Spatiotemporal pattern extraction from functional neuroimaging

We present an adaptation of a tensor decomposition framework – SWOTTED – for spatiotemporal pattern extraction in functional neuroimaging. SWOTTED operates without prior assumptions, allowing the detection of recurrent patterns and potentially, rare signal variations, such as those observed in pathological conditions. Applied to functional ultrasound (fUS) recordings of rat brains under visual stimulation, SWOTTED successfully identified stimulus-related activation patterns while jointly capturing spatial and temporal dynamics. Beyond task-based paradigms, its ability to extract patterns without predefined assumptions makes it particularly suited for investigating intrinsic brain activity and spontaneous neural fluctuations. Additionally, SWOTTED efficiently processes entire fUS stacks and can be applied to multiple subjects simultaneously, making it a promising tool for expanding neuroimaging applications. These findings position SWOTTED as a more expressive approach to explore complex brain dynamics beyond traditional tensor decomposition methods. Perspective works encompass performance robustness optimization via pattern selection and hyperparameter tuning.