Abstract Functional ultrasound imaging (fUSI) is a promising tool for studying brain activity in awake and behaving animals, offering insights into neural dynamics that are more naturalistic than those obtained under anesthesia. However, motion artifacts pose a significant challenge, introducing biases that can compromise the integrity of the data. This study provides a comprehensive evaluation and benchmarking of strategies for detecting and removing motion artifacts in transcranial fUSI acquisitions of awake mice. We evaluated 792 denoising strategies across four datasets, focusing on clutter filtering, scrubbing, frequency filtering, and confound regression methods. Our findings highlight the superior performance of adaptive clutter filtering and aCompCor confound regression in mitigating motion artifacts while preserving functional connectivity patterns. We also demonstrate that high-pass filtering is generally more effective than band-pass filtering in the presence of motion artifacts. Additionally, we show that with effective clutter filtering, scrubbing may become optional, which is particularly beneficial for experimental designs where motion correlates with conditions of interest. Based on these insights, we propose two optimized denoising paradigms tailored to different experimental constraints, providing practical recommendations for enhancing the reliability and reproducibility of fUSI data. Our findings challenge current practices in the field and have immediate practical implications for existing fUSI analysis workflows, paving the way for more sophisticated applications of fUSI in studying complex brain functions and dysfunctions in awake experimental paradigms.
Diebolt et al. (Thu,) studied this question.