American Society of Functional Neuroradiology–Recommended fMRI Paradigm Algorithms for Presurgical Language Assessment

Sentence Completion Task

The Sentence Completion language task is one of the many language paradigms that can be used for language localization and hemispheric lateralization for identifying the primary language cortex. SC is a semantic language paradigm that is effective in activating the superior temporal gyrus in the Wernicke area.¹¹ SC can also activate the Broca area in the dominant hemisphere because the task requires both receptive and expressive language processing, though activation of Broca is slightly less robust than in Wernicke.¹¹ Additionally, the pattern of activation with SC is less lateralized to Broca compared with SWG because of involvement of the homologous right hemisphere in speech-comprehension tasks, which invoke executive processing.^7,12 Since verbal comprehension tasks such as SC involve visual processing, activation of the ventral language stream, including the visual word form area, can also be seen.^13,14 SC has been found to produce increased activation in both temporal and frontal regions compared with a word generation task, suggesting more robust activation of language networks because it combines language comprehension as well as production in a naturalistic fashion.¹⁵

The control task for SC uses gibberish sentences devoid of the semantic, syntactic, and lexicalization demands that are present in the active task.

Silent Word Generation Task

Silent Word Generation is a commonly applied clinical language fMRI paradigm for presurgical mapping with extensive comparisons with Wada testing.^{11,14,16⇓⇓–19}

Like most word generation paradigms, SWG tasks activate mainly frontal lobe language and cognitive support areas but are less consistent activators of temporal language regions.^11,14,19,20 Specifically, SWG tasks yield reliable activation of the inferior frontal gyrus, dorsolateral prefrontal cortex, and superior frontal gyrus, with variable activation in the anterior cingulate gyrus, presupplementary motor area, and inferior, middle, and superior temporal gyri.^{14,16,17,21,22}

Several studies have validated the SWG paradigm for both language lateralization and localization.^{21⇓⇓⇓⇓–26} For example, Yetkin et al²³ reported 100% sensitivity of fMRI by using a SWG task for activation within 20 mm of the electrically stimulated cortical site during ECS and 86% sensitivity within 10 mm in a mixed series of 28 patients with predominantly epilepsy (n = 22) and cerebral lesions. Brannen et al²⁴ studied the reliability, precision, and accuracy of word generation tasks in mapping the Broca area in 34 patients with cerebral lesions by comparing regions of activation with awake (ECS) mapping of speech function during craniotomy. They noted that SWG tasks activated Brodmann areas 44 or 46 either individually or both unilaterally or bilaterally in most patients in the series with variable activation of Brodmann areas 9 and 45. Activation was noted in the same gyri when the patient performed a second iteration of the SWG task, and speech areas located with ECS coincided with areas of the brain activated with the SWG task. Comparing hemispheric language dominance by using an SWG task with results of Wada testing in a large series of 100 patients with epilepsy, Woerman et al²⁷ found 91% concordance between both tests. Similarly, Sabbah et al²¹ demonstrated that fMRI language lateralization based on the SWG was concordant with the Wada test in 19 of 20 patients with intractable partial epilepsy.

Pillai's group compared the localization (as locally detectable statistically significant percentage signal change) and lateralization among 5 language paradigms: SWG, Sentence Completion, Visual Antonym Pair, Auditory Antonym Pair, and Noun-Verb Association in 5 ROIs: inferior frontal gyrus, middle frontal gyrus, and superior frontal gyrus for expressive language activation; and middle temporal gyrus and superior temporal gyrus for receptive language activation in a group of 12 healthy volunteers.¹⁴ The results of this study found that SWG was the most robust paradigm for language localization and the most effective for determining language lateralization in the expressive ROIs of the dominant language hemisphere. In contrast, the analysis of patterns of activation in the receptive ROIs in the temporal gyri demonstrated a weaker BOLD percentage signal change for SWG tasks in both hemispheres when compared with the other 4 paradigms that include some form of language comprehension in the active blocks of the paradigm, which are localized in the angular and middle temporal gyri.²⁸

A typical control used for the SWG task employs nonsense symbols without the phonemic fluency-processing demands of the active task.

