Enhanced parietal cortex activation during location detection in children with autism

Seventeen high-functioning children and adolescents with ASD (mean age: 13.45 years, SD: 1.73; male/female: 16/1) and 19 typically developing (TD) control participants (mean age: 12.41, SD: 1.56; male/female: 16/2) took part in this fMRI study of object recognition and location detection. Both groups were matched on age, and IQ, measured by the Wechsler Abbreviated Scale of Intelligence (WASI) (see Table 1 for demographic information).

Scans were acquired for 22 TD and 21 ASD participants. This study was approved by the Institutional Review Board of the University of Alabama at Birmingham, and all participants provided informed consent for their participation in the study. Each participant’s data were examined for continuous head motion, intermittent spikes, and drifts in x, y, and z translational directions after the realignment during data preprocessing, and stringent head motion criteria were applied to ensure data quality. Participants were excluded if more than 20% of their functional images displayed excessive motion (more than 2 mm) in any direction. These criteria resulted in a total of 19 TD and 17 ASD participants. Any remaining motion within each subject’s scans was entered into the general linear model in SPM8 (Wellcome Department of Cognitive Neurology, London, UK) as a regressor of no interest. A two-tailed, independent sample Mann-Whitney U test was performed using IBM SPSS 22 (using the TD group as the reference group) to determine the distribution of head motion across six dimensions (translation: x, y, and z; rotation: pitch, roll, and yaw). This analysis did not reveal any statistically significant difference in head motion between the two groups in any of the dimensions: x: U(33) = 144, z = -0.297, p = 0.766; y: U(33) = 140, z = -0.165, p = 0.668; z: U(33) = 148, z = -0.165, p = 0.869; pitch: U(33) = 146, z = -0.817, p = 0.817; roll: U(33) = 139, z = -0.462, p = 0.644; and yaw: U(33) = 140, z = -0.429, p = 0.668.

We used a simple object recognition and location detection paradigm to examine functional connectivity in participants with autism. A series of grey-scale photographs of small common household objects overlaid against a black background were presented in the MRI scanner in a blocked design format. The objects presented included, but were not limited to, the following categories: miniature animals, children’s toys, kitchen objects, and clothing items. These stimuli were adapted from Pennick and Kana [8]. Each stimulus item presented during the experiment was unique, and the presentations of the blocks were pseudorandomized with two tasks (four blocks of object recognition and four blocks of location detection) and five iterations of a fixation baseline each lasting 24 s. During the object recognition blocks, participants were asked to identify a given object by choosing the name of the object from a list of four answer choices. During the location detection block, participants identified the location of a given object relative to a fixation cross at the center of the screen; the answer choices (left, right, above, below) indicated the object’s location relative to the cross. Responses were recorded using fiber optic buttons. The recorded responses provided the reaction time and performance accuracy data for the object recognition and location detection tasks. For both tasks, each stimulus was presented for 6 s, during which the participant chose an answer. Each block consisted of six pictures with an inter-stimulus interval of 1 s.

All participants practiced the experiment on a laptop computer before the scanning session. While in the scanner, E-Prime 1.2 (Psychology Software Tools, Inc., Pittsburgh, USA) was used to present the stimuli. An integrated functional imaging system (IFIS) interface projected the data onto a screen behind the participant’s head which was viewed using a mirror. Images were acquired using a 3 T Siemens Allegra head-only scanner housed at the Civitan International Research Center, University of Alabama at Birmingham. Structural images were acquired using high resolution T1-weighted scans using a 160 slice 3D MPRAGE volume scan with a repetition time (TR) = 200 ms, echo time (TE) = 3.34 ms, flip angle = 7°, field of view (FOV) = 25.6 cm, 256 × 256 matrix size, and 1-mm slice thickness. To record functional imaging data, a single-shot gradient-recalled echo-planar pulse sequence was used, which offers the advantage of rapid image acquisition (TR = 1,000 ms, TE = 30 ms, flip angle = 60°, FOV = 24 cm, matrix 64 × 64). This sequence covers most of the cortex (seventeen 5-mm-thick slices with a 1-mm gap were acquired in an oblique-axial orientation) in a single cycle of scanning (TR = 1) with an in-plane resolution of 3.75 × 3.75 × 5 mm³.

