Super-resolution wave-dispersion analysis in optical coherence elastography using multiple signal classification (MUSIC).

In vivo wave-based optical coherence elastography (OCE) is often constrained by short propagation distances, sparse spatial sampling, and low signal-to-noise ratios (SNRs), which degrade the wavenumber localization and phase-velocity accuracy of conventional two-dimensional fast Fourier transform (2D-FFT) analysis. Here, we introduce Multiple Signal Classification (MUSIC), which exploits signal-noise subspace orthogonality to improve wavenumber localization. Numerical simulations and agar phantom experiments compared MUSIC with 2D-FFT under impulse (0-800 Hz) and chirp (0-5000 Hz) excitation. Under sparse sampling (6 points over 2 mm), MUSIC maintained absolute mean biases below 1.5% for impulse excitation and 0.2% for chirp excitation across prescribed velocities of 2-20 m/s, whereas the 2D-FFT bias approached 20% at higher velocities. MUSIC also produced usable phase-velocity estimates at SNRs of 2:1 for impulse excitation and 1:5 for chirp excitation. Joint simulations spanning 6-121 spatial samples and 2-40 mm apertures further showed that sample number affected 2D-FFT accuracy more strongly than aperture length within the evaluated range. In dispersive Kelvin-Voigt simulations (μ 1 = 25 kPa, η = 0.1-5 Pa·s), the absolute mean phase-velocity bias of MUSIC remained below 0.4%, whereas 2D-FFT underestimated phase velocity by up to 17.7% under sparse sampling. Agar phantom experiments showed that MUSIC-derived phase velocity and fitted elastic shear modulus were less sensitive to spatial sample number, whereas viscosity estimates remained variable and model dependent. These findings support MUSIC as a promising alternative for dispersion analysis with improved effective wavenumber localization under practical OCE acquisition constraints.