Functional magnetic resonance imaging of the brain: A quick review
Correspondence Address: Source of Support: None, Conflict of Interest: None DOI: 10.4103/0028-3886.73735
Source of Support: None, Conflict of Interest: None
Ability to non-invasively map the hemodynamic changes occurring focally in areas of brain involved in various motor, sensory and cognitive functions by functional magnetic resonance imaging (fMRI) has revolutionized research in neuroscience in the last two decades. This technique has already gained clinical use especially in pre-surgical evaluation of epilepsy and neurosurgical planning of resection of mass lesions adjacent to eloquent cortex. In this review we attempt to illustrate basic principles and techniques of fMRI, its applications, practical points to consider while performing and evaluating clinical fMRI and its limitations.
Keywords: BOLD-MRI, brain function, fMRI, MRI
Functional magnetic resonance imaging (fMRI) is a method which detects physiological changes occurring in the brain in response to any given task by using blood-oxygen-level-dependent (BOLD) changes. Major application of this method is to localize eloquent regions of brain and their relation with surgically removable lesions. Though other methods like Wada test or direct cortical stimulation are available for such purpose, fMRI offers unique advantages of non-invasiveness, repeatability and lack of radiation. Although at this stage it may not be a standalone investigation for all the applications described below, it has tremendous scope for being so in the near future and it is definitely complementary to other more invasive methods in many areas of neuroscience research.
BOLD MR phenomenon was first observed by Ogawa and colleagues, who suggested that change in blood oxygenation has a measurable effect with MR signal and can be used to monitor brain activation. Generation of signal in fMRI is due to local reduction of deoxy-Hb as a result of neurovascular coupling and later uncoupling, know as the BOLD effect. The sequence of events occurring at the molecular level, which ultimately leads to signal generation is illustrated in [Figure 1]. During any particular task (e.g. finger tapping), there is local increase in neuronal activity and consequent oxygen consumption (i.e. cerebral metabolic rate of oxygen, CMRO 2 ) which increases deoxyhemoglobin. These along with other metabolic changes in turn send local signal, which results in increased cerebral blood flow (CBF) and cerebral blood volume (CBV) due to vasodilatation. Actually, this increase in oxygen delivery (CBF) overcompensates the local O 2 need (CMRO 2 ) and results in an excess of oxygenated blood at the local site. MRI can measure this physiological change as oxyhemoglobin is diamagnetic and deoxyhemoglobin is paramagnetic and it induces local field inhomogeneities.  It has been known since 1936 that magnetic state of hemoglobin (Hb) changes with its oxygenation state.  Deoxyhemoglobin induces microscopic field gradients that make the field more inhomogeneous and results in signal decay. Oxygenated blood makes the local magnetic environment more uniform. As the field is more uniform, the signal decays less rapidly in this region and results in small increase in local signal. This small percentage increase in signal is recorded on fMRI. So it must be clear now that what we are measuring in fMRI is not the neuronal activity but the hemodynamic response to such activity and hence the signal in fMRI is delayed by 1-2 s and reaches its peak at 5-10 s. 
The basic requirement to get fMRI signal is to identify activated voxels while patient performs the task. For this purpose, fast imaging sequence like echo planar imaging (EPI) is used, which allows collection of data from entire brain within few seconds at the cost of low spatial resolution. fMRI setup requires availability of audiovisual system and various tailor-made paradigms, which are used to guide the patient through the various stages of control and activity conditions. The experiment begins with explaining the patient regarding the task to earn patient co-operation which is vital for a proper result. Within the scanner, various paradigms are presented to patient (e.g. finger movement for motor examination, tactile stimulus for somatosensory and visual/auditory commands for language examination). Images are obtained while patient performs the task, which is divided into control and active state. This results in generation of large amount of data that are analyzed statistically to identify the area of brain activity corresponding to the task performed. This analysis could be done by real-time fMRI processing or by offline post-processing tools. Finally, data regarding the activated brain areas are superimposed on high-resolution images like 3D SPGR/ FLASH or 3D-FLAIR.
A wide variety of clinical and research applications of fMRI has immerged in recent years. Out of these a few relevant and practically implementable clinical uses are described in this section; more sophisticated and research oriented uses are mentioned later.
fMRI is now an integral part of evaluation of patients with intractable epilepsy. The main applications of the fMRI in epilepsy include preoperative mapping of eloquent brain areas (motor, language, memory), predicting the memory outcome in temporal lobe surgery and localization of seizure foci. According to the western statistics, more than 30% of patients with medically intractable epilepsy require surgery. , Significant impact of fMRI has been found on diagnostic work-up and treatment planning in patients with seizure disorders, who are candidates for surgical treatment. 
