Skip to content

Selected figures

Figures

Figures from the lab's published papers, grouped by article. Swipe or use the arrows; click any figure to enlarge.

Naturalistic movie viewing is an effective functional localizer of the fusiform face area in adolescents with and without autism

Steeby CJ et al. Imaging Neuroscience (2026). DOI

  • Fig. 2 . Levels of motion (quantified by mean framewise displacement) across tasks, separated into ASD and NT groups. Motion was significantly greater in the ASD group for TL Run 4 (*p < 0.05).
  • Fig. 5. Z-normalized mean timecourses from each ROI type applied to an independent run of the traditional localizer task in the left hemisphere, across ASD and NT groups. Right hemisphere time course looks very similar and so is not pictured here. White columns within blocks indicate presentation of flower stimulus.

Lesion network mapping of focal injury-related aggression finds two distinct network injury patterns

Miller GN et al. Brain Communications (2026). DOI

  • Graphical abstract.
  • Figure 2 Lesion network localization of each cluster. (A) Voxels functionally connected at least 60% of lesions in the ACC Cluster (n = 14 lesions). (B) Voxels functionally connected at least 60% of lesions in the vmPFC Cluster (n = 47 lesions). (C) Two-sample t-test comparing functional connectivity of the ACC Cluster (n = 14 lesions) to the large cohort of lesions causing other symptoms (n = 716 lesions) (FWE corrected P < 0.05). (D) Two-sample t-test comparing functional connectivity of the vmPFC Cluster (n = 47 lesions) to the large cohort of lesions causing other symptoms (n = 716 lesions) (FWE corrected P < 0.05).
  • Figure 3 Alignment of cluster networks with lesions causing other symptoms. Box plots showing the distribution of correlation coefficients between connectivity maps derived from individual lesions (n = 716 total lesions) associated with 22 clinical syndromes and each cluster network. Each data point represents the correlation coefficient for a single lesion-derived connectivity map, and each box represents the distribution of lesion-level correlations within a syndrome. (A) Correlation of syndrome lesions and the ACC cluster network map. Akinetic mutism showed the strongest correlation (average r = 0.56) with this cluster network. Paired t-tests revealed that lesions causing akinetic mutism (shown in at far left in red) were significantly more correlated to the ACC Cluster network compared with the vmPFC Cluster network [t(27) = 11.678, P < 0.0001]. (B) Correlation of syndrome lesions and the vmPFC cluster network map. Confabulation showed the strongest correlation (average r = 0.46) with this cluster network. Paired t-tests revealed that lesions causing confabulation (shown in orange, ninth from the left) were not significantly more correlated to the vmPFC Cluster network compared with the ACC Cluster network [t(24)=−1.056, P = 0.302].

Coordinate Network Mapping of Focal Brain Volume Differences in ADHD Reveals Common Patterns That Lack Specificity: A Systematic Review

Wall J et al. Annals of the Child Neurology Society (2025). DOI

  • FIGURE 1 | Coordinate network mapping methods and results. In this study, we first created experiment‐level maps of the coordinates (A), which we then used to calculate functional connectivity maps using a pediatric normative connectome (B). Finally, we conducted a one‐sample t test with permutation testing and multiple comparisons corrected to identify which regions were significantly and consistently connected to the original coordinates (C).
  • FIGURE 3 | Original coordinates. This map shows the distribution of the original coordinates collected from the 38 papers used in this meta‐ analysis. Although the coordinates were widespread across the brain, a maximum of only five studies contained the same overlapping coordinate.
  • FIGURE 5 | Comparison with control groups. The two‐sample t test between CNM of ADHD and psychiatric disorders (A) is largely nonsignificant, while the two‐sample t tests between two separate neurodegenerative disorder cohorts (B–C) largely reflect increased consistency in the two control cohorts. Comparison against lesion network mapping of acute stroke solely revealed the ACC (D), while comparison against a random set of atrophy studies was nonsignificant (E). For visual comparison, we show our CNM of ADHD one‐sample t test results (F, upper row) compared with the same analysis done with random coordinates (F, lower row). ADHD, attention‐deficit/hyperactivity disorder; CNM, coordinate network mapping; PALM, Permutation Analysis of Linear Models.

