Multi-parametric NIR-II fluorescence perfusion imaging towards quantitative stroke evaluation

Preclinical stroke research suffers from the absence of high-throughput, quantitative perfusion imaging tools that can effectively connect functional hemodynamics with structural injury. NIR-II fluorescence imaging, a promising technique, faces limitations due to a dearth of standardized analysis protocols and its inability to capture deep infarct regions.

The researchers present FIMPA, an integrated analytical framework for standardized, multi-parametric stroke evaluation utilizing NIR-II fluorescence perfusion imaging. FIMPA systematically processes image sequences by correcting motion and registering to an atlas, creating an anatomically aligned library of 125 vascular and hemodynamic parameters.

Statistical screening identifies 67 stroke-correlated features, allowing for objective, data-driven quantification of perfusion deficits. FIMPA-SI, a novel deep learning-based index developed within FIMPA, utilizes a ViT-UNETR framework to resolve spatial infarct boundaries. By applying self-supervised pre-training with MRI data, FIMPA-SI effectively suppresses dominant vascular signals, translating multi-parametric dynamics into precise stroke severity maps.

Tested in tMCAO mice models, FIMPA-SI demonstrated high spatial concordance with MRI and significantly outperformed conventional laser speckle contrast imaging and single-parameter metrics. This innovative solution promises to expedite preclinical therapeutics and advance optical neuroimaging.