Pre-surgical localization of eloquent cortex
Success of brain surgery depends not only on proper localization of epileptic focus but also on presurgical identification of brain areas that are close to the epileptic focus and have likelihood of getting damaged during surgery. Another reason for presurgical mapping of brain areas is that the brain function may be reorganized to other areas other than primary areas probably as a compensatory/complimentary mechanism and their localization become necessary to avoid inadvertent surgical damage.,
Under this heading, we will discuss three basic fMRI applications: mapping of motor and somatosensory cortex, language lateralization and mapping of memory system. Traditional methods available for this localization are Wada test, intraoperative cortical stimulation in awake-patients and intraoperative recording of sensory-evoked potentials. All of these methods are invasive and carries risk of major or minor complications. Moreover these are time and technique intensive methods as compared to fMRI, which is being non-invasive and repeatable and thus scores over the other methods at least as an initial mode of investigation.
Pre-operative mapping of motor cortex
This is relatively easy to perform as compared to language and memory functions and so we will begin with this. When a structural lesion (e.g. focal cortical dysplasia (FCD), tumor, heterotropic grey matter etc) is located close to the central sulcus, its preop localization is necessary so that during surgical resection that area can be spared and postoperative deficit can be avoided or minimized. Many authors have studied ability of fMRI to accurately localize sensorimotor cortex. ,, Pattern of activation of brain area depend on part of body that is moving and movement parameters that are selected. The magnitude of activation depends on the rate of the movement.  Common motor stimuli used are finger tapping, opposition of finger to the thumb or repetitive opening and closing of fist. Gross movement (face and limb) should be avoided as it may lead to patient movement and fMRI artifacts. To minimize the movement artifact, it is imperative to instruct the patient regarding the procedure and if possible to fix the head and proximal limb. 
Mapping of somatosensory cortex
This can be achieved by application of some sensory stimuli to the patient's limb in the form of touching with cotton ball, or proprioception by joint movement. ,, Corresponding area of activation in the brain include contralateral post-central gyrus, parietal operculum and peri-insular areas. The extent of response in these areas depends on factors similar to described in motor area mapping. Once the mapping of the sensorimotor area is done its relation with primary pathology (tumor, FCD etc.) is determined by superimposing it a on a high-resolution 3D images, which will help in decision-making such as operative versus conservative management, extent of possible resection and the probability of post-op deficits.
Intracarotid amobarbital injection (Wada test) which is considered as a gold standard for preoperative language and memory lateralization has many limitation such as invasiveness, vascular anomalies and variations interfering with the results, over or under anesthetization and inability to give intracortical localization. Some of these limitations can be there in direct cortical stimulation also. fMRI now provides similar information avoiding all these complexities. Many authors have found high degree of agreement between these methods and fMRI. ,, However, it should be remembered that, fMRI is an 'activation' study and Wada is a 'lesional' study, and they may have inherent differences between them.
As compared to mapping of motor and sensory areas, language mapping is technically demanding and less accurate. Several language paradigms are available for language mapping. Purpose of such tasks is to assess lateralization of language (left or right hemispheric dominance) [Figure 2]. Most of the right-handed individual typically show left dominance (typical) with very few showing symmetric or pure right dominance (atypical). , The role of fMRI is to find out any atypical language dominance that is symmetric or right lateralized as incidence of such atypical lateralization is higher in patients with epilepsy.  Patients with early age of onset of seizure have been shown to have shifting of the language towards right hemisphere.  Sabsevitz et al. in his study of 24 patients showed that accuracy of fMRI in predicting postoperative naming decline is higher than Wada test. However, accuracy of fMRI in language lateralization is still not totally established due to variability in paradigm selection, patient population and analysis. No uniform method of language mapping still exists. 
Mapping of memory system
This is one of the most challenging aspect of clinical fMRI study. The memory is a complex process arbitrarily divided into encoding, consolidation and retrieval. The brain areas, in particular the medial temporal structures are involved in all these steps variably and exact contribution in each process is yet to be discovered. That is why it is difficult to confidently say that the activation occurring in a particular brain area is secondary to memory stimulus or as part of more generalized cognitive process.
Goal of fMRI by mapping memory is to predict post-op memory decline and to localize the seizure focus. Many investigators have studied the ability of fMRI in lateralization of memory and its validity with Wada test. , fMRI can predict the post-op memory decline by showing more activation ipsilateral to the side of seizure.  One study has shown that fMRI can also show reorganization of mesial temporal lesion (MTL) activity to right MTL in patient with left temporal lobe epilepsy (TLE).  Recently, Dupont et al. have shown in a retrospective study that accurate prediction of postoperative memory deficit may be achieved by combining left functional MR imaging activation during delayed recognition, side of the epileptic focus, and preoperative verbal memory score. 
