Correlation of Preoperative Functional Magnetic Resonance Imaging (fMRI) with Intraoperative Cortical Stimulation in Surgeries of Eloquent Brain Lesions
Correspondence Address: Source of Support: None, Conflict of Interest: None DOI: 10.4103/0028-3886.355146
Source of Support: None, Conflict of Interest: None
Keywords: Direct Cortical Stimulation (DCS), eloquent cortex, functional MRI (fMRI)
Motor movements and language are considered as one of the basic functions of human being that helps in their survival. Various studies show that different cortical areas are crucial for different functions and that some regions of cortex are indispensable for a defined cortical function. These areas are referred as Eloquent cortex.
In pre-surgical planning, the determination of eloquent afferent and efferent tracts is important to avoid any damage to normal functions. Due to complex location of eloquent areas and the ability of functional areas to change their usual location because of the lesion, it is difficult to localize eloquent areas based only on anatomic landmarks. The classical concept of a constant localization has been proven wrong by the phenomenon called “natural plasticity,” that is, the ability of brain leading to redistribution of the functional maps within a patient.,,,,,,
Direct Cortical Stimulation [DCS] is usually used for localizing the eloquent cortex during the surgery but it has some disadvantages of invasiveness, complexity, seizures and extended operation time. Functional Magnetic Resonance Imaging [fMRI] is a newer imaging technique that is noninvasive and is also relatively easily available for the eloquent brain surgeries. Although diffusion tensor imaging is most helpful to illustrate structural connectivity, fMRI is used to depict functional connectivity.
In this study, we evaluate the reliability and accuracy of fMRI as compared to DCS along with the processing and reporting of the data related to the neurosurgical cases from our institute. We have also given focus on the clinical condition of the patient before and after the surgery thus trying to correlate the safety of these investigations for the patient.
The study was a prospective study of diagnostic efficacy to assess the reliability of fMRI by comparing it with DCS and to achieve maximum extent of resection with minimum disruption of patient's neurological status. The identity of the patient was not declared in any part of the project. The study was approved by the ethical committee of the hospital.
This study involved thirty patients with brain lesions within or adjacent to eloquent cortex undergoing surgery at our institute. This study was conducted prospectively for 2 years from April 2016 to March 2018. Inclusion criteria were brain lesions within or adjacent to eloquent areas and age of patient more than 10 years. Exclusion criteria were lesions in non-eloquent brain areas, children below 10 years of age and the patients incompatible for MRI.
Informed written valid consent was taken. Preoperative neurological examinations were performed a day before the surgery to evaluate the existing neurological deficits. Karnofsky score and MRC grading of power were noted preoperatively. Preoperative fMRI and diffusion tensor imaging sequences were obtained to localize functional areas and white matter tracts. Philips Ingenia 3T digital scanner was used along with EPrime with ESys In vivo hardware and software to perform fMRI scans. Imaging data were analyzed offline, loaded into Dynasuite Neuro workstation. All the processed data was then fed into the neuronavigation system and the activated areas were co-registered. A Stealth Station S7 neuro-navigational system was used for localization during the surgeries.
Lateral position was preferred for Awake craniotomy patients and supine position for general anesthesia patients. After the craniotomy and durotomy, eloquent cortex was identified using fMRI data embedded in Neuronavigation system. SSEP was used to confirm the motor-sensory cortex junction with the phase reversal phenomenon. The contralateral upper limb, lower limb and face were visualized for any abnormality in movement. Stimulation mapping starts with identification of the motor cortex. A 4-point contact strip electrode (5 mm spacing, 60 Hz, 100 us phase duration and 10 s train duration) was positioned on the brain surface initially depending on identification of the rolandic fissure/fMRI motor map.
DCS was performed with the current amplitude between 1 and 15 mA using Nicolet/Nihon Kohden Cortical Stimulator incorporated with the ongoing electrocorticography.
The motor strip was stimulated with an initial current of 1 mA and later increased by 1 mA units until a motor response was visually identified. The amplitude of current that identified the motor cortex usually was 5 mA or less in the awake patient.
At the same time, fMRI embedded neuronavigation was used to confirm the location of eloquent cortex. The findings of fMRI and DCS monitoring regarding the recognition of eloquent cortical areas were compared. If the location of fMRI and cortical stimulation areas was same or less than 5 mm apart, these were considered concordant studies. However, if the distance between the fMRI and the DCS defined eloquent areas was more than 5 mm, these were considered discordant studies.
