Tumor location and reduction in functional MRI estimates of language laterality

Abstract

OBJECTIVE Brain tumors located close to the language cortex may distort functional MRI (fMRI)–based estimates of language dominance. The nature of this distortion, and whether this is an artifact of numerous confounders, remains unknown. The authors hypothesized tumor bias based on laterality estimates independent of confounders and that the effects are the greatest for tumors proximal to Broca's area. METHODS To answer this question, the authors reviewed more than 1113 patients who underwent preoperative fMRI to match samples on 11 known confounders (tumor location, size, type, and grade; seizure history; prior neurosurgery; aphasia presence and severity; and patient age, sex, and handedness). The samples included 30 patients with left hemisphere tumors (15 anterior and 15 posterior) and 30 with right hemisphere tumors (15 anterior and 15 posterior), thus totaling 60 patients (25 women; 18 left-handed and 4 ambidextrous; mean age 47 [SD 14.1] years). Importantly, the authors matched not only patients with left and right hemisphere tumors but also those with anterior and posterior tumors. Standard fMRI laterality indices (LIs) were calculated using whole-brain and region of interest (ROI) approaches (Broca's and Wernicke's areas). RESULTS Tumors close to Broca's area in the left hemisphere decreased LIs independently of known confounders. At the whole-brain level, this appeared to reflect a decrease in LI values in patients with left anterior tumors compared with patients with right anterior tumors. ROI analysis replicated these findings. Broca's area LIs were significantly lower (p = 0.02) in patients with left anterior tumors (mean LI 0.28) when compared with patients with right anterior tumors (mean LI 0.70). Changes in Wernicke's area–based LIs did not differ as a function of the tumor hemisphere. Therefore, in patients with left anterior tumors, it is essential to assess language laterality using left posterior ROIs. In all remaining tumor groups (left posterior tumors and right hemisphere tumors), language laterality derived from the anterior language ROI was the most robust measure of language dominance. CONCLUSIONS Patients with tumors close to Broca's area showed more bilateral fMRI language maps independent of known confounders. The authors caution against the assumption that this reduced language laterality suggests no or little risk to language function following tumor resection in the left inferior frontal gyrus. Their results address how to interpret fMRI data for neurosurgical purposes, along with theoretical questions of contralesional functional compensation and disinhibition.

Organizing Variables Affecting fMRI Estimates of Language Dominance in Patients with Brain Tumors

“…Conversely, tumors affecting Wernicke's area did not significantly distort fMRI language dominance. The fMRI values of language dominance in these patients were similar to those observed in patients with tumors in the right hemisphere [25]. (c) Tumor grade-patients with low-grade tumors are more likely to show functional reorganization because their tumor progression is typically more extended.…”

“…However, the results of the studies are difficult to compare due to differences in the methodological design, including, for instance, not matching samples based on confounding factors that may have biased the results. A recent study [25] addressed these challenges. Sixty patients were matched on a case-by-case basis to control for 11 confounding and modulating factors known to affect fMRI estimates of language dominance (e.g., tumor grade, patient demographics).…”

“…By obscuring actual language activation, previous neurosurgery is assumed to be a significant confounding factor in clinical language mapping [31]. A recent study [25] compared patients with a history of prior surgery who were diagnosed with recurrent tumors and patients with tumors without prior surgery. The patients all had their tumors around the left inferior frontal gyrus of the language-dominant left hemisphere (Brodmann areas 44/45/47), as determined clinically.…”

“…Thus, previous surgery needs to be considered in patients who underwent prior resections and need to undergo preoperative language fMRI because they anticipate another surgery [25,51]. In these individuals, using an ROI approach within unaffected areas of the language network is recommended [25]. Such patients constitute a significant portion of all neurosurgical patients with brain tumors.…”

“…As will be presented below, different approaches to an fMRI analysis can significantly alter the LI values in a single subject. The approaches include using threshold-dependent versus threshold-independent calculations [22,23], applying a full-hemisphere versus an region of interest (ROI) approach [24,25], and administering a single task versus a panel of language tasks [11].…”