Results

To showcase results achieved through the use of MatMRI and MatHIFU, four applications that were developed at TBRRI and the Hospital for Sick Children (Toronto, Ontario, Canada) are presented. These applications show the ability of MatMRI and MatHIFU to control MR imaging in real time for pre-clinical animal studies, generate thermal maps using T2-based thermometry, drive individual transducer elements to perform acoustic characterization, and execute custom ablation patterns.

Integration of MatMRI with a small animal MR-HIFU system for the treatment of abscesses in a murine model

MatMRI was integrated with existing software used to control a table designed for small animal MR-HIFU studies (FUS Instruments, Toronto, Ontario, Canada). Experiments were conducted to test the hypothesis that MR-HIFU can be used as a therapeutic option for the treatment of abscesses related to methicillin-resistant Staphylococcus aureus (MRSA) [8]. The animal protocol was approved by the Animal Care Committee of Lakehead University (AUP 08 2012). A 50- μ L subcutaneous injection of an MRSA strain, USA-400 bacteria, at a concentration of 7 × 10³/μ L was performed on the left flank of BALB/c mice, and an abscess of 6 ±2 mm in length formed after 48 h. The abscess was targeted using a transducer operating at 3 MHz with a focal length of 50 mm and diameter of 32 mm. The focal point was positioned 2 mm underneath the abscess, and an ultrasound exposure was applied over 9 s with an acoustic power of 25 or 35 W. Temperature maps were calculated from the coronal MR images of the subcutaneous region of the left flank using the PRFS technique. Magnetic drift was monitored in the non-heated muscle region, and a correction was applied. MR imaging parameters for thermometry were as follows: field of view (FOV) = 80 mm, pixel size = 1 mm, slice thickness = 3 mm, echo time/repetition time (TE/TR) = 16/23 ms, flip angle = 19°, acquisition matrix = 68×63, reconstruction matrix = 80, echo train length (ETL) = 9, number of excitations (NEX) = 1, and dynamic time = 0.35 s. Figure 2 shows a screenshot of the graphical user interface used to monitor and control the experiments.

End points were at days 1 and 4 after MR-HIFU treatment. For each end point, 18 animals were randomly assigned into three groups with 6 animals each: control, treatment with an acoustic power of 25 W, and treatment with an acoustic power of 35 W. Results indicated that an exposure with an acoustic power of 35 W was able to induce a significant ( p < 0.05) reduction of the bacteria concentration in the abscess when compared to non-treated abscesses and low-power exposures. The bacteria concentration (mean ± standard deviation) at the 4-day end point was 1.0 ± 1.3 × 10⁴, 0.6 ± 0.6 × 10⁴, and 0.09 ± 0.2 × 10⁴ colony-forming units/ μ L for the control, low power, and high power groups, respectively.

T2-based thermometry with MatMRI for the MR-HIFU treatment of the bone marrow

MatMRI was used to test the hypothesis that changes in bone marrow temperature can be accurately determined by measuring changes in the transverse magnetization decay time (T2) [9] since magnetic relaxation times increase linearly in fat during heating [10, 11]. A 2-cm-thick cross section of bovine femur (cut from the diaphysis region) was coupled to the acoustic membrane of the Sonalleve system with degassed water. Calibration of T2-based thermal maps involved heating the marrow in a bovine femur and simultaneously measuring T2 with MRI, and absolute temperature with T-type thermocouples was placed in the bone marrow, cortical bone, and the surrounding soft tissue. The femur received a continuous ultrasound exposure for 60 s with an acoustic power of 50 W. This relatively low power was selected to induce a mild temperature increase of approximately 5°C in the marrow while avoiding irreversible T2 changes to enable correction of T2 and temperature during tissue cooling. Dynamic T2 maps were calculated and displayed in real time during both HIFU exposure and tissue cooling. MR imaging parameters were as follows: FOV = 250 mm, pixel size = 1.5 mm, slice thickness = 5 mm, TE₁ = 40 ms, TE₂ = 180 ms, TR = 2.4 s, NEX = 1, and dynamic time = 9.7 s. Figure 3 depicts the user interface for dynamic T2 mapping experiments. Results showed a positive T2 temperature dependence of 20 ms/°C in the bone marrow during the HIFU exposure. The cooling phase showed a temperature dependence of 21 ms/°C, indicating that the measured temperature elevation did not cause irreversible changes in T2 relaxation.

Hydrophone measurements with MatHIFU

MatHIFU was used to automate pulsed ultrasound exposures and to perform acoustic pressure measurements. The Sonalleve clinical MR-HIFU system was controlled with MatHIFU to apply ultrasound exposures on a per-element basis and to generate changes in acoustic pressure within the focal region. The pressure was measured using a fiber-optic acoustic hydrophone system (Precision Acoustics, Dorchester, UK). In this application, each element of the transducer was driven one after another using a pulsed driving signal with 40 cycles at 1.2 MHz and with a repetition rate of 40 Hz. An average of 64 signals was calculated using an oscilloscope (MDO4054-3, Tektronix, Beaverton, OR, USA). Using MatHIFU, a protocol was created to automate the ultrasound exposures with the following steps:

1. 1.

