What problem does it solve? Raw fMRI BOLD time series contain unwanted frequency content and noise that corrupt downstream connectivity, decoding, and statistical analyses. This Skill provides model-level guidance for classical temporal filtering workflows, producing denoised BOLD images and cleaned ROI time series without deep-learning methods. ## Core Features & Use Cases - Temporal Filtering Routes: Apply high-pass, low-pass, or band-pass filtering to preprocessed BOLD time series using a specified TR and frequency band. - Confound-Aware Cleaning: Optionally combine filtering with confound regression, standardization, and masking for cleaner voxel-wise or ROI-wise signals. - Delegation Workflow: Coordinates with fmri-skill for preprocessing planning and nilearn-tool for concrete filtering execution and cleaned image export. - Use Case: A researcher with preprocessed resting-state fMRI data wants to retain the 0.01-0.08 Hz band before functional connectivity analysis; this Skill guides the band-pass filtering run and exports the denoised image plus QC summaries. ## Quick Start Ask the assistant to band-pass filter your preprocessed resting-state BOLD image between 0.01 and 0.08 Hz with TR 2.0 and export the denoised image to a run_models_output directory.
