An Experimental Protocol for Assessing the Performance of New Ultrasound Probes Based on CMUT Technology in Application to Brain Imaging

A new family of ultrasound probes based on MEMS transducers is currently emerging. CMUT components are constructed on silicon and have a significant potential in terms of wideband operation, good thermal efficiency and low fabrication cost. This video describes a robust and repeatable experiment for protocol for accessing the capabilities of CMUT ultrasound probes in application to brain imaging during the early prototyping stages of design and development.

Capacitive Micromachined Ultrasound Transducers or CMUT for short are micro-fabricated arrays of tiny, electrostatic actuators, each acting as a two-ultrasound transducer. More in detail, in each CMUT cell an electrical excitation produces a dynamic force that then uses mechanical vibration of the moving plate. The acoustic waves thus generated propagate through the medium.

The effect is reversed during reception. Acoustic waves generated by tissue scattering cause vibrations of the moving plate that change the capacitance of the CMUT cell. This produces an electrical signal which is further amplified and transmitted to the ultrasound system.

This is the linear CMUT probe used for the experiments shown in this video. The probe head on the left contains the the CMUT array covered by an acoustic lens. The image on the right shows the packaged probe head together with the multichannel cable that connects it to the ultrasound system.

This is the overview of the protocol that will be shown. In the preliminary design step, the positioning strategy for landmark glass spheres is defined and an Agar Phantom is prepared for calibration. In the sample preparation set, the bovine brain is prepared by inserting the glass spheres in predefined positions and fixating it in formalin.

Then the prepared bovine brain is passed into the MR scan to acquire 3D images. Qualitative poses are then defined on 3D MR images with the help of experts. Setting up the experimental environment requires planting the passive tracking markers to the ultrasound probe and connecting the ULOB scanner and the motion tracking system.

Two calibration sets are then required, one for clamped markers on the ultrasound probe and another for converting motion tracker positions into positions in the target 3D MR image. Ultrasound image acquisitions will be performed using our real time visualization tool for matching predefined, qualitative poses. In the final step of post processing, ultrasound images are registered on the target 3D MR image.

Freehand moving images are also reconstructed into 3D volumes. A preliminary operation for the calibration activity to be shown later on involves using a phantom made with argala jelly with patterns of glass spheres embedded in it. For maximum realism and reliability of the assessment protocol, bovine brains obtained from the standard food supply chain were chosen as imaging targets.

Specimens were prepared by inserting glass spheres at predefined positions. Specimens were then immersed in a formalid solution and kept in this way for at least three weeks until complete fixation. Multiple magnetic resonance scans of the bovine brain were then acquired using different waiting sequences.

As it can be seen, glass spheres are well evident in MR images. The repeatability of the acquisition of ultrasound images is enforced by the definition of 12 virtual poses. Each of these predefined poses points to a region in the bovine brain which is rich in morphological features and tissue heterogeneity.

A specific software routine provides realtime visual feedback about the current position and orientation of the ultrasound probe using data from the motion tracking system. This is the complete and integrated ultrasound system. On the mechanical arm, the CMUT probe connected to the ULA-OP system, which uses a computer connected by a USB for control, programming, and visualization.

And the ultrasound probe used in this protocol must be clamped to an arrangement of reflective markers so that it's position and orientation can be recorded by the motion tracking system. The phantom is immersed in a water tank within the field of view of an infrared position sensor which records the position and the orientation of a rated pattern of reflective markers and transmits them by a USB to a house computer. The position of each glass sphere is recorded with a pointer tool.

The recording is driven by the operator. The objective of the calibration is finding the geometric transformation from the intrinsic image frame of the probe to the reference frame of reflective markers using the positions of the glass spheres as extracted from the ultrasound images. Once the spheres are made visible in the ultrasound images, the mechanical arm is blocked and the image is acquired.

Calibration ultrasound images are then imported in a visualization tool in which the positions of the spheres are marked manually and saved in a file. The geometric transformation is then computed automatically by comparing all these positions with the corresponding positions recorded by the motion tracking system. For the position of ultrasound images, which is the objective of the protocol, after complete washing, the bovine brain is first immobilized on a plastic plate and then immersed in a water tank within the field of view of the motion tracking system.

The positions of all the glass spheres embedded in the bovine brain are manually recorded with the motion tracking system and the pointer tool. All the positions obtained in this way are stored in a log file. The objective of the second calibration activity is finding the geometric transformation from the physical space of the motion tracking system to the face of each MR image.

This image shows all the recorded pointer positions inside the 3D phase of one MR image. Here we see the visual feedback routine being initialized after calibration. The 12 predefined virtual poses are visible to the operator in the 3D view.

Here the red cone corresponds to the ultrasound probe being handheld by the operator. The probe position is recorded by the motion tracking system and transformed in realtime to provide visual feedback. Here the ULOB system is recording images at two frequencies simultaneously.

Once an edge is considered satisfactory, the ultrasound image is frozen and saved on disk. A different acquisition modality is by a freehand sweep of the ultrasound probe made while the ULOB system is programmed to perform fast, repeated acquisitions in sequence. In this way, entire sub volumes of the bovine brain can be covered in each sweep.

The result is a sequence of slices, each being a bmod ultrasound image that is saved to the disk as a whole. Each sequence is intended for post-processing. When each slice is associated with the instantaneous position and orientation provided by the motion tracking system and combined to form a three-dimensional picture of the corresponding sub volume.

Overall, the objective of image post-processing in this protocol is to enable a comparative visual analysis of ultrasound images with MR images after special registration. Here we see the dataset of 12 images corresponding to predefined poses combined in a 3D reconstruction after spatial registration. Using a visualization software tool, each ultrasound image corresponding to a predefined pose can be seen in superposition to the corresponding slice of the three-dimensional MR image and the purity of the image contents can be evaluated.

In this final animation, we can see the results of the three-dimensional reconstruction of the entire sequences of the sliced images acquired with a single sweep in freehand mode. Apart from a few reconstruction artifacts, the quality of these images gives a clear idea of the performance level that can be achieved with the technology and application to brain imaging. The main result achieved with the experiments described is proving the effectiveness of the protocol for the assessment of the CMUT probe imaging performance and on the integrated system of hardware and software tools for each execution.

In addition, complete data sets of registered ultrasound and MR images have been obtained and this can be used to support farther status in automated image analysis. In conclusion, the experimental protocol presented in this video is a effective and repeatable method for the assessment of imaging capabilities of innovative CMUT ultrasound probes. In its realization, the integration of open hardware and software tools proved fundamental.

The comparative ultrasound to MR visual analysis that the method allows can be further enhanced by integrating more sophisticated software techniques.