Going into (cortical) depth: Laminar imaging and applications for cognitive neuroscience

Magnetic resonance imaging (MRI) is a popular technique to map the structure and function of the human brain. Over the past two decades, this imaging method has seen rapid developments, resulting in ever-increasing technical possibilities and applications. With the advent of ultra-high (7 Tesla and higher) field MRI scanners, it has become possible to investigate both structure and function of the human brain at a sub-millimeter resolution. Sub-millimeter resolutions enable the investigation of cortical depth-dependent signals reflecting contributions from different cortical layers, also known as laminar MRI. A particularly interesting new possibility is studying feedforward and feedback influences on information processing in the cortex as these afferents terminate in different cortical layers. While laminar fMRI is increasingly popular, some challenges remain to be addressed. One of the major assumptions underlying nearly all fMRI data-analysis techniques, is that fMRI responses are linearly proportional to a local average neuronal activity over a period of time. This linearity assumption largely holds for conventional-resolution fMRI. However, due to contributions of hemodynamic changes at other cortical depths to the local signal, this assumption may not be valid for laminar fMRI. Additionally, the majority of laminar fMRI studies thus far have focused on sensory and motor functions. Ultimately, however, we would want to extend this to cognitive functions and higher-order laminar information processing. In this thesis, I first show that the linearity assumption commonly made for conventional-resolution fMRI also holds for laminar fMRI. I then apply laminar fMRI to answer questions about the information processing flow in cognitive functions. Specifically, I look at processing of numerosity information -perception of the set size of a group of items- across cortical depth in a piece of numerosity-responsive cortex in the parietal lobe. For this, I use a population receptive field-based approach to model the preferred numerosity and tuning width for each voxel. The current analysis shows that laminar information processing in numerosity-selective cortex appears to be different from laminar information processing in early visual cortex. While laminar fMRI has a great spatial resolution, this comes at the cost of a limited temporal resolution. To obtain a higher temporal resolution while maintaining a reasonable spatial resolution, I use intracranial electrodes to directly record activity of neuronal populations. I find a neuronal population tuned to numerosity seven, with decreasing responses for displays of higher and lower numerosity. The location of this population roughly corresponds to the parietal site of the most consistently identified numerosity map in fMRI studies. Moreover, I show that the recorded responses do not reflect typical visual responses, further supporting the apparent differential information processing of cognitive processes. In summary, I show that the linearity assumption holds for laminar fMRI. This enables wider application of laminar fMRI to study sensory and cognitive processes. Using our own methods, I conclude that laminar fMRI is a useful tool to study cognitive processes -in this case numerosity perception- and that this can lead to new insights about information processing in the human brain.