Lesions associated with autism symptoms map to a cerebellar brain network in tuberous sclerosis complex

Herman WX et al. Annals of the Child Neurology Society (2025). DOI

  • FIGURE 4 | Overview of lesion network mapping. All tuber voxels in totality for each patient were characterized as that patient's “lesion” (A) Permutation Analysis of Linear Models (PALM) was applied by comparing the functional connectivity between each lesion voxel against all other voxels in the brain of 1000 control participants from a connectome (B) while comparing with a measurement of interest, in this case ADOS subscores (C) The connectome used in this analysis was from the Adolescent Brain Cognitive Development (ABCD) Study, comprised of functional connectivity measured from 1000 individuals aged 9–10 years (see acknowledgments for reference). At the completion of those comparisons, a lesion network map was constructed with t‐statistics on each brain voxel. Maps of p‐values after family‐wise error (FWE) and false‐detection rate (FDR) corrections were then overlayed onto the t‐map at p > 0.95 (D) When applying PALM to our cohort's lesions along the ADOS‐SA score, we found a region of interest in the R cerebellar crus V area with the anatomical view above and the cerebellar flatmap view below. Note that the process of flattening a NIFTI ROI involves averaging of values, and therefore the range of t‐values plotted here is different than the direct result of PALM. Alt text: Composite figure with first column showing tuber locations in example axial brain slices of three patients, second column showing the functional connectivity map of each patient's tubers, third column showing ADOS testing subscores for SA and RRB, and last column showing the cerebellar region of interest plotted in three dimensions as well as on a cerebellar flatmap.
  • FIGURE 5 | Correlation between ADOS‐SA and connectivity. Graph shows correlation between ADOS‐SA residuals after subtracting the effect of covariates against each patient's tuber connectivity to the R cerebellar ROI. Spearman correlation yielded r = 0.39. Solid line indicates the linear regression with dashed lines indicating the 95% confidence interval. Alt text: Graphical representation of the residual ADOS‐SA score for each patient on the y axis and tuber functional connectivity to the cerebellar region of interest on the x axis. The line of best fit as well as 95% confidence is also plotted.

Comparison of Brain Normalization Software and Lesion Compensation Techniques in Chronic Perinatal Stroke Imaging

Miller GN et al. Imaging Neuroscience (2025). DOI

  • Fig. 2. Representative brain segmentation created using SynthSeg v2.0 with and without the robust setting for a patient with a large cystic lesion (A). Note that without the robust setting (B), there is a lack of cortical ribbon delineation through the area of cystic lesion in the right hemisphere, while using the robust setting (C) generates a simulated cortical ribbon representation that - while not an accurate representation of the patient’s current anatomy - allows for boundaryregistration algorithms to be utilized.
  • Fig. 3. Brain grafting procedure. The patient’s structural image and associated lesion mask are affine warped to template space. The mask is then flipped onto the contralateral hemisphere and used to extract a “graft” of healthy tissue. In parallel, the lesioned area is removed from the structural image. The brain graft is then flipped back to the ipsilateral side and combined with the prepared structural image, replacing the area of lesioned tissue. This “corrected” image is then normalized by the three registration software packages.
  • Fig. 10. Selected brain grafting comparison examples. Patient I with EasyReg (top): brain grafting appears to reduce accuracy; note the slightly larger posterior left lateral ventricle and incorrectly reduced lesion volume with brain grafting compared to without brain grafting. Patient K with ANTs (middle): brain grafting appears to improve accuracy; note the smaller right lateral ventricle and more anatomically accurate posterior-occipital white matter with brain grafting compared to without brain grafting. Patient K with FNIRT (bottom): brain grafting appears to improve accuracy; note the smaller right lateral ventricle and more anatomically accurate brain edge.

