Abstract Although quantitative ultrasound has crossed the threshold from research tool to routine clinical adjunct, current techniques still only interrogate tissue at the millimeter scale. Direct, micrometer-resolved insight into tissue structure, comparable to histology, remains an unmet need. The Scatterer Reconstruction (ScatRec) method, a non-stationary, deconvolution-based technique, shows promise in addressing this need.We improved the ScatRec algorithm and introduced three upgrades to improve its robustness: (i) Anisotropic total-variation, (ii) a Gaussian-noise fidelity term, and (iii) amplitude bound constraints. Additionally we bridge the gap to real work applicationby utilizing a spatially invariant point spread function. We then evaluated the enhanced reconstruction capabilities using in silico scatterer phantoms. For the first time, we analyzed the resolution limits with several two-scatterer phantoms with different scatterer distances. We tested the reconstruction quality and accuracy with phantoms containing randomly distributed scatterers and a signal-to-noise ratio (SNR) ranging from infinity to 10.Our two-scatterer phantoms showed that our proposed method at 18 MHz has an effective scatterer resolution of 38.5 µm x 156 µm in the axial and lateral directions, respectively, which is 2.6 times better than conventional B-mode. For randomly distributed scatterers, we quantified the reconstruction quality (measured by the normalized correlation coefficient, NCC) and the accuracy (indicated by the relative deviation of the effective acoustic concentration, EAC, compared to the ground truth). Compared to the original ScatRec, the NCC improved 3.7-fold, and the EAC 15.5-fold across realistic SNR of 40.Our feasibility analysis suggests that in vivo micro-structural ultrasound for scatterer reconstruction is within reach, opening a path toward "ultrasonic histology" for diseases that are currently diagnosed only by biopsy.
Rupinski et al. (Thu,) studied this question.