During the procedure, DCS findings were analyzed by the Neurologist team and fMRI findings were analyzed by the radiologist. Both the teams were blind about each other's findings.
The eloquent brain areas were divided into true positive, true negative, false positive, and false negative. Based on it, specificity, sensitivity, positive predictive value, negative predictive value of fMRI as compared to DCS was calculated.
Patient's preoperative and postoperative neurological status were compared using Karnofsky scale and MRC scale. The outcome was measured as Improvement, same or worsening of neurological status.
Data analysis was done with the help of Microsoft Excel and SPSS V.23. Statistical analysis of our data was performed by employing Pearson Chi-Square, Fisher's Exact test, Mann–Whitney U test and the level of significance was set at 5%. The receiver operative characteristic (ROC), sensitivity, specificity, PPV, and NPV for each region were calculated.
In this study, patients were classified into various age groups, with most (50%) being within 40-59 years group followed by 9 patients in 20–39 year group and 5 patients above 60 years age. Only 1 patient (age = 13 year) was below 20 year age. Mean age was 45 years with standard deviation of 14 years. 20 patients (66.7%) were males and 10 patients (33.3%) were females [Table 1]. Most of the patients presented with history of seizures (23/30, 77%). There was overlapping of symptoms in some patients in the form of seizures and motor weakness. Ten patients had motor weakness and four patients had speech difficulty [Table 1].
Most of the lesions were present in frontal lobe, predominantly posterior frontal in the motor cortex region [22/30]. Five lesions were present in parietal lobe, 2 lesions were in temporal lobe and 1 involved the insula. Majority of the lesions were gliomas [19/30]. There were 5 cases of meningioma, 3 cases of AVM, 2 cases of lymphoma and 1 case of metastasis [Table 1]. Local and general anesthesia was used in 21 cases and 9 cases, respectively. General anesthesia was used in cases where lesions were expected to be vascular or expected to have a long operation time like meningioma and AVM. The relationship between the type of anesthesia and concordance was calculated however the results were not significant (χ2 value. 055, P = 0.815) [Table 1].
Concordance of fMRI with DCS was noted in 26 cases (87%) in our study of 30 patients [Table 2]a. Overall mean sensitivity, specificity, positive and negative predictive value of fMRI as compared to DCS was 95%, 92.48%, 85.56%, and 96.08%, respectively [Table 2]b.
Preoperative and Postoperative Karnofsky score stayed same in most of the cases [25/30] [Figure 1]. Seizures were seen during the direct cortical stimulation in 7 cases. Focal seizures were noted in 4 cases and generalized seizures in 3 cases. No case was abandoned because of seizures. Discordance between SSEP and DCS was seen in 6 cases [Table 2]a.
Cortical stimulation is considered the gold standard for mapping eloquent brain cortex. DCS is a real-time mapping technique compared to virtual mapping by fMRI. DCS is an old yet trustful technique, when it comes to eloquent cortex mapping. However, DCS allows only mapping of a local region, and not of the whole brain. Moreover, DCS is invasive, labor-intensive and time-consuming with associated complications such as seizures. Therefore, merger with other non-invasive methods seems desirable.
There have been various studies in the past comparing the functional MRI with DCS with conflicting results. Atlas et al. published that they localized eloquent cortex in 71.5% cases by fMRI and they reasoned it by blaming the glioblastomas and high-grade gliomas for altering BOLD signals. Fandino et al. published 82% fMRI precision in outlining the eloquent cortex. 92.3% fMRI precision was noted by Spena et al. in localizing sensory-motor cortex. Mueller et al., Schulder et al., Jack et al. and Roux et al. found 100% fMRI precision in recognizing the eloquent cortex.,,, However, study populations were limited in these publications.
Lehericy et al. and Li et al. noted 92% and 100% fMRI precision in pinpointing the motor cortex., In one prospective study, Yousry et al. reported that the fMRI precision in recognizing the eloquent cortex was 100% with an error margin within 10 mm. Similarly, Hirsch et al. published a series of 125 patients with brain lesion and of 63 healthy volunteers. There was accuracy of 100% in identifying central sulcus in healthy cases and of 98.4% in patients. FitzGerald et al. published sensitivity of 81% and specificity of 53% for fMRI when the margin error was 10 mm, that changed to 92% and 0%, respectively, with a margin error of 20 mm. Pouratian et al. mentioned in his study that the specificity and sensitivity for frontal lobe were 66.7% and 100%, respectively, whereas 96.2% and 69.8% for the temporal lobe, respectively. Krings et al. compared fMRI and PET scan with DCS and found overlapping results. In 2005, Duffau et al. published a study of low-grade glioma patients, that were divided into two groups––first group operated using DCS and the other group without DCS. The localization of eloquent areas was better in DCS group along with decrement in the percentage of severe permanent deficits to 6.5% in the DCS group as compared to 17% in the non-DCS group.