Rhyming Task

Rhyming is considered a phonologic task that has been shown to robustly activate the Broca area in the dominant hemisphere.¹⁷ Rhyming also activates the Wernicke area though not as strongly as seen in Broca.¹¹ In addition to the above regions, Rhyming has been shown to activate the inferior parietal lobule, the dorsolateral prefrontal cortex, and posterior lateral gyrus.²⁹ Rhyming and SWG are the 2 paradigms found to be most helpful for language lateralization compared with Sentence Completion and Noun-Verb Association.¹¹ As with Sentence Completion, there is activation of the posterior temporo-occipital regions due to involvement of the visual processing pathways.¹³

Studies comparing the relative effectiveness of different commonly used expressive and receptive language paradigms by using both threshold-independent and threshold-dependent methods in patients with brain tumors demonstrated that SWG and Rhyming were more helpful for language lateralization than SC or Noun-Verb Association.^11,19 The Rhyming task has several advantages over the SWG task: It results in more specific activation of language areas than SWG, has a higher mean laterality index (LI) value, and is less threshold-dependent than SWG for determining the LI.^11,17,19 Nevertheless, both SWG and Rhyming demonstrate adequate language lateralization, even in a subgroup of patients with brain tumors located in the left hemisphere and in the frontal or parietal lobes.¹⁹ This has clinical implications since the presence of susceptibility artifacts or neurovascular uncoupling associated with tumors in the dominant hemisphere could affect the strength of the BOLD signal.

The control task most often employs nonsense symbols wherein the patient is asked whether 2 sets of symbols are oriented in the same fashion, which allows monitoring of the patient's attention and task engagement without stimulating phonologic processing like the active task.

Object Naming Task

Simple Object Naming is an expressive language paradigm in which the subject is shown an object and is asked to silently name the presented object.^30⇓–32 The control block consists of a nonsense symbol to exclude the visual component from the activation maps. The objects are presented in black and white.

ON tasks yield activation in the inferior frontal gyrus, middle frontal gyrus, and ventral occipitotemporal cortex, with variable activation in the posterior temporoparietal cortex.^30,31,33,34 Several studies have validated the ON paradigm for both language lateralization and localization.^{8,19,31,34⇓–36} For example, Rutten et al⁸ reported that a combination of 3 different fMRI language tasks (ON, Verb Generation, and Sentence Processing) were able to localize critical language areas with 100% sensitivity and 61% specificity when compared with ECS mapping in a group of patients with epilepsy. Similarly, Hirsch et al³¹ demonstrated 100% concordance of localization between fMRI and ECS when using ON and word-listening tasks in neurosurgical candidates, most of whom had brain tumors,³⁶ and Pouratian et al³⁵ demonstrated 100% sensitivity and 67% specificity for frontal language regions when correlating ECS mapping and fMRI in patients with vascular malformations who performed Object Naming, Word Generation, and auditory-response naming tasks.

Although ON provides relatively good localization predominantly of frontal language regions, the general consensus is that the activation pattern associated with ON does not allow as effective hemispheric language lateralization as with other expressive language paradigms such as SWG, Verb Generation, and Rhyming.^19,34 Moreover, like other covert paradigms, the ON task does not allow monitoring of patient responses. Despite these limitations, the ON task is easy for most patients to perform and is therefore often utilized in patients with cognitive impairment or pediatric patients, and it is frequently used during intraoperative cortical stimulation.

Antonym Generation Task

This word generation paradigm typically presents the patient with 1 word at a time, and the patient is asked to silently think of the word that means the opposite. Typically, during each activity period, 5–10 words are sequentially presented on the screen, and during the rest period, a black blank screen is presented with a white crosshair in the center for visual fixation.

This paradigm stimulates phonologic, working memory, lexical search, semantic, and orthographic processes of the speech and language system. Functional maps typically demonstrate activation in the inferior frontal gyrus, middle frontal gyrus, premotor cortex, and frequently posterior peri-Sylvian speech areas (posterior aspect of the superior temporal lobe, supramarginal and angular gyri). Frequent weak activation is also seen in the posterior aspect of the inferior temporal lobe, possibly due to engagement of the visual word form systems.^{37⇓⇓–40} Because there is no control for visual sensory processing during the resting phase, frequently, bilateral primary visual cortex activation is also seen.

