Ultrasound Super-Resolution Imaging of Neonatal Cerebral Vascular Reorganization

Abstract

During the first days of neonatal growth, the central nervous system (CNS) develops self-regulatory mechanisms to ensure constant cerebral perfusion. However, this vascular neogenesis takes place at a microscopic scale that cannot be observed with current clinical imaging techniques. Ultrasound localization microscopy (ULM) allows us to observe micro-vessels of the order of a few microns at depths of several centimeters. This can be done using conventional clinical ultrasound scanners and contrast sequences (CEUS). In this study, ULM is used to observe the human microvasculature in neonatal patients undergoing treatment for life-threatening malformations forming direct connections between the cerebral arterial and venous systems. It is observed that neuroendovascular treatment of neonatal arteriovenous malformations causes remodeling and reorganization of the cerebral vasculature by also activating corticomedullary vascular connections. ULM enables us to follow microvascular changes in human neonates with high spatio-temporal resolution. ULM may provide a novel clinical translatable tool, particularly including cerebral imaging in very young patients.

Keywords: CEUS; Vein of Galen malformation; microbubbles; super‐resolution imaging; ultrasound; ultrasound localization microscopy.

Figure 1

Perfusion dynamics of the cerebral vasculature during therapy. A) Schematic illustration of dynamic flow parameter measurements in both hemispheres. Rise (RT) and fall time (FT) can be retrieved from the time‐intensity curve of the overall microbubble signal. B) Representative ultrasound (US), contrast‐enhanced ultrasound (CEUS), and color‐coded maps of rise (RT) and fall time (FT) (left to right) during therapy (T1–T3, upper to lower rows) of a single patient. More red color‐coding indicates an increase in rise time (faster arterial inflow). Identical measurements were performed in n = 7 consecutive patients with Vein of Galen malformation. C) RT measurement of the right and D) left hemisphere for T1–T3. E) FT measurement of the right and F) left hemisphere for T1‐T3. Parameters in relation to baseline were tested with the Kruskal‐Wallis test and Dunn's post‐test. G−J) Comparison of RT and FT of the first time point in both hemispheres to n = 8 patients with no cerebral pathology. Significance tested with unpaired, non‐parametric t‐tests. K) Correlation matrix using Spearman's correlations coefficient comparing reduction in RT (ΔRT) and FT (ΔFT) with white (MRI(WM)_pre), grey matter (MRI(GM)_pre) and total magnetic resonance imaging scores (MRI(total)_pre) before and after (MRI(WM)_post, MRI(GM)_post, MRI(total)_post) intervention (The WM score consisted of four grades: normal is 5–6 points, mildly abnormal is 7–9, moderately abnormal is 10–12, and severely abnormal 13–15. Pre‐ and post‐treatment grey matter abnormalities were assessed using a three‐point scale in three regions. The total GM score was calculated by summing up the points from each region and graded into normal (3–5 points) and abnormal (6–9). Using GM and WM subscores, a total score was calculated. A total score of 19–24 points was considered as severe brain injury), Bicêtre neonatal evaluation score before (Bicêtre_pre) and after (Bicêtre_post) (Bicêtre neonatal evaluation score evaluates five objective clinical parameters: cardiac function, respiratory function, cerebral function, renal function, and hepatic function, resulting in a score between 0–21 points, lower scores indicate higher disease severity)^{[ 11b ]} and maximum neonatal multiple organ dysfunction scores (NEOMOD_max) (The NEOMOD assessment scores eight organ systems: central nervous, hemocoagulation balance, respiratory, gastrointestinal, cardiovascular, renal, acid‐base balance, and microvascular systems, with a score between 0 and 16 points, higher scores indicate higher disease severity).^{[ 14a ]} L) Magnetic resonance imaging of a single patient pre and post‐treatment. White circles represent the area of the malformation. Created with BioRender.com.

Figure 2

Macrostructural adaptions of the cerebral vasculature during therapy. A) Schematic illustration of dynamic flow parameter measurements in different cerebral regions during all time points (T1–T3). B) Representative mean time‐intensity curves with standard deviation (bright area) during therapy (T1–T3) in n = 7 patients. C) Representative color‐coded maps of rise time (RT) during therapy (T1–T3) of a single patient. Bar represents 1cm. D) RT measurement of the right and E) left cortex, F) right, and (G) left white matter, H) right and I) left putamen. J − O) Similar measurements on the left hemispherical regions. Parameters in relation to baseline were tested with the Kruskal‐Wallis test and Dunn's post‐test. Bars represent SD. Created with BioRender.com.

