Layer connective fields from ultrafast resting-state fMRI differentiate feedforward from feedback signaling

Deciphering the directionality of information flow in cortical circuits is essential for understanding brain dynamics, learning, and neuroplasticity after injury. However, current non-invasive methods cannot distinguish feedforward (FF) from feedback (FB) signals across entire networks, including deep brain regions. Here, we present a novel approach combining ultrafast fMRI with a Layer-based Connective Field (lCF) model to disentangle FF from FB signaling. Our findings reveal that lCF size, an indicator of spatial information integration, differentiates FF and FB activity through distinct layer-specific connectivity patterns during spontaneous activity, challenging the notion that FF signals are solely stimulus-driven. FF connectivity follows an inverted U-shape, peaking in layer IV, while FB exhibits a U-shaped pattern, with peaks in layers I and VI. These profiles generalize across sensory pathways (visual, somatosensory, and motor) and reveal injury-induced network reorganization, such as LGN bypassing V1 to provide direct FF input to higher visual areas.