the amount of information extracted from fMRI data sets. In the present study, multivariate statistical pattern recognition methods, including linear discriminant analysis and support vector machines, were used to classify patterns of fMRI activation evoked by the visual presentation of various categories of objects. Classifiers were trained using data from voxels in predefined regions of interest during a subset of trials for each subject individually. Classification of subsequently collected fMRI data was attempted according to the similarity of activation patterns to prior training examples. Classification was done using only small amounts of data (20 s worth) at a time, so such a technique could, in principle, be used to extract information about a subjects percept on a near real-time basis. Classifiers trained on data acquired during one session were equally accurate in classifying data collected within the same session and across sessions separated by more than a week, in the same subject. Although the highest classification accuracies were obtained using patterns of activity including lower visual areas as input, classification accuracies well above chance were achieved using regions of interest restricted to higher-order object-selective visual areas. In contrast to typical fMRI data analysis, in which hours of data across many subjects are averaged to reveal slight differences in activation, the use of pattern recognition methods allows a subtle 10-way discrimination to be performed on an essentially trial-by-trial basis within individuals, demonstrating that fMRI data contain far more information than is typically appreciated.

Voluntary Attention Modulates fMRI Activity in Human MT–MST

How does voluntary attention to one attribute of a visual stimulus affect the neural processing of that stimulus? We used functional magnetic resonance imaging to examine the attentional modulation of neural activity in the human homolog of the MT-MST complex, which is known to be involved in the processing of visual motion. Using a visual stimulus containing both moving and stationary dots, we found significantly more MT-MST activation when subjects attended to the moving dots than when they attended to the stationary dots, even though the visual stimulus was identical during the two conditions.

Auditory and visual word processing studied with fMRI

Brain activations associated with semantic processing of visual and auditory words were investigated using functional magnetic resonance imaging (fMRI). For each form of word presentation, subjects performed two tasks: one semantic, and one nonsemantic. The semantic task was identical for both auditory and visual presentation: single words were presented and subjects determined whether the word was concrete or abstract. In the nonsemantic task for auditory words, subjects determined whether the word had one syllable or multiple syllables. In the nonsemantic task for visual words, subjects determined whether the word was presented in lower case or upper case.

Generalizable patterns in neuroimaging: how many principal components?

Generalization can be defined quantitatively and can be used to assess the performance of principal component analysis (PCA). The generalizability of PCA depends on the number of principal components retained in the analysis. We provide analytic and test set estimates of generalization. We show how the generalization error can be used to select the number of principal components in two analyses of functional magnetic resonance imaging activation sets.

Visibility of low-spatial-frequency sine-wave targets: Dependence on number of cycles

The number of cycles in a low-frequency sinusoidal display is a crucial variable in determining the visibility of the display. In particular, the threshold contrast is essentially independent of spatial frequency for these displays. We have extended the above experiments, using more cycles and a variety of targets and observer tasks. The results confirm previous findings; they also show that the type of target or task has little influence. For low-frequency sinusoids that contain up to about 3 cycles, the threshold contrast is determined by the number of cycles. For high-number-of-cycles targets with spatial frequencies above 6-10 cycles per degree, visibility is predominantly dependent on the spatial frequency. The results suggest that the low-frequency decrease in reported MTFs is due to the decrease of the number of cycles used in determining them.

Chemosensory Cues to Conspecific Emotional Stress Activate Amygdala in Humans

Alarm substances are airborne chemical signals, released by an individual into the environment, which communicate emotional stress between conspecifics. Here we tested whether humans, like other mammals, are able to detect emotional stress in others by chemosensory cues. Sweat samples collected from individuals undergoing an acute emotional stressor, with exercise as a control, were pooled and presented to a separate group of participants (blind to condition) during four experiments. In an fMRI experiment and its replication, we showed that scanned participants showed amygdala activation in response to samples obtained from donors undergoing an emotional, but not physical, stressor. An odor-discrimination experiment suggested the effect was primarily due to emotional, and not odor, differences between the two stimuli. A fourth experiment investigated behavioral effects, demonstrating that stress samples sharpened emotion-perception of ambiguous facial stimuli. Together, our findings suggest human chemosensory signaling of emotional stress, with neurobiological and behavioral effects.

