Streak-Aware Localization Microscopy Enables High-Throughput Brain Imaging Across Platforms

Abstract

Optical, ultrasound, and optoacoustic localization microscopy based on microparticle tracking has enabled surpassing the resolution limits imposed by ultrasound diffraction and optical diffusion in tissues. However, its reliance on high-speed (kilohertz) data acquisition systems for precise emitter localization and tracking substantially increases methodological complexity and data storage demands, limiting scalability and applicability beyond specialized benchtop platforms. Here, we present streak-aware localization microscopy (SALM) that employs localization- and tracking-free deep learning model to convert motion-blurred streaks originating from low frame rate recordings of flowing emitters into super-resolved structural and functional readouts. In optical implementations, SALM exploits streaks captured by low-speed cameras to recover capillary-level cerebrovascular maps across a variety of benchtop, miniaturized, and second near-infrared preclinical imaging platforms, achieving over 30-fold reduction in the reconstruction time compared to conventional localization pipelines. We further introduce three coded excitation strategies that embed finer time-varying vectorial flow signatures into individual streaks, enabling single-frame velocimetry and video-rate hemodynamic imaging. Extending SALM to ultrasound imaging enables high-fidelity vascular imaging with centimeter-scale penetration in rhesus macaque and rat brains while reducing plane-wave compounding frame rates by up to one order of magnitude. By overcoming long-standing trade-offs between spatiotemporal resolution and hardware complexity, SALM offers a flexible and scalable framework for next-generation super-resolution microscopy.