0237

Cortical depth-dependent fMRI: heterogeneity across tasks, across participants, across days and along the cortical ribbon
Laurentius Huber 1, Daniel A Handwerker1, Andrew Hall1, David C Jangraw2, Javier Gonzalez-Castillo1, Maria Guidi3, Dimo Ivanov4, Benedikt A Poser4, and Peter A Bandettini1

1SFIM, NIMH, Bethesda, MD, United States, 2NIMH, United States, 3Max Planck Institute for human cognitive and Brain science, Leipzig, Germany, 4MBIC, Maastricht University, Netherlands

Synopsis

Measurements of depth-dependent cortical activity provide insights on directional activity between brain areas. While previous studies demonstrated the feasibility of human depth-dependent fMRI, the stability and reliability of depth-dependent results are less studied. In this work, we investigate sources of inconsistencies in depth-dependent activity profiles. We find that depth-dependent activity profiles are highly reproducible across different scanning sessions. They are, however, quite variable within cortical areas across different cross sections along the cortical ribbon. Only when depth-dependent profiles are considered with respect to their location along the cortical ribbon, task-driven modulations of input-output activity become consistent across participants.

Purpose

Measurements of layer-dependent cortical activity provide insight on how feedforward/feedback functional connectivity affects a given cortical area. A few promising studies looking at differently modulated feedback activity [1,2,3], show that depth-dependent fMRI can capture depth-dependent activity modulations. However, individual studies come to contradictory conclusions (feedback in supragranular vs. infragranular layers) [1,2]. In this study, we sought to investigate the reproducibility, consistency, and heterogeneity of cortical depth-dependent fMRI results. We focus on differences of depth-dependent activity A) across different sensory-motor tasks, B) across scan sessions within participants on different days; C) across different cross sections within the cortical ribbon that we call ‘cortical distances’; D) across different participants with different anatomical features and folding patterns.

Methods

Experiments in 14 volunteers were performed on a 7T Siemens scanner with a 32-channel NOVA Medical head coil. Data acquisition procedures were used as previously presented in [4]. In short: We used a multi-contrast VASO-BOLD sequence with slices positioned to be approximately perpendicular to the M1 cortex. (Nominal resolution = 0.75×0.75×1.5 mm3, TI1/TI2/TR = 1.1/2.6/3.0 s, 12 slices with FLASH GRAPPA-2 [5], matrix 44x132, 3D-EPI readout [6]. Five 12-min functional scans were collected per participant: (1) right-hand tapping with touch, (2) left-hand tapping with touch, (3) right-hand tapping without touch, (4) no tapping while being touched with an abrasive cushion, (5) resting state. Segmentation and cortical layering algorithms (based on the equi-volume approach [7]) were applied directly on the EPI images from the functional scans.

For cross-participant comparison and cross-participant averaging, we developed a sub-sampling method across cortical depths and cortical distances: depths are normalized based on GM/CSF and GM/WM borders. Cortical distances are normalized based on the bending radius (red arrows in Fig. 2A) and hand knob templates described in [8]. Task-specific responses are then averaged in this 2D-grid.

Results

The method provides sufficient sensitivity and specificity to identify depth-dependent activity features (Fig. 1) as shown earlier [4]. Depth-dependent activity features, like a double-layer response, can be reproduced across multiple scan sessions up to 6 months apart. Activity maps acquired across different days look very similar; two distinct active depths can be identified across all scanning sessions (black arrows in Fig. 1). Fig. 2A/B depicts how a standardized 2D grid that is seeded in the hand knob can be used to average the depth-dependent fMRI responses for multiple tasks across cortical distances (Fig. 2C) and cortical depths (Fig. 2D). These averaged 2D grids are transformed back into the EPI space of one representative participant in Fig. 3 to investigate the heterogeneous fMRI responses across cortical distances. Cortical distances within M1 have different responses (for instance, white and black arrows in Fig. 3) [9]. For the response heterogeneity across tasks, average depth-dependent profiles are replotted on top of each other in Fig. 4A. Using a depth-dependent model assuming superposition of input and output across M1 layers [10] (Fig. 4B), cross-participant results can be investigated for task-specific heterogeneity in a scatter plot (Fig. 4D).