   Turn off all elements.

    
2. 2.

   Turn on only element n.

    
3. 3.

   Turn on pulsed ultrasound exposure.

    
4. 4.

   Turn off ultrasound after 1.8 s of exposure.

    
5. 5.

   Remain inactive until all data collected by the oscilloscope are transferred.

    

Prior to the protocol execution, the oscilloscope was prepared to capture 64 samples of the hydrophone signal over time. Querying the status of the MR-HIFU system was used to determine when the ultrasound exposure was finished, after which the average reading of the 64 acquisitions was transferred through an Ethernet link to the external computer running MatHIFU. Post-transfer, a modification to change the active transducer element was executed and the protocol was restarted. The pseudo-code for the whole acquisition, execution, and modification step of the treatment protocol was as follows:

Do For All Elements In Transducer Prepare Oscilloscope If First Element of Transducer Execute Protocol Else Execute Modification Go to Step 1 of Protocol Wait Until Status of MR-HIFU Indicates End of Ultrasound Exposure Collect Data from Oscilloscope Do Next Transducer Element

Figure 4 shows a screenshot of the graphical user interface that was used to perform and control the ultrasound exposures and hydrophone measurements. The time required to perform an acquisition was 2.3 s per channel and included the overhead for protocol modification, waiting time (1.8 s) to collect 64 acquisitions over time, as well as time to transfer the data. The total time required to capture the measurement data for all 256 transducer elements was 589 s (just under 10 min).

The ‘HIFU’ application using MatMRI in conjunction with MatHIFU

An application was developed to test and showcase the combined capabilities of MatMRI and MatHIFU. This application was used to produce a heating pattern that would form the letters in the word ‘HIFU’ using straight lines, as shown in Figure 5. A total of ten straight segments were used in the pattern. The heating was performed using a polymer-based tissue-mimicking heating phantom (Philips Healthcare, Vantaa, Finland). Using MatHIFU, a treatment protocol with the following steps was created:

1. 1.

   Move the transducer to the center of the letter.

    
2. 2.

   Pause for 2 s. This pause is used to acquire a reference image for thermometry.

    
3. 3.

   Apply HIFU with an acoustic power of 80 W and electronically steer the focal point in 2-mm steps along the 18-mm-long line. The time between steps (trajectory interval) was set to 25 ms.

    
4. 4.

   Turn off the ultrasound after 8 s of exposure.

    
5. 5.

   Remain inactive. This allows protocol modifications to be performed.

    

The coordinates for the transducer position and focal point steering were initially set to produce the first line segment (Figure 5). A cooling time of 60 s was observed after the ultrasound was turned off, and then a protocol modification was created and executed to change the transducer position as well as the steering coordinates. MatMRI was used to collect MR magnitude and phase images that were utilized to calculate temperature change using the PRFS technique. Magnetic drift was monitored in non-heated regions, and a correction was applied [1]. The MR parameters for the thermometry were as follows: FOV = 200 mm, pixel size = 1.1 mm, slice thickness = 7 mm, TE/TR = 20/30 ms, flip angle = 19.5°, acquisition matrix = 180, reconstruction matrix = 224, ETL = 9, NEX = 1, and dynamic time = 0.59 s. Reference images for thermometry were acquired at the beginning of each letter. The protocol duration was 637 s, during which 1,080 images were acquired. The pseudo-code for the image acquisition as well as for the protocol execution and modification steps was:

Do Until All Segments Are Finished If first segment Execute Protocol Else Execute Modification If New Letter Go to Step 1 of Protocol Else Go to Step 3 of Protocol If Transducer Movement Just Finished Collect Magnitude and Phase Image For Reference Else Collect Image Calculate Temperature Change If End of Cooling Do next segment

Figure 6 shows the maximum temperature change projection over the protocol execution. The heating pattern reconstructs the word ‘HIFU’ as expected. The peak maximum temperature was 10.4°C over the baseline.

Discussion

MatMRI and MatHIFU are software tools to perform and facilitate pre-clinical studies. These tools leverage research centers’ multi-million dollar investments in clinical MRI and MR-HIFU hardware, and their greatest strength resides in integrating real-time MRI measurements and controlling the MR-HIFU system with minimal effort. Both MatMRI and MatHIFU toolboxes operate within the MATLAB^®; computational software environment, which is well known in many research laboratories. In the present study, four projects were presented to showcase the capabilities of these software tools to simplify and accelerate the development of new MR-HIFU applications. MatMRI was seamlessly integrated with an existing MR-HIFU pre-clinical small animal system. In addition, the development of a new framework to perform T2-based thermometry was facilitated through the use of MatMRI. Furthermore, MatHIFU was used to activate and precisely control the transducer element driving signals of a clinical MR-HIFU system to enable acoustic pressure measurements using a hydrophone. Finally, MatMRI and MatHIFU were combined in a demo application to highlight the capabilities of these tools in pre-clinical thermal therapy applications.

