11.1.1. Intervention planning

In this pre-interventional phase, the goal is to define the main steps of the intervention that include (i) delimiting the location to be reached and/or the region to be treated via the MTD and critical adjacent tissues to be preserved; (ii) the choice of the access path: percutaneous, vascular or non-invasive; and (iii) the choice of the MTD and its interventional mode. For this purpose, a map of the subject’s anatomy (ideally in 3D) that covers the entire region penetrated by the MTD is usually obtained before the treatment. This map can be an anatomical image, a 3D model (e.g. a 3D mesh) or a functional representation (e.g. a 3D map of the electrical cardiac activity). Such pieces of information can be acquired several days before the intervention or at the very beginning of the interventional session if the imaging device makes it possible.

11.1.2. Pre-operatory imaging

This phase requires real-time imaging to interactively guide the intervention by combining live images with data obtained during the planning phase. This sets a minimum image acquisition rate for a smooth and, ideally, latency-free display of relevant information to the person performing the intervention. At first, the MTD is typically moved (ballistically) toward the area to be treated, and then the intervention is performed. Ideally, these steps are guided via real-time imaging. However, if this is impossible (e.g. if the device cannot be operated during image acquisition), intermittent imaging is performed, for example, by iteratively re-positioning the therapeutic device at the target area.

11.1.3. Post-operative follow-up imaging

Once the therapeutic intervention has been performed, one or several imaging modalities can be used to evaluate the treatment effect. Such post-operative imaging can be performed immediately after the intervention and/or several weeks later, if necessary.

Table 11.1 lists the different imaging modalities for each intervention step.

MRI presents undeniable advantages for several procedural steps. However, the issues of costs and instrumentation accessibility have so far hindered the clinical development of interventional MRI in favor of other imaging modalities such as ultrasound and X-ray scans or fluoroscopy. On the other hand, a certain number of technical barriers related to the acquisition and image reconstruction techniques, hardware of medical devices that are used in highly magnetic environments, and numerical computation capabilities have also challenged the growth of this discipline. Nevertheless, all these obstacles are getting tackled, paving the way for interventional MRI applications with significant potential.

Figure 11.1 provides a schematic representation of an MTD guided via MRI.

Before continuing, the notion of real-time imaging has to be defined.

It would not make any sense to provide absolute values of imaging rate in seconds or minutes. In fact, the rate strongly depends on the application and type of information that has to be obtained. The notion of real-time imaging has to be put in perspective with the characteristic time of the biological phenomenon that has to be investigated, the specific constraints related to the foreseen treatment and the targeted organ. To illustrate this point, let us consider the example of thermal radiofrequency therapies that are commonly used to treat liver tumors (via percutaneous access) and cardiac arrhythmias (via vascular access). Although the MTD technology is similar in the two applications, the constraints are very different. The treatment duration is in the order of 10–15 min for the liver and about l min for the atrium. Hence, while it is acceptable to update the images every 10 s for the liver, such a rate is not sufficient in the heart. On the other hand, the dimensions of thermal lesions induced in the liver are in the order of several centimeters and a few millimeters in the heart. As a result, the spatial image resolution has to be much higher for the imaging of cardiac arrhythmias than for liver tumor treatments. Furthermore, motion (only respiratory versus respiratory and cardiac) that occurs in these two organs also leads to different constraints regarding the MRI acquisition sequences.

Therefore, the concept of real-time imaging has to take into account all the hardware and physiological constraints to provide relevant information to the operator.

11.2.1. Choice of the MRI acquisition sequence

Interventional MRI involves more significant technical constraints than diagnostic imaging because useful information has to be immediately available during the therapeutic intervention. For this purpose, the same image acquisition has to be repeated to analyze and visualize the evolution of contrasts over time in the reconstructed images. Several fast MRI sequences exist, and their choice depends on the requirements in terms of spatial resolution, volume coverage, acquisition rate and radiofrequency energy deposited during the scan in the sample or around an implanted medical device (risk of tissue heating; see the SAR notion in section 2.6.1). A trade-off among acquisition parameters thus has to be met in order to guarantee sufficient signal-to-noise ratio (SNR) while providing suitable contrasts for visualizing the relevant phenomenon.

For all such reasons, fast gradient echo sequences, such as FLASH (Markl and Leupold 2012) or True-FISP (Bieri and Scheffler 2013), represent the preferred choice. If the scans have to be accelerated, some variations can be employed, such as echo planar imaging (EPI) (Mansfield 1984) as a multishot EPI (ms-EPI) or single-shot EPI (ss-EPI). These techniques acquire several lines (ms-EPI) or all the lines (ss-EPI) of the k-space following an RF pulse. Therefore, the acquisition time is shorter compared to FLASH, since the number of pulses is reduced to record all the k-space lines, with the highest acceleration factor provided by ss-EPI where all the lines are acquired after a single pulse. However, this choice entails longer echo times, which can decrease the SNR in organs with short T₂*. On the other hand, artifacts can occur, including in particular geometric distortions and magnetic susceptibility artifacts around medical devices positioned inside the organs during the treatment.

11.2.2. Image reconstruction

An additional intrinsic constraint in interventional MRI is the reconstruction of images at a rate at least as fast as the acquisition rate. Indeed, if the image reconstruction process leads to delays with respect to the scans, there will be lags between the acquisition and display of useful information for the personnel performing the intervention. Today’s computational performance is capable of providing reconstructed images in real time at rates of at least 10 images/s, which is sufficient for many applications. Among the usually available algorithms on the clinical scanners, parallel imaging techniques (e.g. SENSE, GRAPPA and ASSET) can reduce the number of acquired lines (by a factor of 2 or 3) in the k-space and reconstruct good-quality images in real time by means of simultaneous measurements of an undersampled k-space using multi-element receive coils positioned around the patient.