The concordance of fMRI with DCS was noted in 26 cases (87%) in our study of 30 patients. Discordance was labeled whenever the distance between fMRI activation areas and DCS stimulated areas was more than 5 mm. Overall mean sensitivity of fMRI as compared to DCS was 95%. Overall mean specificity of fMRI as compared to DCS was 92.48%. Overall mean positive predictive value of fMRI as compared to DCS was 85.56%. Overall mean negative predictive value of fMRI as compared to DCS was 96.08% [Table 2]b.
Preoperative and Postoperative Karnofsky score stayed same in most of the cases [25/30]. Improvement in Karnofsky score was noted in 3 cases where motor weakness improved after surgery. The relationship between the preoperative Karnofsky score and concordance was calculated however the results were not significant (Mann Whitney U value 32.00, P = 0.14). The relationship between the postoperative Karnofsky score and concordance was calculated however the results were not significant (Mann–Whitney U value 31.00, P = 0.11). There was worsening of score in one case from 80 to 40. The patient had cortical vein injury during the surgery which can act as a confounding reason for hemiparesis and hence drop in Karnofsky score. One of the cases has been shown in the figures [Figure 2]a and [Figure 2]b.
During surgeries of large tumors, there is a risk of brain shift because of mass effect, CSF leak, surgical retraction, or even the extent of resection that can lead to false mapping by fMRI even after an accurate registration before the surgery. In our series, DCS was carried out just after opening of the dura mater to avoid effects of surgical retraction, CSF leak, or extent of resection.
In spite of taking all possible precautions, 4 cases had discordance between fMRI and DCS. One of the possible reasons can be deformation induced by craniotomy, which reduces image registration and thus decreases the accuracy of fMRI. Also, the paradigms applied for fMRI and DCS cannot be the same because of different setups. Activation areas of fMRI are based on the statistical thresholds that were decided by the radiologists and directly influence the localization of eloquent cortex by fMRI. In our case series, only 1 radiologist was involved to avoid inter-observer error. A low threshold can increase the number of non-essential areas and the size of the critical areas, whereas a high threshold will result in critical areas not reaching statistical significance.
Additionally, mapping principles of fMRI and DCS are basically different. fMRI shows all the brain areas that are involved during the execution of a voluntary task; however, DCS only localizes the area that is essential for the task, by disturbing that particular function against the patient's will. The poverty in complete concordance raises the question about the precision of specificity, sensitivity, and predictive values of fMRI when compared with the DCS. DCS is and will remain the gold standard for mapping eloquent cortex. On the other hand, we cannot ignore the importance of fMRI as an adjuvant to DCS during planning of eloquent cortex surgeries. Although fMRI cannot replace DCS, it can guide and increase the efficacy in resection, select high-risk patients for intraoperative monitoring, help in preoperative stratification of risk counseling and preservation of neurological status in eloquent brain lesions.
Concordance of fMRI with DCS was noted in 26 cases (87%) in our study of 30 patients. Overall mean sensitivity, specificity, positive and negative predictive value of fMRI as compared to DCS was 95%, 92.48%, 85.56%, and 96.08%, respectively. The other advantages of fMRI were non-invasive assessment of eloquent brain regions (both surface and deep) with ease, assessing surgical risk, guiding surgical approaches, and exposure for the resection of eloquent cortical lesions. However, motion artifact, observer dependence, infiltrative tumors and venous effects can lead to inaccuracy of fMRI, and hence DCS, the real-time mapping technique still remains the Gold standard despite its invasiveness, risk of seizures (21 percent in our series), and labor-intensive procedure. Moreover, fMRI can be used to select high-risk patients that would require DCS thus reducing the final number of patients requiring DCS.
Declaration of patient consent
The authors certify that they have obtained all appropriate patient consent forms. In the form, the patient(s) has/have given his/her/their consent for his/her/their images and other clinical information to be reported in the journal. The patients understand that their names and initials will not be published and due efforts will be made to conceal their identity, but anonymity cannot be guaranteed.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
[Figure 1], [Figure 2]
[Table 1], [Table 2]