Ultrasound matrix imaging for 3D transcranial in vivo localization microscopy

TL;DR

This study demonstrates that ultrasound matrix imaging (UMI) can robustly compensate skull-induced aberrations and multiple scattering to substantially enhance transcranial ultrasound localization microscopy (ULM) in vivo. By recording a high-dimensional reflection matrix and extracting aberration phase laws via iterative phase reversal, UMI enables post-processing synthesis of accurate focusing across depth, markedly improving microbubble detection and localization for 3D brain vasculature imaging. In sheep, UMI-corrected ULM yields higher-contrast, higher-resolution vascular maps that align more closely with gold-standard MRA, and reveals smaller vessels previously obscured by skull-induced distortions. The approach offers a nonionizing, transcranial imaging pathway with potential relevance to human stroke observation and other cerebral microvascular pathologies, while also providing a benchmarking framework against other aberration-correction methods. Overall, the combination of UMI and ULM delivers artifact-free, micrometer-scale 3D cerebrovascular imaging in vivo, paving the way for clinical translation and cross-domain wave-field imaging applications.

Abstract

Transcranial ultrasound imaging is usually limited by skull-induced attenuation and high-order aberrations. By using contrast agents such as microbubbles in combination with ultrafast imaging, not only can the signal-to-noise ratio be improved, but super-resolution images down to the micrometer scale of the brain vessels can also be obtained. However, ultrasound localization microscopy (ULM) remains affected by wavefront distortions that limit the microbubble detection rate and hamper their localization. In this work, we show how ultrasound matrix imaging, which relies on the prior recording of the reflection matrix, can provide a solution to these fundamental issues. As an experimental proof of concept, an in vivo reconstruction of deep brain microvessels is performed on three anesthetized sheep. The compensation of wave distortions is shown to markedly enhance the contrast and resolution of ULM. This experimental study thus opens up promising perspectives for a transcranial and nonionizing observation of human cerebral microvascular pathologies, such as stroke.

Figures (16)

• Figure 1: Sheep experiment : A multi-sequence acquisition. (A) Angiographic MRI. (B) Transcranial power Doppler for positioning the probe. (C) Acquisition of the reflection matrix to estimate local aberration laws with 3D-UMI. (D) Ultrafast imaging for super-resolved ULM images. The images shown in this figure are for illustrative purposes only and do not provide quantitative results.
• Figure 2: Principle of ultrasound matrix imaging. (A) Focused reflection matrix contains the time-gated response between virtual source ($\bm{\rho}_{\textrm{in}}$) and detector ($\bm{\rho}_{\textrm{out}}$) located at the same depth $z$. (B) An output projection of the focused reflection matrix in the transducer basis ($\mathbf{u}_{\textrm{out}}$) provides the reflected wave-fronts induced by each virtual source ($\bm{\rho}_{\textrm{in}}$) at depth $z$. (C) Those wave-fronts are realigned to extract the wave-front distortions induced by the mismatch between the real speed-of-sound distribution and the wave velocity model. Seen from the focused basis, this operation leads to an angular de-scan of each virtual source at the same position, leading to the synthesis of a virtual guide star. (D) Exploiting the correlations between each distorted wave-front, an iterative phase reversal algorithm extract an aberration phase law that can be used to compensation for wave distortions induce by the skull and ideally retrieve a diffraction-limited focal spot across the field-of-view.
• Figure 3: Influence of cranial heterogeneities on aberration and multiple scattering. (A) Micro-CT of each sheep skull. Left: sheep n$^{\textrm{o}}$4, middle: sheep n$^{\textrm{o}}$5, right: sheep 6. (B) Corresponding RPSF maps at depth $z=35$ mm. (C and D) RPSF extension and multiple scattering rate as a function of depth, extracted by UMI, for sheeps n$^{\textrm{o}}$4 (green), 5 (purple line) and 6 (orange line).
• Figure 4: Ultrasound matrix imaging in the sheep brain. (A) Confocal volume extracted from the diagonal of the focused $\mathbf{R}-$matrices (Methods, Eq. \ref{['confocal']}). (B, C, and D) Maps of initial RSPFs, aberration phase laws, and corrected RPSFs, respectively, at three different depths $z=32$ mm (top), $z=50$ mm (middle) and $z=68$ mm (bottom). Each RPSF is displayed in a scan range $\boldsymbol{\Delta \rho}=\boldsymbol{\rho}_\textrm{out}-\boldsymbol{\rho}_\textrm{in}$ (see Methods) that varies from $-15$ to $+15$ mm in both $x$ and $y$ directions. (E) Confocal image after aberration correction. The results shown here correspond to ultrasound data acquired on sheep n$^{\textrm{o}}$6.
• Figure 5: Enhancing micro-bubble localization accuracy with UMI. (A and B) Sagittal middle slice of the ultrasound image without and with aberration correction at a given time, respectively. The position of the detected microbubbles is highlighted by red dots. The blue and orange boxes show areas where false alarms are reduced by reducing the side lobes of the PSF. The purple region shows a region where aberration compensation allows the detection of a microbubble. The red square highlights a zone where a microbubble is reassigned at a new location. (C) Number of localized microbubbles as a function of the acquisition time $t$ with (green) and without (pink) UMI. (D) Increase of the microbubble detection rate with UMI for different track lengths. The data shown here correspond to sheep n$^{\textrm{o}}$ 6 (acquisition 4).