Research: Neuroimaging

Overview

Our research takes advantage of various imaging modalities, including structural and functional imaging of Primary Progressive Aphasia (PPA). In collaboration with the Memory and Aging Center (MAC) Imaging Core and Rab Lab, we assist in the development of various brain imaging pipelines, which we use regularly alongside behavioral and cognitive assessments. Currently, our structural magnetic resonance imaging (sMRI) protocol comprises T1- and T2-weighted imaging for gray matter structure as well as diffusion-weighted imaging (DWI) for white matter structure. In addition, we analyze functional imaging data acquired using functional MRI (fMRI), magnetoencephalography (MEG), and positron emission tomography (PET). Together, these imaging modalities allow us to investigate the clinicopathological correlates of PPA from different, but complementary, angles.

Our areas of interest with respect to our imaging work include:

Neuroimaging (sMRI, fMRI, DTI, MEG, PET) of PPA clinical and pathological correlates
Distributed semantics via novel behavioral assessments and MEG studies
Prediction of longitudinal gray matter changes in PPA based on pre-defined healthy functional networks
Deep learning and advanced mathematical models in PPA for prediction of in-vivo pathological changes in PPA
Longitudinal tracking of structural changes in PPA and its association with behavioral and cognitive impairment
Multilingualism-related differences in clinical presentation and neural correlates of PPA
Theory- and data-driven clustering of PPA into reproducible clinical syndromes and associated brain atrophy patterns
Dissection of white matter pathways related to language and speech features in PPA
Intrinsic connectivity networks critical for specific aspects of language in PPA
Machine learning predictive modeling of cognitive and language decline in PPA
Behaviorally relevant compensatory neural mechanisms in PPA
Treatment-induced structural and functional neuroplasticity in PPA
Structural and functional organization of language in the human brain

sMRI: Neural Correlates in PPA

We utilize structural magnetic resonance imaging (sMRI) to characterize changes in grey and white matter structure in individuals with PPA, including quantitative measurements of cortical thickness and subcortical volume. Analytical methods, such as voxel-based morphometry (VBM), enable us to understand the relationship between brain anatomy, cognitive impairment, and disease progression. For example, in a prior study, we found that an impairment in the repetition of long phrases was associated with thinning of the left temporoparietal cortex across PPA variants (Lukic et al., 2019). In another previous study focusing on semantic variant PPA (svPPA) specifically, we found that gray matter integrity in the right middle temporal lobe was associated with an individual’s ability to judge whether or not a face was familiar (Borghesani et al., 2019).

**Fig. 3**
Cortical thickness maps showing areas where cortical thinning was associated with repetition performances on all phrases, and long (over short) phrases across PPAs. The presented map is thresholded at p < 0.05 family-wise error (FWE) corrected both at peak-level and cluster extent-based thresholding. Covariates: age, gender, handedness, and severity (measured by the Clinical Dementia Rating box score).
Lukic S, Mandelli ML, Welch A, Jordan K, Shwe W, Neuhaus J, Miller Z, Hubbard HI, Henry M, Miller BL, Dronkers NF, Gorno-Tempini ML. Neurocognitive basis of repetition deficits in primary progressive aphasia. Brain Lang. 2019 Jul;194:35-45. doi: 10.1016/j.bandl.2019.04.003. Epub 2019 May 2. PMID: 31055171; PMCID: PMC6669076.

**Fig. 6**
Isolating naming, semantic retrieval, and familiarity judgment. The three effects are overlaid on five axial slices and a rendered template brain (left and right hemisphere).
Borghesani V, Narvid J, Battistella G, Shwe W, Watson C, Binney RJ, Sturm V, Miller Z, Mandelli ML, Miller B, Gorno-Tempini ML. "Looks familiar, but I do not know who she is": The role of the anterior right temporal lobe in famous face recognition. Cortex. 2019 Jun;115:72-85. doi: 10.1016/j.cortex.2019.01.006. Epub 2019 Jan 23. PMID: 30772608; PMCID: PMC6759326.

fMRI: Functional Language Networks

We utilize functional magnetic resonance imaging (fMRI) to understand regional differences in cerebral blood flow across PPA variants, measured through the blood oxygenation level-dependent (BOLD) signal. In task-based fMRI, the research participant is asked to complete an experimental task (e.g. repeat a word) while inside the scanner. In task-free (or resting-state) fMRI, the research participant is allowed to let their mind wander (i.e. no explicit task) while inside the scanner. Because functional changes in the brain tend to precede observable structural changes in neurodegenerative disease, fMRI allows us to capture subtle changes in the brains of patients with PPA that we would not otherwise be able to quantify. Using task-free, or resting-state fMRI, we have studied correlated patterns of spontaneous brain activity and found four reproducible intrinsic connectivity networks that differentially relate to aspects of language such as speech perception, phonology, and semantics, across the three PPA variants (Battistella et al., 2020). In our collaboration with the UCSF Dyslexia Center, we utilize task-based fMRI in children with dyslexia to investigate differences in neural activation across cognitive tasks relative to neurotypical children.

