Three-dimensional evaluation of microvascular networks using contrast-enhanced ultrasound and microbubble tracking

Abstract

Evaluating tumor microvascular networks with use of contrast-enhanced ultrasound (CEUS) imaging and one-dimensional (1D) linear array transducers have inherit limitations as tumors exist in volume space. The use of a mechanical sweep allows users to overcome this limitation. To that end, we have developed a new method by which a 1D linear array transducer can be mechanically scanned over a region-of-interest to capture a volume of data allowing for the evaluation of microvasculature structures in 3D space. After intravascular injection of a microbubble (MB) contrast agent into a developing chicken embryo, a sequence of CEUS images were acquired using a Vevo 3100 scanner (VisualSonics Inc) and taken at multiple tissue cross-sections. The CEUS images were processed with a singular value filter (SVF) to help remove any clutter signal. MB localization was performed, and frame-to-frame MB movement was analyzed to produce spatial maps depicting blood flow and velocity at each tissue cross-section. Reconstruction of all images allowed visualization of microvascular networks and blood velocity distribution in volume space.

I. Introduction

Contrast-enhanced ultrasound (CEUS) imaging involves the administration of intravenous microbubble MB) contrast agents to help visualize blood flow in vascular structures [1]. Real-time CEUS has become an attractive clinical imaging modality as it allows for dynamic and repeat examinations. This functionality has been exploited for applications including the assessment of tissue perfusion in both 2-dimensional (2D) [2]-[5] and 3-dimensional (3D) space [6]-[8]. A more recent innovation was in the development of super-resolution ultrasound (SR-US) imaging systems and methods [9]. This new US modality is based in part on the accurate localization and enumeration of individual MB positions from a sequence of CEUS images [10]. By allowing visualization of microvascular structures considerably smaller than the wavelength of the US pulse used during imaging, SR-US has shown tremendous promise for evaluating tumor angiogenesis [11], [12] and microvascular dysfunction in skeletal muscle tissue due to diabetes [13]-[15]. The incorporation of MB tracking from a sequence of the CEUS images is a processing approach that produces SR-US images and allows quantification of blood flow and velocity at the microvascular level [16]-[19].

Experimental studies that are conducted using a one-dimensional (1D) linear array transducer allows one to make decisions from the acquired planar CEUS images. When tracking MBs, the assumption is that the MBs stay within the CEUS image section during the tracking process [17]. To overcome limitations with MB tracking using 2D images, Christensen-Jeffries et al. [20] used a pair of linear array transducers (one active and one passive) placed orthogonally to each other. By combining the data, it was shown that 3D SR-US images of microvascular structures and blood velocity could be produced.

The purpose of this research project was to introduce an approach for generating 3D super-resolution and super-resolved velocity maps using a preclinical US scanner equipped with a linear array transducer. Mechanical translation of the US transducer allows acquisition of CEUS images. By finding the MB centroids and utilizing MB tracking, velocity maps at each tissue cross-section can be generated and combined in 3D space.

II. Methods

Fertilized White Leghorn chicken eggs were obtained from a commercial vendor (Texas A&M University, College Station, TX) and incubated in a forced-draft incubator (GQF Manufacturing Company Inc, Savannah, GA) maintained at 37°C and 60% humidity until Hamburger and Hamilton (HH) stage 18. CEUS imaging was performed using a high-frequency preclinical US system (Vevo 3100, FUJIFILM VisualSonics Inc, Toronto, Canada) equipped with a MX250 linear array transducer. This transducer has a center frequency of 21 MHz and an axial resolution of about 75 μm. A pulled glass pipette was loaded with 10 μL of a MB contrast agent (Definity, Lantheus Medical Imaging, N Billerica, MA). While visualizing the embryo with a stereomicroscope, MBs were introduced into the embryo’s vascular network by microinjection. Chicken embryos were then imaged as the fixed US transducer was mechanically scanned over a region-of-interest (ROI; steps = 8, step size = 0.09 mm) while acquiring a sequence of images at each spatial location (N = 1000).

The acquired CEUS images were processed using a singular value filter (SVF) to remove any stationary clutter artifacts [21]. After MB localization at each cross-section, MB tracking was performed using a nearest neighbor (NN) search algorithm and measurement of Euclidean distance [22]. MBs were defined to be moving in a positive direction if a change in position along the x-axis between the n and n + 1 frames was calculated. The first condition set to determine if a MB was being tracked properly was that the change in the x-axis must be the same between all locations along the same track. This means that if the first pair in a MB track shows a change in the x-axis as being positive, then this value should be true for the remainder of the track. The second condition set was that a MB would only be considered to have been tracked if locations were validated in 5 or more sequential frames [9]. The third condition for MB tracking was establishing a threshold of maximum expected blood velocity, e.g. 20 mm/s, in microvasculature with a diameter of less than 200 μm [23]. If a MB was considered to have been successfully tracked, then the Euclidean distance between each location is placed into a new image where the value is the distance between locations. If there is more than one MB detected at a given location the final value was represented as the mean distance. The Euclidean distances where converted to velocities by multiplying them by the US imaging frame rate.

A zoomed in microvascular segment from a SR-US image is depicted in Figure 1. Using this ROI, workflow for the proposed method is illustrated in Figure 2. Starting with MB localization and SR-US image formation, the approach then performs MB tracking at each spatial location to create mean blood velocity maps. These planar maps are combined to produce a 3D SR-US and super-resolved velocity images.

Fig. 1.

Super-resolution ultrasound (SR-US) map derived from a sequence of contrast-enhanced ultrasound (CEUS) images from a developing chicken embryo injected intravascularly with a microbubble (MB) contrast agent.

Fig. 2.

Flowchart depicting the process for creating a three-dimensional (3D) velocity map during SR-US image formation. Starting with MB localization from 1000 frames at each of the 8 different spatial positions, two-dimensional (2D) MB tracking is used to create a velocity map. These 2D SR-US and super-resolved velocity maps are combined to form the final 3D volume.

III. Results And Discussion

CEUS image processing was able to simultaneously track MB position and movement. A representative 3D super-resolved microvascular velocity map reconstruction from the developing chicken embryo is shown in Figure 3. This velocity map was created from a sequence of CEUS images that were acquired at set positions as the US transducer was mechanically translated over the target tissue. These results indicate the blood velicity in this microvascular network ranged from 1 to 5 mm/s, which is consistent with previous reports [18].

Fig. 3.

3D super-resolved microvascular velocity map reconstruction from the developing chicken embryo.

It was shown previously that a pair of linear array transducers can be used to obtain a 3D super-resolved velocity map [20]. The method introduced in this study overcomes an orthogonality limitation and potential clinical obstacles by using a single linear array transducer and CEUS-based MB tracking. While the methodologies are different, 3D super-resolved velocity results are comparable. To make improvements upon this study, a more robust MB tracking algorithm could be used to increase the ability to follow MBs and throughout a larger ROI. Further directions would be to implement this method with previous studies that have expressed limitations due to being confined data acquired from 1D linear array transducers.

IV. Conclusion

This research introduces a method for creating 3D super-resolved velocity maps using a preclinical US scanner and mechanical translation of the imaging transducer.