Development of a Clinical Functional Magnetic Resonance Imaging Service

One of the limitations of anatomy-based imaging approaches is its relative inability to identify whether specific brain functions may be compromised by the location of brain lesions or contemplated brain surgeries. Of the many techniques available to the surgeon, functional magnetic resonance imaging (fMRI) has become the primary modality of choice because of the ability of MRI to serve as a “one-stop shop” for assessing both anatomy and functionality of the brain. This article discusses the specific requirements for establishing an fMRI program, including specific software and hardware requirements. In addition, the nature of the fMRI CPT codes is discussed.

From the discovery of the Roentgen Ray more than a century ago, the primary focus of radiology has been the noninvasive demonstration of structural anatomy of the human body. Through the demonstration of alterations of normal anatomy, the identification of pathologic conditions was elicited. The wide availability of cross-sectional imaging modalities, first with computed tomography (CT) followed by magnetic resonance imaging (MRI), resulted in marked improvements in the quality of the depiction of anatomic structures in 3-dimensional (3D) space. The anatomic region most affected by these techniques was the neuraxis, which had been previously hidden by the calvarium and vertebral column.

As this issue of the Clinics demonstrates, a new focus on physiologic imaging has developed over the past 2 decades. Through these methods, the imaging aim has shifted from the demonstration of anatomy to the evaluation of physiology. Many of these so-called functional imaging approaches are being refined, including the imaging of blood flow (perfusion), specific metabolites (spectroscopy), and regional blood oxygenation (blood oxygen level–dependent, or BOLD imaging). Taken as a whole, each of these methods can be considered as functional imaging, although the aim of this article is to describe approaches involving BOLD imaging.

As more completely described elsewhere in this issue, BOLD imaging is based on a phenomenon first described by Roy and Sherrington in Cambridge in 1890, demonstrating that neural activation induces increased blood flow in the blood vessels supplying and draining that area of cerebral cortex. This phenomenon remained unexploited for nearly a century until the development of positron emission tomography (PET) imaging in the 1980s. The phenomenon was further leveraged with the development of BOLD functional MRI (fMRI) in 1990. The power of the BOLD method was made evident in early studies when regions of brain activation associated with various motor, sensory, visual, and language tasks were mapped in normal subjects. These early demonstrations of feasibility were followed by the development of paradigms used to evaluate pathologic states, including brain tumors, multiple sclerosis, strokes, epilepsy, and Alzheimer disease.

The fMRI technique remained largely a research modality until 2007, when the Centers for Medicare and Medicaid Services (CMS) issued 3 separate current procedural terminology (CPT) codes for its clinical use ( Table 1 ). The recommendation for the development of CPT codes was based largely on experience at major academic medical centers that had begun to use fMRI as a presurgical technique for mapping eloquent regions of the brain, typically under the auspices of a protocol approved by the institutional review board. Using this approach, the typical study consisted of developing fMRI paradigms for the identification of motor areas, speech centers, and sensory regions that should be avoided by the surgeon during a craniotomy. This development of clinical CPT codes has led to significantly increased interest in the use of fMRI by clinicians and consequent increases in the number of studies ordered in many institutions. The aim of this short article is to describe both the rationale for establishing a clinical fMRI service and the basic requirements for hardware and software. In addition, there is a short discussion on the implementation of the CPT codes.

Table 1

Summary of the CPT codes for fMRI

CPT Code	Description
70554	Functional MRI selected and performed by a technologist. Do not report with CPT 96020 or CPT 70555.
70555	Complex paradigms for fMRI selected and performed by a physician or psychologist, involving neurofunctional testing. Report with CPT 96020.
96020	Neurofunctional testing by physician or psychologist, with review of results and report. Report with CPT 70555.

Rationale for establishing an fMRI center in support of brain tumor centers

According to the Brain Tumor Society, there are in excess of 200,000 newly diagnosed brain tumors (primary or metastatic) each year within the United States. Despite significant advances in both chemotherapy and radiation treatments, neurosurgery remains the preeminent treatment modality in most cases. Numerous outcomes of gross-total resection of a brain tumor include relief of mass effect, decreased risk of epilepsy, increased time to tumor progression, and increased survival. However, the preoperative goal of gross-total resection can be thwarted intraoperatively by the difficulty in determining tumor boundaries. This is particularly true for infiltrating gliomas, which may be difficult to distinguish from healthy brain due to indistinct infiltrative margins with healthy brain parenchyma. Moreover, the proximity of tumor to eloquent brain carries the risk of functional loss after surgery, if those regions are violated. By mapping the boundaries of eloquent brain in the context of tumor-distorted anatomy, neurofunctional testing approaches have been used over the past 25 years to improve surgical outcomes.