The data were preprocessed and statistically analyzed using SPM8 (Wellcome Department of Cognitive Neurology, London, UK). Images were corrected for slice acquisition timing, motion-corrected, and normalized to the MNI template, resampled to 2-mm³ voxels, and smoothed with a 6-mm full width half maximum (FWHM) kernel. Statistical analyses were performed on individual data using the general linear model, while group analysis used random-effects models. Areas of statistically significant activation were determined using t-statistics on a voxel-by-voxel basis. For statistical significance, the data were examined using family-wise error-corrected multiple comparisons (p < 0.05) for the contrasts between the tasks with fixation. For direct contrasts between conditions, we applied Monte Carlo simulations to the data using 3dClustSim in AFNI [28] within an average oblique-sliced mask generated from each subject’s functional images to determine the minimum number of voxels in each cluster threshold of p < 0.05. Based on this simulation, an uncorrected threshold of p = 0.005 and a cluster extent threshold of eighty-two 2-mm³ voxels were used for within and between-group comparisons.

Functional connectivity (FC; the synchronization of brain activation between regions) was computed (separately for each participant) as a correlation between the average time course of the signal intensity of all activated voxels from a given region of interest (ROI) with the average time course of the signal intensity of all activated voxels from every other ROI. The ROIs for FC analysis were defined by examining the clusters of significant activation for all participants (ASD + TD) for the contrast object + location vs. fixation so that it best represented all regions activated for object and location tasks. A total of 15 ROIs were defined which included the following: the supplementary motor area (SMA), bilateral inferior parietal lobule (LIPL, RIPL), thalamus (LTHAL, RTHAL), inferior temporal gyrus (LITG, RITG), superior parietal lobule (LSPL, RSPL), occipital cortex (LOC, ROC), left middle frontal gyrus (LMFG), left precentral gyrus (LPRCN), medial prefrontal cortex (MPFC), and right hippocampus (RHIP) (See Additional file 1: Figure S1 for the locations of these ROIs). Statistical t-maps from contrasts of the normalized and smoothed images were high-pass filtered and had the linear trend removed. Activation values from the t-maps that did not exceed a t-threshold of 3.0 were not included in statistical comparisons of the correlations. The time courses across the ROIs were correlated, and Fisher’s r to z transformation was applied to the correlation coefficients prior to averaging and performing statistical comparisons. In addition to the functional connectivity analysis across 15 ROIs, a connectivity network analysis was conducted by grouping 12 of the 15 ROIs into different networks based on the lobes and hemispheres to which they belong. These networks and the ROIs of which that they consist are: left parietal (LIPL, LSPL), right parietal (RIPL, RSPL), inferior temporal (LITG, RITG), occipital (LOC, ROC), and frontal (LMFG, LPRCN, SMA, MPFC). In addition to significantly reducing the number of statistical comparisons in connectivity analysis, the network analysis also allowed to test the functional connectivity among different lobes in mediating object recognition and location detection.