MatMRI substantially simplifies the acquisition and processing of real-time, dynamic MR data, and it may have potential beyond thermal therapy applications. As indicated by the presented examples, applications requiring real-time processing of MR data, such as MR thermometry or functional MRI, may benefit from this software tool. On the other hand, MatHIFU facilitates the exploration of new therapeutic applications using the Philips Sonalleve clinical MR-HIFU system. MatHIFU opens up new opportunities for rapidly adapting this system for application-specific needs in pre-clinical research. MR-HIFU thermal ablation, HIFU-mediated mild hyperthermia and drug delivery, as well as applications utilizing HIFU-induced mechanical effects may benefit from the use of these tools in pre-clinical work.

Both MatMRI and MatHIFU are intended to be freely available as open-source projects to other research groups under the coordination of Philips Healthcare. These tools are meant to facilitate pre-clinical MR-HIFU research and aid collaboration between research teams. The first collaboration under this model initiated by TBRRI took place with the Hospital for Sick Children. The exchange of ideas has been mutually beneficial for both teams, and we aim to expand the community of users to drive the future development of these tools.

MatMRI

MatMRI uses the eXTernal Control (XTC) protocol [12] that was developed to provide low-latency access to MR data as well as to control an MRI scanner in real time. Among other applications, the XTC protocol is used to provide real-time MR data to the Sonalleve clinical MR-HIFU system. A .Net (Microsoft, Redmond, WA, USA) class was written in the C# programming language (Microsoft, Redmond, WA, USA) to encapsulate all the low-level communication protocols. Furthermore, a class in the MATLAB ^®  programming language was written to expose high-level functions to the end user. These functions are utilized to acquire MR data and to control the scanner.

MatMRI communicates with the MRI scanner through a 1-Gbit/s Ethernet link. Multi-threading capabilities were added to maximize real-time processing speed. An example in which MatMRI is used to acquire individual slices is shown in Figure 7. In this example, image acquisition is initiated on the MRI console and remains paused until the external computer sends an indication to continue. The end-user program subscribes to the MR images and resumes scanning. As shown in Figure 8, the scanner console sends the images, and MatMRI collects them in the background. The end-user program queries for any new image and then processes it. The user program may stop scanning as required. This scenario allows the development of applications that utilize MatMRI for dynamic image acquisition and MATLAB ^®  code for real-time image processing. An example of such application is the calculation of temperature change from temperature-sensitive MR images.

MatHIFU

MatHIFU operates as a client-server system (Figure 9) where the server component is executed in a workstation that is part of the standard Sonalleve clinical MR-HIFU system.

The client component is a MATLAB ^®  class that exposes the functions required to prepare custom-made treatment protocols to the user. A treatment protocol is a series of commands, a mini-program in itself, where the parameters, such as ultrasound exposure and transducer positioning, are specified. The client and the server communicate with each other using transmission control protocol (TCP) sockets. JavaScript Object Notation (JSON) is used to encapsulate the communication messages.

The types of commands that can be specified in a treatment protocol include, among others, the position of the transducer, the start and stop of ultrasound delivery, and low-level R&D commands. MatHIFU allows creating, executing, and modifying protocols, each with any number of commands as required by the user. Only one protocol can be executed at a time. It is possible, however, to prepare any number of protocols in advance and execute them one at a time according to application-specific needs. A very important feature of MatHIFU is the modification of a protocol execution in real time. Figure 10 illustrates an example to prepare, execute, and modify a treatment protocol.

Contribution of Philips Healthcare

Under a research agreement between Philips Healthcare and the Thunder Bay Regional Research Institute, Philips Healthcare provided libraries, documentation, and example code for the use of the XTC protocol. Philips Healthcare also provided specialized training and example code for the customization of the control of the Sonalleve system.

Access to toolboxes

As indicated in the ‘Discussion’ subsection, both MatMRI and MatHIFU are aimed to be freely available to researchers in coordination with Philips Healthcare. The tools are hosted in a private repository in Github (http://www.github.com). To gain access to the repository, please contact Samuel Pichardo (spichard@lakeheadu.ca), with copy to Charles Mougenot (charles.mougenot@philips.com) and Ari Partanen (ari.partanen@philips.com). You can also get in contact with Samuel Pichardo through his personal web page (http://www.proteus-mrighifu.net).