Multicycle Temporal Compounding Sonography for Despeckling at Specified Respiratory Phases.

Abstract

This work presents a breath-hold-free abdominal ultrasound method for both clinical and ICU settings, addressing speckle noise and motion artifacts. Its key contribution lies in respiratory phase-guided multicycle temporal compounding - identifying motion states across cycles to fuse images for target despeckling. A synergistic framework combines detail-preserving anisotropic diffusion (for speckle removal) and guided filtering (for structure transfer), while the outputs progressively refine as acquisition proceeds. Evaluated on 31 ventilator-assisted and 66 spontaneous respiration datasets, the method achieved average $\beta =0.931$ /0.930, the speckle suppression and mean preservation index (SMPI) of 0.483/0.584, and the Pratt's figure of merit 0.842/0.827, outperforming prevalent despeckling methods in edge preservation and despeckling. With computational efficiency (~2.65s/iteration) comparable to established methods and applicability across respiration styles, it enables reliable motion-artifact-suppressed imaging without breath-holding, streamlining abdominal US workflows in critical care and routine practice.