Edge computing for biomedical imaging

Biomedical imaging is increasingly constrained not by acquisition but by computation: instruments generate high-bandwidth data streams that must be reconstructed, compressed, interpreted, and governed under tight limits on latency, cost, and clinical reliability. Here we review how edge computing addresses these challenges by relocating key stages of the instrument-to-inference stack (i.e. preprocessing, reconstruction, compression, visualization, and AI inference) closer to the point of measurement. We organize the literature by architectural patterns, including heterogeneous hardware acceleration and split edge-cloud pipelines. We highlight whole-slide edge tomography for autonomous cytopathology as an exemplar enabling deployable throughput at scale. We also discuss evaluation and translation considerations that link algorithmic performance to system-level outcomes, such as throughput, tail latency, data volume, and auditability.