Spontaneous ictal activity localization
Considerable proportion of epileptic patients may not show structural brain abnormality. Secondly, those who show some epileptogenic abnormality may not have lesion within the abnormal area and EEG lacks spatial resolution to exactly localize such lesions. It has been shown that there is evidence of increased perfusion during the period of abnormal interictal electrical discharges, which can be localized by the fMRI and the corresponding seizure activity can be recorded by EEG.  This has inspired the wisdom of combining fMRI with high spatial resolution and EEG with higher temporal resolution. Another application is to guide placement of intracranial electrodes in patients with multifocal epileptic spikes on EEG.
However the technique is expensive and labor intensive and hence the use of such sophisticated methods should be done only when epilepsy team agrees that it will provide additional information and should also be tailored to answer specific questions not provided by more conventional methods such as EEG and structural MRI. It is more practical to use concurrent EEG and fMRI in patients with frequent interictal discharges to make the examination productive.
All the above applications of fMRI mentioned in epilepsy are applicable in preoperative brain tumor assessment. fMRI will provide the information regarding location of eloquent cortex, its relation with and distance from the brain tumor and the probability of postoperative deficits [Figure 3]. It is important for the neuroradiologist to tailor the fMRI study according to the location and extent of the brain tumor, which will reduce the overall procedure time. It is also necessary to directly communicate with the operating surgeon regarding all the information the study has provided so as to facilitate the operative decision. The information could be integrated to neuronavigation systems also, but it may be technically more demanding.
A decrease in activation of MTL has been found in the patients with mild AD as compared to healthy volunteers using visual memory encoding. , Patient of AD or people with high risk of AD have shown larger area of activation on fMRI study during memory task as compared to normal adults, which may explain the compensatory mechanism of brain. In a recent prospective study, the author has shown that functional changes in the posteromedial cortices (precuneus and posterior cingulate gyrus) may be a better fMRI marker of memory impairment than those in the hippocampus. Thus fMRI could potentially be used to identify person with high risk for development of AD based on these observations. Recently many resting state fMRI studies have provided considerable insight into the understanding of this elusive disease. ,,
Considerable research is going on in this field. fMRI has provided answers in many areas of dyslexia research, resulting in better understanding of this condition. These applications can be divided as (1) identification of abnormal circuit in dyslexic patients: Pattern of activation on fMRI is different between normal and dyslexic individuals with latter showing dysfunction in posterior reading circuit (e.g. superior temporal gyrus, temporo-parietal gyrus and angular gyrus); , (2) assessment of compensatory mechanisms in a dyslexic person: increased activation of left and right inferior frontal gyrus has been found in dyslexic readers than in normal controls and (3) effect of training: with reading interventions change in brain activation of dyslexic reader has been found, which is similar to normal reader in the form of activation of left occipito-temporal area. ,
The fMRI can be challenging in patients with AVM due to flow abnormality interfering with BOLD signal. Many studies have assessed the reliability and technical feasibility of pre-op fMRI in AVM patients. , It may give information such as possible postop deficit, eloquent areas to be avoided by the neurosurgeon and reorganization of functional area to same or opposite hemisphere. Correlation has been found between postop patient recovery/deficit and extent of reappearance/disappearance of functional activity on fMRI. 
Brain has an important property of plasticity, which aims at maintaining optimal brain function through reorganization of brain areas from damaged or functionally sub-optimal areas to normal areas. Motor reorganization in patients with stroke and brain tumors has been shown by many researchers. , Study of brain plasticity in case of AVM has shown evidence of interhemispheric transfer of language function (verbal fluency related to left frontal lobe) to right frontal lobe.  However, it may still be unclear whether these activations are a compensatory mechanism or just a reorganization of existing alternate mechanisms.
Prefrontal cortex, anterior cingulate cortex (ACC), insula and striatal structures are the major areas for the regulation and modulation of mood in human beings. Increase in ventral striatal and dorsal ACC activity and decrease in prefrontal cortical activity has been shown in manic patients.  In patients with post-traumatic stress disorder while viewing of emotional facial expressions, there is evidence for increased amygdala activity and diminished medial prefrontal cortex responsively. 
fMRI has been used to assess impairment of cognition in patient after trauma. In one study during a working memory task, pattern of activation in frontal, temporal and parietal region was similar in patients and controls with more dispersed and right lateralized activation in the former.  Another author has shown relative decrease in ACC activity in traumatic brain injury patient as compared to controls probably secondary to destruction of neural networks after diffuse axonal injury. 
Real-time applications of fMRI are emerging in clinical and research applications. 
In the past two decades fMRI has stretched its horizon from being a mere research tool to a highly relevant clinical investigation for surgical planning, be it in epilepsy surgery or brain neoplasms. Its role in dyslexia, Alzheimer disease, brain AVM, psychological disorder and assessment of brain plasticity has been recognized and increasing number of new applications are emerging every day.
[Figure 1], [Figure 2], [Figure 3]