Functional Imaging

fMRI is a noninvasive MR-based brain mapping technique that requires no exogenous contrast agents. Through the investment in a minimal amount of hardware and software that can be obtained at incremental cost, fMRI can be added to most existing MR scanners. Evaluation of multiple functions is feasible in patients with cerebral tumors using fMRI, and preoperative maps depicting brain areas activated during motor, sensory, and language tasks can be obtained using task paradigms specific to the function or functions of interest. Typical paradigms include motor mapping (eg, hand clenching, finger tapping, toe movement), language (word generation, sentence completion, and similar tasks), and visual stimulation (typically a flashing checkerboard or annulus). The resulting maps are useful for presurgical planning and can be integrated into neuronavigation systems to guide intraoperative decision making ( Fig. 1 ). Strong evidence that a more radical tumor resection may be achieved by using fMRI information during neurosurgery has been demonstrated by Krishnan and colleagues and Haberg and colleagues. Diffusion tensor imaging (DTI)–based tractography has recently emerged as another potentially valuable tool to visualize white matter anatomy for preoperative planning and postoperative follow-up of surgically treated brain tumors and vascular malformations ( Fig. 2 ). In addition, the use of fMRI has been shown to significantly reduce operative time and facilitate preoperative decision making as to whether to perform surgeries awake or under general anesthesia.

Software and hardware requirements

Although most mid-field and high-field MRI scanners in clinical use are capable of performing fMRI, most require upgrades of scanner software and hardware to perform the imaging. In addition, software “keys” for specialized pulse sequences required for performing fMRI on the MRI scanner, such as BOLD or DTI echo planar sequences, may be required. Specialized software is required for both administering the test and interpreting the test results properly. These results are discussed separately later. There is also a specific need for specially trained individuals to administer and interpret the test.

Stimulus Delivery Software and Hardware

The specialized software required for performing the study is known as stimulus delivery software. This software handles several tasks that are not part of standard MRI studies. Specifically, the software must perform 3 tasks: (1) synchronize with the MRI scanner software to initiate the proper pulse sequences to be performed by the scanner at the appropriate time, (2) provide the appropriate stimulus to the subject to activate the brain area of interest at the proper time, and (3) record any associated data, such as feedback from response devices, at the appropriate times. To perform these tasks, there are many solutions offered by both the original equipment manufacturers (OEMs) and third parties.

The ability to synchronize with the native scanner software to perform the appropriate pulse sequence is central to the fMRI experiment. The synchronization is critical because precise knowledge of the task timing relative to image acquisition is required to tease out the small signal differences between task and control conditions. When using OEM stimulus software, it may reside on the scanner console itself. In contrast, when using third-party stimulus software, it typically resides on another computer that is linked to the scanner computer, usually through a specialized hardware synchronization device ( Fig. 3 ).

There are several different stimulus devices that are available that may need to be powered by the stimulus software. In many instances, visual stimuli are provided to the subject, as such, several different kinds of visual display devices may potentially be used. Typically a video projector or special MRI-compatible eye goggles are used to display a visual stimulus at appropriate times to the subject (see Fig. 3 ). More recently, several vendors have developed MRI-compatible high-definition liquid crystal display (HD-LCD) monitors. In fact, the development of the HD-LCD monitors has necessitated computer hardware upgrades because older systems do not have the high-definition video output that is capable of supporting the HD-LCD monitors. Alternatively, auditory stimuli are commonly used and delivered through nonmagnetic headsets or ear buds, and high-quality audio systems are also available for auditory stimulus presentation. The third component of the stimulus delivery software is the ability to record behavioral data and subject response data at the proper time point. The types of data recorded can be from a variety of devices, such as button boxes and eye tracking cameras (see Fig. 3 ).

Analysis Software

The analysis software generates the work product of the entire process. To be useful, it must be able to accept both BOLD datasets and high-resolution anatomic datasets. At present, analysis software typically accepts other types of data, including perfusion, DTI, and angiographic datasets. In almost all instances, the analysis software packages are run on a postprocessing workstation, whether the software has been provided by the OEM or not. The analysis software performs two main tasks: data reduction and data display. Software subroutines that are useful for the analysis of each of these datasets frequently include BOLD data, DTI data, and high-resolution anatomic data. It should be noted that the typical fMRI experiment generates thousands of images and that a separate data transfer protocol from the scanner to the post processing station is usually required. The images then need to be appropriately grouped according to the analytic module and statistically analyzed. Activations are then identified based on a user-defined threshold and displayed as a 3D map that is superimposed on the anatomic images. Typically, it possible to both modify the data in real time and cut planes and rotate projections viewed in the 3D and orthogonal view displays. More advanced systems allow the operator to display activations from multiple task paradigms simultaneously and toggle them on and off ( Fig. 4 ).