This acceleration can be exploited in several ways, for example, to increase the number of slices for a fixed scan time, reduce the number of lines acquired at each TR, reduce the echo time (decreasing the T₂* weighting) or decrease possible susceptibility artifacts or geometric distortions. Another approach consists of simultaneously exciting several slices (simultaneous multi-slices techniques) to gather a composite signal resulting from a mixture of all the slices, and then extracting each slice via a reconstruction algorithm that uses the spatial sensitivities of the different receive coils. This approach can be combined with a FLASH, ms-EPI, or ss-EPI sequence, among others, or include k-space undersampling to benefit from the advantages of parallel imaging and fast sequences.

Nevertheless, it should be reminded that the final quality of the images and the eventual artifacts induced by the combination of these acceleration and parallel imaging techniques have to be taken into account.

11.2.3. Image analysis and display

Once the images are reconstructed, they should be processed to extract useful information and deliver it to the operator or MTD to guide the treatment. This typically requires a real-time transfer of reconstructed MRI images to a separate console that performs this operation via specifically developed software for the target application. The provided images can be visualized inside or outside the MRI room depending on the application, for example, when the operator performs the intervention by means of interactive imaging (e.g. insertion of a device into the tumor). The image analysis software should not introduce any latency, especially if high amounts of data are generated at each second.

11.2.4. Motion management

Physiological motion, in particular cardiac and respiratory movements, is quite constraining for MRI. Typically, synchronizing the acquisition of a part of k-space with the respiratory and/or cardiac phase can tackle motion, although at the price of lower temporal resolution, an essential aspect in real-time MRI. On the other hand, using breath holds is difficult to envision. Employing acquisition sequences that are faster than the duration of the considered motion (typically 6 s for the respiratory cycle and 1 s for the cardiac cycle) is thus a viable solution (see section 11.2.1). In this way, the displacement during the acquisition of a given k-space is minimized, although it may be necessary to correct the displacement between subsequent scans for the same k-space plane. The most straightforward way is to use some physiological synchronization (respiratory/cardiac sensor, navigator echo). Another approach consists of correcting the motion that is directly visible on the images of the same slice over time by means of image co-registration. For this purpose, an estimated vector field representing the voxel displacement between an image obtained during the treatment and a reference image has to be calculated. This displacement is then applied to the present image to make it correspond as much as possible to the reference image. The process is then repeated for each newly acquired image. Several algorithms based on rigid and/or deformable transformations have been introduced for different target applications. Such functionalities are particularly important when the analysis of the intensity variation in a pixel during treatment has to be displayed. Another approach to remove motion effects consists of using 3D positioning systems integrated with the MTD (when these are available) and synchronizing the image acquisition with the images of the position of such a system. For example, several existing technologies exploit electromagnetic triangulation systems or MRI measurements from projections along three orthogonal directions by means of small MRI detectors integrated into the MTD.

Knowing the instantaneous position in 3D of the organ or medical device also enables the automatic repositioning of a slice at each scan. This dynamic monitoring makes it possible to choose the image refresh rate without necessarily depending on the characteristic time of the physiological motion. However, this requires greater interaction with the MR system by implementing a dynamic feedback loop that analyzes this position in real time to instantaneously reposition the MRI slices over the target region.

11.3.1. Intracorporeal medical devices

11.3.2. Extracorporeal therapeutic medical devices

HIFU and magnetic hyperthermia devices are non-invasive and operate inside the MRI bore. Their considerations regarding access, MRI compatibility and electromagnetic interference risks are analogous to percutaneous MTDs. However, the risks of artifacts in the target region are almost non-existent, which offers more possibilities in the choice of acquisition sequences. The combination of MRI with a radiotherapy device (MR-Linac) is a relatively recent technology. It has the advantage of an MTD located outside the MRI bore, with a modification of the magnet to let the radiotherapy beam pass through it. For the three MTDs mentioned above, one of the challenges is to manage the relative displacement of a moving target (respiratory or cardiac activity) with respect to the local area where the energy is delivered by the MTD. A dynamic feedback system must be devised to control the energy deposition in real time such that this energy is always oriented toward the target and avoids damaging adjacent healthy tissues.

11.5.1. Principle of MRI thermometry

Several physical parameters affecting the contrast of MRI images are sensitive to temperature (Quesson et al. 2000). They include longitudinal (T₁) and transverse (T₂) relaxation times, magnetization (M₀) and the resonant frequency of water protons (ν₀). Based on the latter, the proton resonance frequency shift (PRFS) technique has become the standard for acquiring images of temperature distribution inside the organs whenever a thermotherapy MTD (radiofrequency, microwaves, laser, HIFU) is used. The variation of ν₀ is almost linear over the range of temperatures in liquid water and depends very little on the investigated organ. The variation of ν₀ with temperature (the PRF constant called σ in the following) is about 0.01 ppm/°C, which corresponds to a variation of 0.63 Hz/°C for a scanner operating at 1.5 T (ν₀ = 63 MHz). Thus, the phase φ of the MRI signal varies as a function of temperature according to the following equation:

Related posts:

Vascular Imaging: Flow and Perfusion Functional MRI Ultra-high Field Imaging MRI Principles, Hardware Components and Quantification Imaging of Dipolar Interactions in Biological Tissues: ihMT and UTE Fast Imaging and Acceleration Techniques

Stay updated, free articles. Join our Telegram channel

Join