Figure 3

Identification of cortical layers to measure microvascular blood flow dynamics during therapy. A) The dashed box represents an example of the selection of the neonatal cortex to be analyzed. B) The upper row shows the magnified region and the lower outline of the grey (gm) and white matter (wm) regions, which are clearly separated by different vascular localizations. SSS = superior sagittal sinus. The following images show the identical section for time points T2 C) and T3 D). Schematic illustration and representative images for analyzed grey and white matter regions E). The upper row outlines the grey (red) and white matter (blue) regions, the middle row a density map, and the lower row a directivity map. Bar represents 1 cm F). Outline of the grey matter region G). Microbubble flow velocities (T1: n = 15 174, T2: n = 17 710, T3: n = 19 596), dispersity (T1: n = 15 156, T2: n = 17 711, T3: n = 19 515), tortuosity (T1: n = 15 585, T2: n = 18 209, T3: n = 20 027) and distance metric (T1: n = 15 587, T2: n = 18 210, T3: n = 20 027) for right grey matter region and all time points H). Microbubble flow velocities (T1: n = 14 841, T2: n = 20 149, T3: n = 19 685), dispersity (T1: n = 14 597, T2: n = 19 698, T3: n = 19 192), tortuosity (T1: n = 15 116, T2: n = 20 569, T3: n = 20 152) and distance metric (T1: n = 15 115, T2: n = 20 571, T3: n = 20 153) for left grey matter region and all time points I). Outline of the white matter region J). Microbubble flow velocities (T1: n = 14 841, T2: n = 19 814, T3: n = 17 600), dispersity (T1: n = 10 146, T2: n = 12 969, T3: n = 11 421), tortuosity (T1: n = 10 929, T2: n = 13 676, T3: n = 11 941) and distance metric (T1: n = 10 931, T2: n = 13 676, T3: n = 11 944) for right white matter region and all time points K). Microbubble flow velocities (T1: n = 10 167, T2: n = 14 440, T3: n = 13 918), dispersity (T1: n = 9940, T2: n = 14 305, T3: n = 13 333), tortuosity (T1: n = 10 514, T2: n = 14 825, T3: n = 14 257) and distance metric for left white matter region and all time points L). The Red dashed line indicates the mean. Parameters tested with one‐way ANOVA with Tukey post‐test. Created with BioRender.com.

Figure 4

Group‐level flow changes during therapy. A) First percentile of absolute microbubble flow velocities in the left cortical region (T1: n = 6, T2: n = 6, T3: n = 6). B) Relative changes of first percentile of microbubble flow velocities with respect to baseline (T1 = 100%) in the left cortical region (T1: n = 6, T2: n = 6, T3: n = 6). C) First percentile of absolute microbubble flow velocities in the right cortical region (T1: n = 6, T2: n = 6, T3: n = 6). D) Relative changes of 1^st percentile of microbubble flow velocities with respect to baseline (T1 = 100%) in the right cortical region (T1: n = 6, T2: n = 6, T3: n = 6). (E) 5th percentile of absolute microbubble flow velocities in the left cortical region (T1: n = 6, T2: n = 6, T3: n 6). F) Relative changes of 5th percentile of microbubble flow velocities with respect to baseline (T1 = 100%) in the left cortical region (T1: n = 6, T2: n = 6, T3: n = 6). (G) 5th percentile of absolute microbubble flow velocities in the right cortical region (T1: n = 6, T2: n = 6, T3: n = 6). (H) Relative changes of 5th percentile of microbubble flow velocities with respect to baseline (T1 = 100%) in the right cortical region (T1: n = 6, T2: n = 6, T3: n = 6). (I) 99th percentile of absolute microbubble flow velocities in the left cortical region (T1: n = 6, T2: n = 6, T3: n = 6). (J) Relative changes of 99th percentile of microbubble flow velocities with respect to baseline (T1 = 100%) in the left cortical region (T1: n = 6, T2: n = 6, T3: n = 6). K) 99th percentile of absolute microbubble flow velocities in the right cortical region (T1: n = 6, T2: n = 6, T3: n = 6). L) Relative changes of 99^th percentile of microbubble flow velocities with respect to baseline (T1 = 100%) in the right cortical region (T1: n = 6, T2: n = 6, T3: n = 6). M) 95th percentile of absolute microbubble flow velocities in the left cortical region (T1: n = 6, T2: n = 6, T3: n = 6). N) Relative changes of 95th percentile of microbubble flow velocities with respect to baseline (T1 = 100%) in the left cortical region (T1: n = 6, T2: n = 6, T3: n = 6). O) 95th percentile of absolute microbubble flow velocities in the right cortical region (T1: n = 6, T2: n = 6, T3: n = 6). (P) Relative changes of 95th percentile of microbubble flow velocities with respect to baseline (T1 = 100%) in the right cortical region (T1: n = 6, T2: n = 6, T3: n = 6). Dots show individual patient values. Parameters tested with non‐parametric Friedman test and Dunn's post‐test. Created with BioRender.com.

Figure 5

CEUS during cortical remodulation: Post‐interventional activity beyond cortical vascularity. A) Ultrasound localization microscopy density (upper) and directivity maps (lower row) for all investigated timepoints (T1–T3, left to right) and schematic outline of the subarachnoid space (green region of interest). Yellow arrows mark individual vessels, white arrows point to the medullary space, and SSS = superior sagittal sinus. Bar represents 1cm. B) Microbubble flow velocities (T1: n = 2123, T2: n = 1878, T3: n = 2124), C) dispersity (T1: n = 1397, T2: n = 1519, T3: n = 1531), D) tortuosity (T1: n = 2204, T2: n = 1995, T3: n = 2264), and E) distance metric (T1: n = 2208, T2: n = 1996, T3: n = 2265) for all time points in one patient. Parameters tested with one‐way ANOVA with Tukey post‐test. Bars represent SD. Created with BioRender.com.

Figure 6

Microvascular dynamics as novel discriminators of cerebral pathology. A) Schematic illustration of cerebral infarction after the treatment (T1 and T2), ultrasound localization microscopy density, and directivity maps (left to right). Blue codes for upward and red for downward localization. Bar represents 1cm. B) Outline of the regions of interest, green = non‐infarcted, viable tissue, orange = infarcted tissue. Bar represents 1 cm. C) Microbubble flow velocities (non‐infarcted: n = 11 116, infarcted: n = 3665), D) dispersity (non‐infarcted: n = 6782, infarcted: n = 2085), E) tortuosity (non‐infarcted: n = 11 160, infarcted: n = 3700), and F) distance metric (non‐infarcted: n = 11 169, infarcted: n = 3703) for T1 and T2. Bars represent SD. SSS = superior sagittal sinus. Created with BioRender.com.