For archival purposes, a threshold level is chosen by the user and then orthogonal views are output in digital imaging and communications in medicine format (DICOM) for loading into institutional picture archiving and communication systems (PACS). Surgeons also appreciate the ability to export the processed fMRI data to neuronavigation systems for use intraoperatively, and this function is now frequently offered in commercial packages, although data transfer work arounds used to be necessary. Given the significant effect that these software packages have in clinical decision making, many vendors have obtained approval by US Food and Drug Administration for their clinical software packages.

Scanner Hardware Requirements

The main hardware requirement obviously is a mid-field or high-field MRI scanner (1.5–3.0 T). There is improved signal observed at higher magnetic field strengths, such that fMRI performed at 3 T MRI is of significantly higher quality than at 1.5 T. Echo planar imaging is typically used for fMRI, and as such, the scanner gradient coils must be capable of rapidly switching gradients. A multichannel head coil also provides improved signal acquisition over a single-channel coil. The scanner should be approved for fMRI use before implementing a program, and the pulse sequences should then be assessed according to routine quality control guidelines by the institutional MRI physicist.

Retrofitting of an existing MRI scanner to perform fMRI requires physical modifications to the scanner room. The specific hardware required for fMRI stimulus delivery should be installed. Devices for visual display include goggles, projectors, or MRI-compatible LCD panels; audio devices include headphones or ear buds, as described earlier. Frequently, optical cables are used to transfer the signals from the control room to the scanner room through a device known as a wave guide. Access to the scanning room via a penetration panel may also be necessary for other instrumentation, such as physiologic recording equipment. In addition, appropriate mounts for cameras and LCD panels may need to be installed. Once the system is set up, several test runs with control subjects should be performed to test all components and train personnel.

Personnel and Training

Personnel administering the fMRI examination need specific training to use the fMRI hardware and software, as well as training regarding methodological issues related to the functional tasks. In many cases the examinations are performed by technologists, and often, the equipment/software manufacturer provides on site training. The visual display devices must be set up to demonstrate clear images to the subject that are observed by both eyes and can stimulate all of the visual fields. Therefore, goggles are often adjustable for the interophthalmic distance and requisite visual corrections as needed for each subject. Similarly, projectors or LCD monitors must be positioned such that the subjects can properly see them. When setting up the audio system, care must be taken to ensure that the subjects can adequately hear the presented content independently in each ear while the scanner is running.

In addition, before scanning, the technologist should instruct the patient and assist with practicing the behavioral task paradigms that will be run. This step is crucial to ensure that the tasks are performed correctly, so that the functions of interest can be assessed. Patients who are aphasic or have motor deficits may require additional coaching. Ensuring patient comfort and reducing head motion with appropriate padding are paramount because both factors can affect task performance. Task performance should also be monitored during the scanning session to ensure the task paradigms are followed. After the scan, the data need to be analyzed by the technologist or additional trained personnel.

The American College of Radiology (ACR) in collaboration with the American Society of Neuroradiology published guidelines for performing fMRI studies in 2007. In these guidelines, the physician supervising and interpreting fMRI is tasked with being clinically informed about the patient and understanding the “specific questions to be answered before the procedure to plan and perform it safely and effectively.” The physician should also have experience or formal training in the performance of fMRI.

As with any MRI study, the supervising physician must understand indications, risks, and benefits of the examination, as well as the alternative imaging procedures, such as a Wada test. Risk assessment involves knowledge of patient factors, including the presence of a pacemaker or other medical device that is potentially hazardous in the MR environment. An understanding of the hazards of MRI contrast encompassing both allergies and the potential of nephrogenic systemic fibrosis in the setting of significant renal impairment is critical if contrast administration is being considered.

The physician interpreting the study should be familiar with the patient’s clinical presentation, relevant prior history, and imaging studies. Obviously, the physician performing the fMRI interpretation must also have appropriate knowledge and understanding of the anatomy and pathophysiology to render a meaningful interpretation. Experience with fMRI paradigm design, selection, administration, and validation is also critical to rendering a quality interpretation. Rigorous quality assessment of every study should take into account factors such as patient compliance with the protocol and patient motion. Confounding issues such as magnetic susceptibility artifact at the skullbase or the presence of blood products or metal is essential to performing high-quality interpretation. Also, the interpreting physician must understand the BOLD effect and potential sources of neurovascular uncoupling that could cause false-negative activations.

Only gold members can continue reading. Log In or Register to continue