**Fig. 2**
Spatial distribution of the four task-free networks. Rendered left-hemisphere lateral views of the four ICNs identified through a statistical conjunction of the two groups. Maps were thresholded at the same statistical threshold used to identify the networks for each group separately, binarized and colored using the same color scheme in Figure 1 (blue for the network seeded from the pITG, yellow for the network seeded from the Spt, red for the network seeded from the opIFG, and green for the one seeded from the aMTG). The render at the center of the figure shows that the four networks cover almost the entire left hemisphere with a small degree of overlap.
Battistella G, Borghesani V, Henry M, Shwe W, Lauricella M, Miller Z, Deleon J, Miller BL, Dronkers N, Brambati SM, Seeley WW, Mandelli ML, Gorno-Tempini ML. Task-Free Functional Language Networks: Reproducibility and Clinical Application. J Neurosci. 2020 Feb 5;40(6):1311-1320. doi: 10.1523/JNEUROSCI.1485-19.2019. Epub 2019 Dec 18. PMID: 31852732; PMCID: PMC7002153.

DWI: White Matter Tracts Subserving Language and Speech

Diffusion-Weighted Imaging (DWI) is a neuroimaging technique that measures the random motion of water within brain tissue. By tracking the movement of these water molecules along fiber bundles in the brain (what is referred to as Diffusion Tensor Imaging or DTI), we can reconstruct the axonal (white matter) pathways of the brain to better understand the connections between cortical and subcortical gray matter structure. To facilitate our DTI analyses, we have developed a Cortically Constrained Shape Recognition Pipeline to automatically generate whole-brain tractography of white matter pathways (Jordan, Lauricella, Licata et al., 2021). We have utilized DTI in healthy and clinical populations to investigate differences in white matter pathways and their relation to language and speech. In our research, we found that specific white matter pathways in the left hemisphere of the brain are involved in aspects of speech fluency and are selectively damaged in individuals with non-fluent variant PPA (nfvPPA) compared to individuals with semantic or logopenic variant PPA (Mandelli et al., 2014).

**Figure 1**
This flow chart shows the steps performed for the automated CCSR segmentation approach. First, the whole-brain tractography is performed in the participant's native diffusion space using High Angular Resolution Diffusion Imaging data, and a FreeSurfer parcellation is performed with the Desikan–Killiany Atlas using the subject's magnetization-prepared rapid acquisition with gradient-echo T1-weighted anatomical MRI. A boundary-based registration is used to bring the FreeSurfer parcellation into the native diffusion space, and the Diffusion Imaging in Python (DIPY) library's whole brain Streamline Linear Registration algorithm is used to align bundle templates with the native diffusion space. Next, the identified a priori inclusion and exclusion regions of interest (ROIs) are used to select the intermediate set of streamlines. Finally, shape recognition is performed using DIPY's RecoBundles to filter out the streamlines that do not match the shape of the tract based on predefined 3-dimensional bundle templates and preset parameters.
Jordan KM, Lauricella M, Licata AE, Sacco S, Asteggiano C, Wang C, Sudarsan SP, Watson C, Scheffler AW, Battistella G, Miller ZA, Gorno-Tempini ML, Caverzasi E, Mandelli ML. Cortically constrained shape recognition: Automated white matter tract segmentation validated in the pediatric brain. J Neuroimaging. 2021 Apr 20. doi: 10.1111/jon.12854. Epub ahead of print. PMID: 33878229.

**Figure 2**
Schematic representation of the speech-production and language tracts. All of the expected tracts are displayed as schematic connections. All of the ROIs are represented as colored spheres in the 3D brain (MNI space).
Mandelli ML, Caverzasi E, Binney RJ, Henry ML, Lobach I, Block N, Amirbekian B, Dronkers N, Miller BL, Henry RG, Gorno-Tempini ML. Frontal white matter tracts sustaining speech production in primary progressive aphasia. J Neurosci. 2014 Jul 16;34(29):9754-67. doi: 10.1523/JNEUROSCI.3464-13.2014. PMID: 25031413; PMCID: PMC4099550.

MEG: Brain Dynamics in PPA

Magnetoencephalography (MEG) is a functional neuroimaging technique that measures the electromagnetic fluctuations at the level of the scalp thought to reflect millisecond-by-millisecond oscillatory brain activity. In collaboration with the UCSF Biomagnetic Imaging Laboratory (BIL), we are able to scan individuals with PPA in order to directly examine dynamic brain activity with high spatial and temporal resolution, compared to other techniques such as fMRI. Utilizing MEG, we have revealed that patients with semantic variant PPA (svPPA) show differences in the spatiotemporal patterns of neural activation during tasks such as irregular word and non-word reading (Borghesani et al., 2020). This work suggests that the neural pathways and cognitive strategies typically engaged during reading aloud are altered in individuals with svPPA relative to healthy adults, potentially signaling a compensatory mechanism by over-recruiting parietal regions of the brain. In another study, we showed that following ATL damage and corresponding semantic loss, svPPA patients categorize items over-recruiting occipital areas of the brain implicated in perceptual processing (Borghesani et al., 2021).

ALBA Language Neurobiology Lab