Network localization of altered auditory and somatosensory sensitivity based on causal brain lesions

Tripathy S et al. Brain Communications (2025). DOI

  • Graphical abstract.
  • Figure 2 Lesion network localization of sensory sensitivity. (A) Representative lesions (n = 61) causing sensory perception changes, demonstrating heterogeneity of lesion location. (B) Lesion overlap map (90%, T > 7). (C) One-sample t-test of lesion connectivity (family-wise error (FWE) corrected P < 0.05). (D) Functional connectivity specific to lesions causing sensory changes compared to a large cohort of lesions causing other symptoms via two-sample t-test (FWE corrected P < 0.05). (E) Cerebellar regions of interest present in both the one- and two-sample t-tests.

Heterogenous brain activations across individuals localize to a common network

Peng S et al. Communications Biology (2024). DOI

  • Fig. 3 | The pattern of the network derived using the activation peak as the seed is highly similar to that derived using the distributed activation map as the seed (spatial correlation r = 0.89 for the unthresholded network overlap map). a Activation peaks of the above three representative subjects from the HCP dataset. b The locations of the activation peaks across individual participants (N = 100) were highly heterogeneous, with only 8 out of the 100 activation peaks commonly located in the most convergent brain region. c However, these heterogeneous activation peaks were functionally connected to common set of brain regions. Notably, since the pattern of the activation peak overlap map (b) and activation network overlap map (Fig. 1c) are highly similar, it indicates that even though the location of the activation peaks are heterogeneous across participants, most of them are located in different parts of the same network derived from ANM analyses.

Mapping lesion-related human aggression to a common brain network

Peng S et al. Biological Psychiatry (2024). DOI

  • Figure 3. Lesion network mapping analysis identified a focal region in right prefrontal cortex whose connection was positively associated with aggression. This region, displayed on both the brain slice (A) and brain surface (B), is referred to as lesion network node. The highlighted region was corrected for multiple comparisons using permutation-based 2-sided cluster-level familywise error p (pFWE) , .05, with a cluster-forming threshold at voxel-level p , .001. L, left; R, right.
  • Figure 6. Relevance to deep brain stimulation (DBS). (A) DBS electrodes from 25 patients with drug-resistant epilepsy exhibit minor variability in electrode location within the anterior thalamus. (B) The stimulation site for each patient was determined by calculating the volume of activated tissue (VAT) using personalized stimulation parameters. (C) Functional connectivity between DBS stimulation sites and the lesion network node significantly predicted the changes in irritability before and after DBS treatment. (D) Notably, this prediction was specific to irritability because the same functional connectivity failed to predict any of the remaining 20 symptoms assessed in Beck Depression Inventory II. The red bar represents significant prediction, whereas blue bars represent nonsignificant prediction. *p , .05. Ant, anterior nucleus of thalamus; mtt, mammillothalamic tract.

Tubers Affecting the Fusiform Face Area Are Associated with Autism Diagnosis

Cohen AL et al. Annals of Neurology (2023). DOI

  • FIGURE 4: Voxelwise lesion symptom mapping (VLSM) of the association between tuber distribution and autism spectrum disorder (ASD) diagnosis found 5 significant voxels in the right temporal lobe that survive false discovery rate (FDR) multiple comparison correction, with a peak that survives familywise error correction (A). Decreasing the statistical threshold to confirm the shape of this peak and examine for other locations with less significant relationships only identified an expanded version of this cluster (B). For illustrative purposes only, tuber burden for this region is shown here (C) in the same format as in Figure 3, demonstrating a significantly higher tuber burden at this location compared to all other regions examined. Results are also shown from using a split-half model to generate this region with one half of the data and predicting ASD diagnosis in the other half (and vice versa). A standard clinical confusion matrix is shown (D), from which accuracy, false positive and negative rates, and positive and negative predictive values can be calculated. FFA = fusiform face area; L = left; R = right.
  • FIGURE 5: The voxelwise lesion symptom mapping-identified region (shown in blue) is consistent with the localization of face processing differences in nonsyndromic autism spectrum disorder (ASD), as demonstrated by an activation likelihood estimation map of “face processing” in ASD comprising 873 foci across 24 studies (A), and with face processing in general, as demonstrated by a Neurosynth meta-analysis of 31,842 activation coordinates from 896 studies (B). This region is also consistent with connectivity to lesions causing sudden onset prosopagnosia, here compared to a set of control lesions to demonstrate connectivity specific to lesion-induced impairment of face recognition (C; adapted from Cohen et al65). L = left; R = right.