Performance accuracy and reaction time data collected in the scanner were analyzed to determine group differences and condition effects. Two-sample t-tests revealed no significant group differences for mean reaction time (location detection: TD mean = 2590.16 ms, ASD mean = 2512.24 ms; t(33) = 0.397; p = 0.397; object recognition: TD mean = 3135.25 ms, ASD mean = 3016.91 ms; t(33) = 0.704; p = 0.092) or for performance accuracy (location detection: TD mean = 70.8%, ASD mean = 63.7%; t(33) = 0.562; p = 0.931; object recognition: TD mean = 79.2%, ASD mean = 64.7%; t(33) = 0.092; p = 0.704) in either experimental condition. For within-group effects, paired sample t-tests revealed a significant effect of condition in TD participants for reaction time (location detection: mean = 2590.16 ms; object recognition: mean = 3135.25 ms; t(17) = 6.51; p < 0.001), with more time needed for object recognition relative to location detection task. There was no significant effect of condition on accuracy of trials with responses in TD participants (location detection: mean = 70.8%; object recognition: 79.2%; t(17) = -1.39; p = 0.183). In the ASD group, there was also a statistically significant effect of condition on reaction time (location detection: mean = 2512.24 ms; object recognition: mean = 3016.91 ms; t(16) = -4.47; p < 0.001), but not for accuracy to trials with responses (location detection: mean = 63.73%; object recognition: mean = 64.7%; t(16) = -0.234; p = 0.838). However, it should be noted that individuals in the ASD group had a much larger incidence of non-responses or responses outside of the 7-s response time compared to controls (t(68) = -2.377; p = 0.017). This difference seems to have been primarily driven by the object condition, as the number of missing data points in this condition for participants with ASD approached significance (t(33) = -2.50; p = 0.061), while that for location detection did not (t(33) = -1.50; p = 0. 143).

Both object recognition and location detection tasks (when contrasted with fixation baseline) primarily activated occipital and parietal-temporal areas, including inferior parietal lobule (IPL), superior parietal lobule (SPL), and inferior temporal gyrus (ITG) bilaterally in ASD and TD participants. In addition, there was increased activity in bilateral frontal areas (precentral and middle frontal) in both groups, particularly during the object recognition task. Hippocampal and IFG activation was unique to contrasts involving object recognition in each group (see Additional file 2: Tables S1 and S2 for a detailed list of regions activated with cluster size for location vs. fixation and object vs. fixation contrasts for each group).

Direct contrasts of location detection and object recognition within groups revealed robust bilateral IPL activation (location > object) in ASD participants along with precuneus and right middle/superior temporal cortex. The TD participants showed increased activation in left precuneus, right superior temporal, and IPL areas during this contrast (see Figure 1 and Table 2). Object recognition, when compared to location detection (object > location), elicited greater activity in bilateral inferior occipital, fusiform gyrus and ITG along with inferior, middle, and medial frontal areas in TD participants. The ASD participants, on the other hand, showed increased activity only in right calcarine and lingual gyrus in this contrast (see Figure 2 and Table 2) (p < 0.005 with a cluster size correction of 82 mm³).

Group differences in brain activation were determined by direct contrasts between TD and ASD children, which revealed significantly increased activity in left IPL (particularly in the angular gyrus, k = 146) and cuneus (k = 141) in ASD participants relative to TD participants during location detection (location > object) (see Figure 3). A similar increase in IPL activity was also seen in the contrast location > fixation (k = 247) in ASD participants, with some overlap with the middle temporal gyrus. There were no statistically significant group differences for object recognition; nor was there any significantly increased activity for TD participants relative to ASD participants.

Based on previous findings of connectivity differences in ASD [10, 21, 22], we hypothesized intact functional connectivity in relatively posterior brain areas in our participants with ASD. Functional connectivity among individual ROIs revealed no statistically significant differences (after correcting for multiple comparisons) between ASD and TD participants in any of the ROI pairs. However, at an uncorrected statistical threshold, there was decreased connectivity in ASD participants in several pairs of ROIs during location detection, and between the ITG and SPL during object recognition, relative to TD participants (see Table 3 for the list of ROI pairings). In order to assess the validity of these results by adjusting for a large number of comparisons, the region-level functional connectivity analysis with 15 seed ROIs was followed up by a network connectivity analysis by grouping 12 of the ROIs into different networks (left parietal, right parietal, inferior temporal, occipital, and frontal) (see “Methods” section).

