Fast and Curious: Unveiling millisecond dynamics of population receptive fields

Understanding how the human brain processes visual information requires insight into both where and when neural activity occurs. However, non-invasive neuroimaging techniques face a fundamental trade-off: imaging techniques such as functional magnetic resonance imaging (fMRI) offer high spatial resolution, while neurophysiological methods such as magnetoencephalography (MEG) provide millisecond temporal precision. This thesis addresses this challenge by introducing a forward modeling framework that combines the spatial detail of fMRI with the temporal accuracy of MEG, enabling precise characterization of processing dynamics in the healthy human brain. Chapter 1 provides a general overview for the reader. In Chapter 2, I introduce the integrated modeling framework and demonstrate its ability to accurately predict measured visual responses with high spatiotemporal precision. Chapter 3 extends this by showing how we can extract precise time-courses from individual visual areas and quantify their temporal contributions during visual processing. Chapters 4 and 5 reveal how both experimental design and model choice affect the accuracy and specificity of the model’s performance, highlighting practical considerations for applying the method to other contexts. In Chapter 6, I present the open-source software developed for this research, which is built to be modular, accessible, and adaptable, allowing researchers to apply this forward modeling approach to a wide range of scientific questions. Chapter 7 concludes with a general discussion and concrete examples of how the modeling approach can be extended. Together, these studies advance our ability to investigate the spatiotemporal dynamics of visual processing and lay the groundwork for new discoveries in visual neuroscience and beyond.