Voets, N. L. | bioRxiv Science

bioRxiv2023

Scene-selectivity in CA1/subicular complex: Multivoxel pattern analysis at 7T

Prior univariate functional magnetic resonance imaging (fMRI) studies in humans suggest that the anteromedial subicular complex of the hippocampus is a hub for scene-based cognition. However, it is possible that univariate approaches were not sufficiently sensitive to detect scene-related activity in other subfields that have been implicated in spatial processing (e.g., CA1). Further, as connectivity-based functional gradients in the hippocampus do not respect classical subfield boundary definitions, category sensitivity may be distributed across anatomical subfields. Region-of-interest approaches, therefore, may limit our ability to observe category selectivity across discrete subfield boundaries. To address these issues, we applied searchlight multivariate pattern analysis to 7T fMRI data of healthy adults who undertook a simultaneous visual odd-one-out discrimination task for scene and non-scene (including face) visual stimuli, hypothesising that scene classification would be possible in multiple hippocampal regions within, but not constrained to, anteromedial subicular complex and CA1. Indeed, we found that the scene-selective searchlight map overlapped not only with anteromedial subicular complex (distal subiculum, pre/para subiculum), but also inferior CA1, alongside retrosplenial and parahippocampal cortices. Probabilistic overlap maps revealed gradients of scene category selectivity, with the strongest overlap located in the medial hippocampus, converging with searchlight findings. This was contrasted with gradients of face category selectivity, which had stronger overlap in more lateral hippocampus, supporting ideas of parallel processing streams for these two categories. Our work helps to map the scene, in contrast to, face processing networks within, and connected to, the human hippocampus.

neuroscience↗

bioRxiv2021

A note on decoding elicited and self-generated inner speech

A recent result shows that inner speech can, with proper care, be decoded to the same high-level of accuracy as articulated speech. This relies, however, on neural data obtained while subjects perform elicited tasks, such as covert reading and repeating, whereas a neural speech prosthetic will require the decoding of inner speech that is self-generated. Prior work has, moreover, emphasised differences between these two kinds of inner speech, raising the question of how well a decoder optimised for one will generalise to the other. In this study, we trained phoneme-level decoders on an atypically large, elicited inner speech dataset, previously acquired using 7T fMRI in a single subject. We then acquired a second self-generated inner speech dataset in the same subject. Although the decoders were trained exclusively on neural recordings obtained during elicited inner speech, they predicted unseen phonemes accurately in both elicited and selfgenerated test conditions, illustrating the viability of zero-shot task transfer. This has significant practical importance for the development of a neural speech prosthetic, as labelled data is far easier to acquire at scale for elicited than for self-generated inner speech. Indeed, elicited tasks may be the only option for acquiring labelled data in critical patient populations who cannot control their vocal articulators.

bioengineering↗