The Antonym Generation task is reported to provide a higher percentage BOLD signal change in the Broca area (versus baseline) compared with the Antonym Decision task (versus baseline).⁴¹ It is reported that activation of the speech areas and speech lateralization in the various Word Generation tasks (letter, category, antonym word generation) is comparable.⁴² However, in our experience with pediatric patients, the Antonym Generation paradigm does much better in terms of the extent of activation in the speech areas and patient compliance versus letter Word Generation tasks; thus, SWG was not included in the pediatric algorithm. One of the limitations of this paradigm is that given the design, it is difficult to assess patient performance.

Passive Story Listening Task

Passively listening to an aurally presented story involves multiple aspects of language function, including syntactic processing and word recognition. This task offers the advantage of being relatively easy to accomplish for young children or adults whose functional impairments limit their ability to adequately partake in more complex language paradigms.

Passive Story Listening has been shown in pediatric populations to produce activation in the bilateral superior temporal gyrus, bilateral superior frontal gyrus, and left posterior superior temporal gyrus.^{43⇓⇓–46} The ease of performing PSL in young or impaired adult patients must be balanced with the lack of a task performance metric and its relatively weak language activation. There is no way to know if the patient is actually listening and attending to the story during the scan, but patients may be asked to recount the stories they heard following the scan as an indirect means of assessing compliance.

Scanner noise can impede the quality and magnitude of language activation during aurally presented tasks.⁴⁷ Vannest et al⁴⁸ compared similar PSL and active-response tasks and found that both produced language activation in similar patterns of the frontal and temporal lobes but the active-response task produced more dorsolateral prefrontal activation, likely due to engagement of working memory and attention during the comprehension questions. No significant difference in hemispheric lateralization was observed. Although adding an active-response component increased memory and attention engagement as well as effect size in the left inferior frontal gyrus, we chose the simpler version of this task because it still provides adequate activation and lateralization of language function while applying to the greatest range of children and/or impaired adult patients. The control task typically involves reversed playback of the same story to remove semantic processing from the sounds presented or simply playing tones.

Breath Hold Task

One task that may be performed for quality control purposes as part of standard clinical presurgical mapping fMRI examinations is the breath hold task. Using such a task, one can alternate brief breath hold periods with periods of normal self-paced respiration in a standard block design paradigm. A general linear model analysis can be performed to evaluate percentage BOLD signal change occurring during the hypercapnia state relative to the normocapnia condition. Although one may utilize either end-inspiratory or end-expiratory breath holds to induce the transient hypercapnia, and both approaches have relative strengths and weaknesses (eg, some studies have shown greater reproducibility with the latter approach), we have found that in general, the end-inspiratory technique is easier for patients to perform and can be performed even by neurologically debilitated patients.^49⇓–51 The purpose of this task is to evaluate cerebrovascular reactivity (CVR) in a relatively quick, easy, and effective manner without the need for exogenous controlled gas administration as is frequently used for quantitative CVR research studies.^{52⇓⇓–55} Although emerging resting-state BOLD imaging approaches have been suggested as well for assessment of CVR, there remains some controversy regarding the reliability of such methods such as resting-state fluctuation of amplitude, and thus they cannot be recommended currently.^56⇓–58

The main value of this method is its utility in detecting neurovascular uncoupling potential in patients with focal resectable brain lesions such as tumors, vascular malformations, or cortical developmental abnormalities. Neurovascular uncoupling (NVU) refers to the breakdown of the normal neurovascular coupling cascade related to factors such as tumor angiogenesis, astrocytic dysfunction, neurotransmitter, or other biochemical dysfunction or abnormalities of intralesional or perilesional hemodynamics. Generally, the final step in the cascade involves the vascular response related to regional neuronal activation, and thus most cases of NVU manifest as abnormally decreased or absent regional vascular reactivity within or immediately adjacent to the brain lesion of interest.^{49,59⇓⇓⇓–63} Determining potential NVU increases the clinical utility of fMRI by uncovering false-negative language activations in the eloquent cortex.