There was decreased connectivity in ASD participants between the left parietal and inferior temporal networks (p = 0.04) during object recognition and between the inferior temporal and occipital network connections and the right parietal-to-inferior temporal connections in the location condition (p = 0.01 and p = 0.04, respectively). However, the network connectivity results for both conditions did not survive stepdown-Bonferroni corrections for multiple comparisons (initial p = 0.005) (see Table 3 for a list of significant ROI pairs and network pairs).

Increased activation of the left inferior parietal cortex, a dorsal stream area, in individuals with ASD during location detection is consistent with similar findings from previous studies, especially with one of our own studies examining global and local processing in autism [21]. The peak of IPL activation reported in the current study is in the angular gyrus (AG), a region associated with multiple functions. It has a critical role in attention [37, 38], especially in the reorienting or shifting of attention [39]. The IPL, including the supramarginal gyrus and the AG, is part of a “bottom-up” attentional subsystem that mediates the automatic allocation of attention to task-relevant information [40], particularly in attending to retrieved memories [41]. It is possible that the participants in our study, especially ASD children, may associate the objects presented to their past experiences with it. The role of AG in spatial cognition is also important in the context of the current study. While IPL has been found to play a significant role in integrating information from dorsal and ventral streams [42], it has also been implicated in other functions, such as attention, praxis, self-other discrimination, visuospatial perception, and visualization of interaction with objects ([43‐48]). The increased IPL activation is also interesting considering the somatosensory and attentional dysfunctions typically reported in individuals with ASD, and increased activity within this region could reflect a pattern of activation unique to individuals with ASD when processing or orienting to basic object stimuli.

Detecting the locations of objects or more “local” processing of objects has been argued to be intact to superior in ASD individuals, but such processing can be task-specific [49]. As such, the increased activity in the parietal cortex could reflect a difference in perceptual strategy used by ASD children for locating the position of objects. This could include a more local or “lower level” perceptual processing in accordance with theories of weak central coherence [50, 51] in autism, or it could reflect increased effort or difficulty in performing the tasks, which is suggested by prior studies finding difficulty in attending to (or failing to disengage from) objects presented in a visual field in individuals with ASD [52, 53]. Increased activation was also seen in cuneus in participants with ASD, relative to TD, while locating the position of objects. Cuneus has been found to act as a link between signals from striate and extrastriate cortices [54], suggesting the increased emphasis of ASD participants on visual coding in this task.

While none of the decreases in connectivity in participants with ASD survived corrections for multiple comparisons, there were a noticeable number of decreased connections in autism between frontal, parietal, and inferior temporal regions during both tasks and decreased temporal-occipital connections during the location task. Studies of neuronal connectivity in monkeys have suggested that the IPL shares dense connections between several frontal (IFG, frontal eye fields, dorsolateral PFC), temporal (superior temporal, insula) somatosensory, and extrastriate visual areas. Thus, IPL may be a key region for visual and sensorimotor integration and response [55‐57]. Thus, despite such widespread anatomical connections reported in literature (and increased activation in IPL seen in the current study), participants with ASD in our study showed decreased connectivity, perhaps suggesting the isolated and autonomous functioning of individual regions like IPL. A prominent model of visual processing suggests that the frontal cortex serves as a “top-down” mediator for processing stimuli between the ventral and dorsal streams in order to recognize objects [58]. The decreased connectivity between known targets of the IPL, coupled with the increased activation in extrastriate visual areas such as the cuneus in individuals with ASD during the location task, may necessitate alternate information processing strategies stemming from general frontal-posterior underconnectivity [26]. This disconnect of targets between visual processing streams may result in adaptive increases in activity to compensate for decreased communication between “bottom-up” (visual cortex, extrastriate and parietal regions) and “top-down” (frontal and temporal) regions of the brain in ASD. However, whether this disconnect could result in increased reliance on alternate perceptual strategies is uncertain.