Network Localization of Awareness in Visual and Motor Anosognosia

Kletenik I et al. Annals of Neurology (2023). DOI

  • FIGURE 2: Method for identifying domain specific and crossmodal networks for awareness. We analyzed 4 groups of lesions associated with (A) visual anosognosia (n = 24), (B) visual deficit with awareness (n = 69), (C) motor anosognosia (n = 95), and (D) motor deficit with awareness (n = 79). Each lesion location (red) was mapped to a common brain atlas (left image). Functional connectivity between each lesion location and all other brain voxels was computed using a large functional connectome, generating a lesion network map for each case (right image). Positive connections are shown in warm colors, and negative connections are shown in cool colors. These 267 lesion network maps were entered into a single voxelwise analysis of variance (ANOVA; modality  awareness) to generate 3 brain maps: (1) a “modality map” comparing visual deficit versus motor deficit (regardless of awareness), (2) a “modality-specific anosognosia map” based on the interaction between modality and awareness, and (3) a crossmodal “awareness map” comparing anosognosia versus awareness (regardless of the deficit).
  • FIGURE 3: Modality-specific and cross-modal anosognosia networks. (A) Modality map showing lesion connections associated with vision deficits (red) versus motor deficits (blue) regardless of awareness. As expected, this map highlights domain-specific brain regions, such as the calcarine sulcus and pre-central gyrus. (B) Modality-specific anosognosia map showing lesion connections associated with visual anosognosia (warm colors) versus motor anosognosia (cool colors). Regions more specific for visual anosognosia include the posterior cingulate/inferior precuneus. Regions more specific for motor anosognosia include the anterior insula/frontal operculum and supplementary motor area/anterior cingulate. (C) Cross-modal anosognosia map showing lesion connections associated with anosognosia (green) versus awareness of deficits (purple) across both vision and motor domains. Regions in the general anosognosia network include the bilateral hippocampi and bilateral superior precuneus. All voxels shown are false discovery rate p < 0.05 (note that panel A is shown at a higher threshold of T > 10 and T < 10 to better highlight the most significant results).

Multiple sclerosis lesions that impair memory map to a connected memory circuit

Kletenik I et al. Journal of Neurology (2023). DOI

  • Fig. 1 Methodology to test if MS memory dysfunction is associated with disruption of memory circuits. A Structural imaging, lesion segmentation and memory assessment collected on patients with MS showing two representative lesion maps from a patient with normal and a patient with low memory score. Lesion maps and memory testing are used in subsequent analyses. B Determine if MS lesion damage to the a priori stroke-derived memory circuit associates with memory dysfunction. B1. Assess lesion overlap with stroke derived memory circuit and then B2. analyze association of MS lesion overlap with stroke derived memory circuit to memory scores. C Derive