Discussion

The result that CBV-fMRI seems to be more distinct across cortical depths compared to BOLD-fMRI (Figs. 1/2D/4A) is consistent with previous single-participant analysis [4]. The reproducibility of depth-specific activity patterns across scans (Fig. 1) provides confidence that the patterns are indicative of layer-dependent, neurally driven activity and not only session-specific artifacts. Accounting for cross-participant variability of anatomical folding patterns, consistent depth-dependent profiles could be obtained for 4 different stimuli with different input-output characteristics. These depth-dependent results are consistent with the animal literature: the modulation of sensory input (e.g., tapping with vs. without touch) modulates the activity in upper layers. Modulation of the output (e.g., tapping with touch vs. touch only) modulates deeper layers.

Conclusion

Depth-dependent results are reproducible across days. Depth-dependent results can be compared across participants with a new approach using a standardized 2D grid. Depth-dependent profiles are highly variable within cortical areas. This suggests that depth-dependent fMRI will be a useful tool to investigate directional activity in neuroscientific application.

Acknowledgements

We thank Kâmil Uludag for the suggestion to introduce the more precise descriptive terminology of cortical ‘depths’ and ‘distances’ instead of ‘layers’ and ‘columns’. Initial single-participant results of this study have been presented at last year’s ISMRM in Singapore #948. This research is supported by the American NIMH Intramural Research Program.

References

[1] Muckli et al., Current Biology, 2015, 20:232690-2695;

[2] Kok et al., Current Biology, 2016, 26: 371-376;

[3] Huber et al., NeuroImage, 2015, 107:23-33;

[4] Huber et al., ISMRM, 2016, #948;

[5] Talagala et al., MRM, 2015, 75:2362-2371;

[6] Poser et al., NeuroImage, 2010, 51:261-266;

[7] Waehnert et al., NeuroImage, 2014, 93:210-220;

[8] Caulo et al., Americ. J. Neuroradiology, 2007, 28:1480-1485;

[9] Havlicek et al., OHBM, 2016, #7273

[10] Weiler et al., Nature Neuroscience, 2008, 11:360-366;

Figures

Functional maps of tapping induced signal change ((Sact-Srest)/Srest). Data are thresholded to show response magnitude variation, not statistical significance. The results show the same participant repeatedly scanned over a 6 month period. Due to slightly different slice positioning across sessions, the thumb areas appear slightly rotated. In addition to the activity maps in the native EPI space, activity maps are also smoothed within layers (FWHM=0.75 mm) for better visibility of the depth-dependent features. Note that while highly active voxels in upper layers align along a relatively clear line, strong activity in deeper layers is confined to a few voxels only.

Accounting for inter-participant variations in the hand-knob folding pattern. A/B) depict cortical depths and cortical distances that are grown within the unique folding pattern of the participant-specific M1 area identified in native EPI space. Within this 2D space of cortical depths and cortical distances, functional responses can be averaged and compared across participants (C). Cortical distances are analyzed with the reference of the smallest bending radius lateral of the hand-knob along the templates in [8] (red arrows). The depth-dependent profiles of the presumed position of the thumb representation are given in panel D.

Average depth-dependent activity along the cortical ribbon for tasks with modulated input-output activity. It can be seen that the depth-dependent activity features are heterogeneous across the cortical distance (black/white arrows). The cortical distances representing thumb and index finger motor activity (black arrow) are highly modulated across different sensory-motor tasks. Dependent on the input-output activity of the respective tasks, those distances show positive and negative activity in upper and lower depths. The cortical distances representing ring/pinky finger and parts of the hand (white arrow) are less modulated by different tasks. Those distances are mostly dominated by signals by superficial depths.

A) Depth-dependent profiles across different tasks within the thumb representations as shown in Fig. 2. Based on simplified circuitry models taken from the animal literature [10], activity in superficial depths can be interpreted as post-synaptic cortico-cortical input (e.g., from premotor, S1, etc.). Activity in deeper cortical depths can be interpreted as post-synaptic activity triggering output activity (e.g., cortico-spinal tract). Plotting normalized activity in ‘input’ vs ‘output’ depths shows the discriminability of the different tasks across participants (D). VASO activity in superficial and deep cortical depths match the expected modulations in cortico-cortical input and cortico-spinal output, respectively.

Proc. Intl. Soc. Mag. Reson. Med. 25 (2017)
0237