Image-domain assessment and quantitative evaluation of acoustic cavitation using singular-value decomposition and magnitude-based spectral gating of B-mode imaging

Acoustic cavitation is a transient and spatially heterogeneous process that governs energy deposition and mechanical effects during high-intensity focused ultrasound (HIFU) exposure. Active B-mode imaging is attractive for monitoring because it is readily available on clinical ultrasound systems; however, therapy-imaging coupling can introduce drifting stripe artifacts that overlap with cavitation backscatter and compromise gray-level thresholding and region of interest (ROI) based quantification. Here we present an image-domain B-mode framework for cavitation-associated feature separation and spatiotemporal quantification based on singular-value decomposition (SVD) and magnitude-based spectral gating. B-mode sequences are rearranged into a spatiotemporal matrix, decomposed into singular components, and reconstructed into full-field maps of cavitation-associated image features from finite frame stacks. Highly coherent components are assigned to quasi-steady HIFU-related echoes and background scattering, whereas candidate cavitation components are selected according to their spectral-magnitude distribution, allowing stripe components observed in the present B-mode sequences to be suppressed while intermittent candidate cavitation-associated image fluctuations are retained. Experiments in egg-white polyacrylamide gels and ex vivo bovine liver showed that the proposed method suppressed stripe artifacts, improved frame-to-frame consistency, and detected cavitation onset earlier than conventional grayscale thresholding. The image-derived cavitation area increased with both acoustic pressure and pulse length and followed the trends observed in post-exposure axial damage morphology. This unsupervised method requires neither training data nor manual ROI selection, and therefore provides a practical image-based metric for retrospective visualization and quantification of cavitation-associated features during therapeutic ultrasound exposures.