A Lesion-Derived Brain Network for Emotion Regulation

Jiang J et al. Biological Psychiatry (2023). DOI

  • Figure 2. Damage to the a priori network is associated with emotion regulation (ER) impairment. (A–C) Managing emotion scores were significantly lower for patients with lesions intersecting the a priori emotion regulation network, i.e., hit vs. no-hit, and were significantly associated with network damage percentage on the a priori network. (D–F) Managing emotion scores did not differ significantly between the 2 lesion groups on the dorsal ER network and were not significantly associated with network damage percentage on the dorsal ER network. (G–I) Managing emotion scores were significantly lower for patients with lesions intersecting the ventral ER network, i.e., hit vs. no-hit, and were significantly associated with network damage percentage on ventral ER network. n.s., not significant. L, left; R, right.
  • Figure 3. Lesion connectivity to the left ventrolateral prefrontal cortex (vlPFC) is associated with emotion regulation impairment. (A) Lesion locations from 3 representative patients. (B) Lesion connectivity maps for each patient, i.e., resting-state functional connectivity maps using individual lesions as seed regions combined with a normative connectome, shown here thresholded at t = 610. (C) Comparing these maps to managing emotion scores from each patient identified negative correlations with several regions including the left vlPFC, superior temporal sulcus (STS), and temporoparietal junction (TPJ) (voxelwise false discovery rate–corrected p , .05, k . 10). L, left; MNI, Montreal Neurological Institute; R, right.

Tuber Locations Associated with Infantile Spasms Map to a Common Brain Network

Cohen AL et al. Annals of Neurology (2021). DOI

  • FIGURE 4: Lesion network mapping identified 3 brains regions with consistent negative functional connectivity to tuber distributions associated with infantile spasms. (A) The 74 tuber distributions associated with infantile spasms were registered to a standardized Montreal Neurological Institute brain template. (B) Brain regions functionally connected to each tuber distribution were identified using a large-scale functional connectivity database of young adult participants. (C) Overlap of these functional connectivity maps identified 3 brain regions connected to >95% of tuber distributions associated with infantile spasms: the left and right globus pallidus and the cerebellar vermis. (D) Consistent connected regions were also identified in 2 independent subsets. Of note, overlap of functional connectivity maps from the 49 children without infantile spasms did not reveal any consistently connected regions. Regions where connectivity was specific to infantile spasms were then identified by voxelwise 2-sample t test between children with infantile spasms and those without. (E) The conjunction of a mask of these significant voxels and the lesion network mapping analysis above generated a map of regions both sensitive and specific for infantile spams, here shown controlling for genetic etiology as a covariate. (F) This process was repeated with an alternate largescale functional connectivity database of 9-year-old participants identifying consistent results.
  • FIGURE 5: Lesion network connectivity with the bilateral globi pallidi and cerebellar vermis predicts locations where tubers are more likely to cause infantile spasms. (A) The intersection of connectivity with the globus pallidi (Gpi, blue shading) and connectivity with the cerebellar vermis (red shading) defined a specific network of areas (purple shading) predicted to be highly likely to cause infantile spasms if lesioned. (B, C) As a demonstration, the same 4 patients shown in Figure 2, 2 with infantile spasms and 2 without infantile spasms, and with either high or low tuber burden, are again shown here with tuber burden (in red) compared to the identified network (in purple). Among patients with infantile spasms (B), it can be seen that tubers are more likely to overlap with the predicted network (blue circles). Conversely, among patients without infantile spasms (C), tubers largely do not overlap with the predicted network.

Network Localization of Unconscious Visual Perception in Blindsight

Kletenik I et al. Annals of Neurology (2021). DOI

  • FIGURE 2: Lesion network mapping of blindsight-negative versus blindsight. (A) Representative lesion locations from a blindsight and blindsight-negative patient. Lesions were consolidated into a single hemisphere for analysis (left analysis shown). (B) Connectivity between each lesion location and the rest of the brain was computed using a normative database of resting state functional connectivity from 1,000 healthy subjects. Pictured are the connectivity patterns derived from the two representative lesion locations shown in A. (C) Connectivity differences between lesion locations from blindsight-negative (n = 35) versus blindsight patients (n = 34) were identified using a 2-sample, voxelwise t test within a mask of regions previously implicated in blindsight (dark gray). Lesions in blindsight-negative patients showed greater functional connectivity to the medial pulvinar (pulvinar in blue outline) compared to lesions in blindsight patients. No voxels in the lateral geniculate nucleus, V1, V5, or superior colliculus (outlined in purple) were identified. Images were corrected for multiple comparisons using a voxel-based family wise error (FWE) rate of p ≤0.05. (D) The analysis shown in A–C was repeated, but consolidating lesion locations onto the right hemisphere rather than the left hemisphere, with identical findings. R = right.