Developmental neural networks in children performing a Categorical N-Back Task

The prefrontal and temporal networks subserving object working memory tasks in adults have been reported as immature in young children; yet children are adequately capable of performing such tasks. We investigated the basis of this apparent contradiction using a complex object working memory task, a Categorical n-back (CN-BT). We examined whether the neural networks engaged by the CN-BT in children consist of the same brain regions as those in adults, but with a different magnitude of activation, or whether the networks are qualitatively different. Event-related fMRI was used to study differences in brain activation between healthy children ages 6 and 10 years, and young adults (20-28 years). Performance accuracy and RTs in 10-year-olds and adults were comparable, but the performance in 6-year-olds was lower. In adults, the CN-BT was highly effective in engaging the bilateral (L>R) ventral prefrontal cortex, the bilateral fusiform gyrus, posterior cingulate and precuneus, thus suggesting an involvement of the ventral visual stream, with related feature extraction and semantic labeling strategies. In children, the brain networks were distinctly different. They involved the premotor and parietal cortex, anterior insula, caudate/putamen, and the cerebellum, thus suggesting a predominant involvement of the visual dorsal and sensory-motor pathways, with related visual-spatial and action cognitive strategies. The findings indicate engagement of developmental networks in children reflecting task-effective brain activation. The age-related pattern of fMRI activation suggests a working hypothesis of a developmental shift from reliance on the dorsal visual stream and premotor/striatal/cerebellar networks in young children to reliance on the ventral prefrontal and inferior temporal networks in adults.

A functional magnetic resonance imaging study of auditory vigilance with low and high information processing demands

This study identified the brain activations associated with auditory vigilance tasks, using functional magnetic resonance imaging. We created auditory continuous performance tests (CPTs) in which a demanding task (working memory task) was made more difficult than a simple vigilance task by increasing working memory and interference filtering demands. Two cohorts of normal male controls performed significantly worse on the working memory CPT than on the vigilance task. Compared to the vigilance task, performance of the working memory task produced significant signal change in lateral and medial prefrontal cortex, precentral cortex, temporal lobe, including insula and hippocampus, parietal-occipital cortex, cingulate, thalamus, and superior colliculus. Performance and degree of activation was associated with an estimate of IQ. Further research should clarify the contributions of working memory and interference filtering to the activated network.

Processing of sub-syllabic speech units in the posterior temporal lobe: an fMRI study.

The objective of this study was to investigate phonological processing in the brain by using sub-syllabic speech units with rapidly changing frequency spectra. We used isolated stop consonants extracted from natural speech consonant-vowel (CV) syllables, which were digitized and presented through headphones in a functional magnetic resonance imaging (fMRI) paradigm. The stop consonants were contrasted with CV syllables. In order to control for general auditory activation, we used duration- and intensity-matched noise as a third stimulus category. The subjects were seventeen right-handed, healthy male volunteers. BOLD activation responses were acquired on a 1.5-T MR scanner. The auditory stimuli were presented through MR compatible headphones, using an fMRI paradigm with clustered volume acquisition and 12 s repetition time. The consonant vs. noise comparison resulted in unilateral left lateralized activation in the posterior part of the middle temporal gyrus and superior temporal sulcus (MTG/STS). The CV syllable vs. noise comparison resulted in bilateral activation in the same regions, with a leftward asymmetry. The reversed comparisons, i.e., noise vs. speech stimuli, resulted in right hemisphere activation in the supramarginal and superior temporal gyrus, as well as right prefrontal activation. Since the consonant stimuli are unlikely to have activated a semantic-lexical processing system, it seems reasonable to assume that the MTG/STS activation represents phonetic/phonological processing. This may involve the processing of both spectral and temporal features considered important for phonetic encoding.

Visibility of low-frequency sine-wave targets: Dependence on number of cycles and surround parameters

Several papers have described experiments which have shown that human visual sensitivity to sinusoidal gratings is dependent on the number of cycles of sinusoid for low-spatial-frequency targets (McCann, Savoy and Hall 1973; Hoekstra, van der Goat, van den Brink and Bilsen, 1974; McCann, Savoy, Hall and Scarpetti 1974; Savoy and McCann, 1975). A recent article by Estevez and Cavonius (1976) presented the argument that dependence on the number of cycles occurred only in the presence of a dark surround They reported a dependence on number of cycles with sinusoids that have a black surround, but they also reported a lack of dependence on number of cycles with targets that have averageluminance areas adjacent to the sinusoidal area. They pointed out that the majority of previous experiments had been performed with a dark surround.’ The exception was one of the experiments by Savoy and McCann (1975) in which the luminance of the area surrounding the sinusoidal grating was equal to the average luminance of the grating. In this case, Estevez and Cavonius (1976) argued that a hairline edge of a mirror, which produced a faint visible line around the sinusoidal portion of the display, was having an effect very similar to that of a black surround. This Letter presents experimental results which replicate the Savoy and McCann (1975) data using targets without any visible line. This shows that the conclusions reached in the 1975 paper reporting a dependence on number of cycles are correct. Furthermore, the experiments show that the experimental data of Estevez and Cavonius are due to substantial differences in non-sinusoidal parameters of their displays compared to those used by Savoy and McCann. We began by repeating the Savoy and McCann experiments without the hairline edge created by the mirror. William Wray. of our laboratory, in collaboration with John Hall designed a display system that provides stimuli without visible lines around the sinusoidal portion. The unusual property of this electronic display system is that it allows us to vary the luminance and the size of the average-luminance surround on all four sides of the sinusoidal portion of the display. Estevez and Cavonius displayed areas of average luminance on the left and the right, but not on the top and the bottom of the sinusoidal portion