Matched neurofeedback during fMRI differentially activates reward-related circuits in active and sham groups

Guler S et al. Journal of Neuroimaging (2021). DOI

  • FIG 1 Experimental design for the functional MRI (fMRI) scans. (A) Order and duration of all five fMRI scans, (B) block design for pre and postfeedback scans, (C) block design for three neurofeedback scans, and (D) exemplary feedback screen that participants saw during a neurofeedback scan. On panel (D), “LEFT TAP” represents the instruction text for the condition block and red rectangular bar width is continuously updated based on the lateralization index
  • FIG 2 User interface of Turbo-BrainVoyager, which was used for real-time functional MRI (fMRI) analysis, region of interest (ROI) selection, ROI timeseries extraction, and transfer to the fMRI computer. Activation maps, ROI timeseries, and subject motion can be monitored on Turbo-BrainVoyager user interface shown on panel (A), and a zoomed-in view of the two exemplary ROIs selected at the end of the prefeedback scan is shown on panel (B). Blue and red voxels represent left and right ROIs, respectively. Note that ROIs are defined as the activated voxels (T > 3) within a cube consisting of 5 × 5 × 5 voxels

Face-Processing Performance is an Independent Predictor of Social Affect as Measured by the Autism Diagnostic Observation Schedule Across Large-Scale Datasets

Zagury-Orly I et al. Journal of Autism and Developmental Disorders (2021). DOI

  • Fig. 2 Face-processing performance compared to ADOS-SA scores for male and female participants.

Looking beyond the face area: lesion network mapping of prosopagnosia

Cohen AL et al. Brain (2019). DOI

  • Figure 3 Lesion network mapping identified consistent regions connected to all lesions causing acquired prosopagnosia. Lesions were traced onto a standardized MNI brain template (A). Brain regions functionally connected to each lesion location were then obtained using a large resting state functional connectivity database (B). Overlap of thresholded functional connectivity maps (t 4 9) from each lesion identified brain regions connected to the greatest number of lesion locations (C). Of note, consistent connectivity to the right FFA was still observed from lesion locations that did not intersect the right FFA (D). In all images, the a priori right FFA region is shown as a blue outline.
  • Figure 5 Connectivity to the right FFA and left frontal cortex is specific to lesions causing acquired prosopagnosia compared to control lesions and lesions causing other syndromes. Using our entire cohort of lesions causing prosopagnosia (n = 44), all lesion locations demonstrated positive and negative correlation to a specific set of locations (A). This pattern of connectivity was specific to lesions causing prosopagnosia compared to a large cohort of control lesions causing non-specific symptoms (B) or to lesions causing specific symptoms other than prosopagnosia (C). The conjunction of our sensitivity and specificity analyses (D) identified five locations including the right FFA, the left anterior prefrontal cortex (APFC), the left middle frontal gyrus (MFG), the dorsal anterior cingulate cortex (ACC), and the left superior frontal gyrus (SFG). In all images, the a priori right FFA region is shown as a blue outline.
  • Figure 8 Lesion connectivity with the right FFA and left frontal cortex predicted subclinical facial agnosia. The intersection of positive connectivity with our identified right FFA (red shading) and negative connectivity with our left frontal regions (blue shading) defined a specific network of areas (purple shading) (A) highly likely to cause prosopagnosia if lesioned. Posterior cerebral artery strokes from an independent dataset that were associated with subclinical facial agnosia (B), versus lesions associated with intact facial perception (C), were significantly more likely to intersect this network (D). (P 5 0.01). Red lines in box-plots indicate medians while stars indicate means.