Developing fMRI protocol for clinical use Comparison of 6 Arabic paradigms for brain language mapping in native Arabic speakers

Participants

This prospective pilot study was done at King Faisal Specialist Hospital & research center (KFSH&RC). Ethical approval was obtained from the institutional review board of KFSH&RC. The study was conducted in accordance with the Helsinki Declaration. Radiological data were kept strictly confidential. Literature review was performed through electronic search (Google Scholar, PubMed). The purpose of the study was explained to participants in both oral and written forms, and informed consent was obtained. Twenty-four healthy adults (50% men) age 22 to 39 years (mean=29.41 years) participated in this study. Each participant was asked to fill a form relating to MRI room safety. Exclusion criteria included presence of known neurological or psychiatric diseases, pregnancy, visual or hearing defects, or any brain pathology or abnormal brain morphology. Participants were all right-handed native Arabic speakers. Handedness has been measured using the Edinburgh Handedness Inventory (mean laterality index was 0.86).24

Task

Based on a review of commonly used fMRI designs in neurolinguistic studies,25-28 we developed 8 paradigms to investigate the neural basis brain regions of different language aspects such as syntactic, semantic, and phonologic processing. Our tasks aimed on assessing language areas in the frontal and temporal lobes. Two of the developed tasks (Sentence Completion and Object Naming) were not used in data collection (one because of challenges relating to collection of response accuracy, and the latter due to being covered by other tasks as to region activation). The remaining 6 paradigms had the same scanning duration time, which is 4:30 minutes, except rhyming (6:48 minutes). Block design was used with action and rest length of 30 seconds each, except rhyming. Each paradigm had four activation onsets and five resting intervals (apart from rhyming, described below), starting and finishing the paradigm with a resting moment. In all paradigms (apart from rhyming), during the rest block, a crosshair was presented to the subjects, and they were asked “to-do-nothing” at rest while keeping eyes open (Table 1).

Paradigms detail

1. Rhyming (RH)

This paradigm is aimed to activate language areas in the left dorsolateral prefrontal cortex (DLPC), Broca’s area and the posterior lateral fusiform gyrus, supramarginal gyrus, and superior temporal gyrus, and is largely based on method used by Lurito and his colleagues.26 Nonetheless, Arabic word rhymes did not yield the same variety in task complexity as the original English task by this group: That is, words written in Arabic ending with the same spelling had much higher chances of rhyming, whereas in English words could be spelled similarly at the end, but do not rhyme (“comb” and “tomb”). This may have landed the task relatively easier to complete than the English format. Words were selected from a pool of high frequency words, and reviewed by a Native Arabic clinical neuropsychologist and an Arabic language teacher. Attention was made to select words with high frequency (based on dictionary by Buckwalter29, and concreteness and familiarity (qualitatively). Imaginability aspects were not addressed. The task started and ended with 12-second rest moment. It consisted of 16 blocks, eight blocks of pair words, and eight blocks of pair symbols, interleaved to each other. Each block contained eight different pairs of word or symbol pairs (total of 64 pairs of word and 64 pairs of symbols) that remained for three seconds, and was displayed to the participant through goggles. Participants were asked to press a button on the response pad when stimuli do not-rhyme (Table 1). Nonetheless, participants responses were not collected during neither format of rhyming, and activation data was collected for the word rhymes only. There were efforts or providing auditory instructions and formats of the tasks (e.g. verb generation as presented in auditory format vs. written word format) in an attempt to have some of these paradigms available for illiterate or low educated people in the future.

2.Semantic Category Generations (SCG)

This paradigm is a single word semantic word fluency task, based on method by Kircher et el.25 Semantic fluency has been associated with activation in the temporal (fusiform gyrus) and left middle frontal lobes.30 It consisted of four different semantic categories, presented in the form of a word display for 30 seconds, followed by rest blocks of 30 seconds, and repeated over the whole scan which lasted for 4:30 minutes. Categories included animals, furniture, fruit and vegetables, and clothes. The participant was asked to think about words belonging to the category displayed (Table 1).

3.Silent Word Generation (SWG)

As an attempt to assess activation in the DLPC and inferior frontal gyrus, this covert word generation task,25,26,28 three Arabic letters were displayed in each block (each letter lasted for 10 seconds), followed by 30 seconds rest, alternating over 4:30 minutes. Participants were asked to think about words starting with the displayed letter, and to do nothing at rest, while keeping eyes open (Table 1). Letter were selected based on a Saudi study exploring difficulty in generating words for each of the Arabic alphabet31 They were presented in the order of difficulty, with letters generating more words (based on the study), were placed earlier. This was proposed to increase participants’ task engagement. Covert format was chosen since data was not collected on response accuracy.

4.Verb Generation - Picture Format (VGp)

Verb generation tasks have been associated with activation in the inferior frontal gyrus (IFG).32,33 In this task, there were six black/white images in each activation block and each single line image lasting for 5 seconds. The total was 24 different images.

Participants were asked to think of as many verbs (action words) they could as prompted by each image shown during the scan. They were asked to relax when the crosshair screen appeared, while eyes kept open (Table 1). The pictures used were derived from database by Sondgrass and Vanderwart34 as well as drawings from a local database. The set of verb prompting stmilui were finalized by a clinical psychologist based on word frequency, as well as familiarity (qualitatively).

5.Verb Generation words (VGw)

Here, same set of stimuli as in the VGp tasks were , but in a different order, and using prompts in a single-written word form. A noun was projected onto a screen, and participants were instructed to think of a verb that was associated with it. For example, the word “ball” might generate the verb “hit” (Table 1).

6.Verb Generation audio (VGa)

In this task, the same set of words as the VGw task above were used, but with the words presented auditory through headphones. As the other two VG tasks, participants were instructed to think of responses, but not to vocalize it (Table 1).

Stimulus Presentation

Stimuli were presented using Nordic Neurolabs (NNL) (Bergen, Norway) head-coil mounted goggles and electrostatic headphones.

Data acquisition and analysis

Imaging acquisitions were performed on a 3.0 Tesla MRI scanner (TRIO SIEMENS) at the imaging facilities of KFSH&RC. A high-resolution anatomical scan was obtained over four minutes. For fMRI, BOLD-sensitive sequences were used to measure the signal over time across a variety of language localization paradigms. Functional volumes were collected every three ms, echo time of 30 ms, matrix size of 64X64, and flip angel of 90o. Slices were selected for each subject based on anatomical landmarks, and the functional tasks employed. Functional scans were conducted to examine six different stimulation paradigms per subjects. The total scanning time was approximately 33 minutes (1988 seconds).

FMRIB Software Library (FSL) software was used to analyze fMRI data. Functional data were preprocessed to remove low frequency linear drift in the signal from in each functional data set. Motion correction, temporal filtering, and spatial smoothing (5mm FWHM Gaussian window) of the data were applied to increase the signal to noise ratio. For this aspect of data preprocessing, standard filtering techniques available in FSL were applied. Individual activation maps from each patient were obtained. The size of the activated area was calculated from the volume of activation exceeding a significance threshold of p<0.001 and forming a cluster significance of p<0.05. Activation peaks and size of areas of activation were directly correlated with participants’ anatomical MRI. Participants had the opportunity to learn and practice the task before running the fMRI experiment. A board-certified neuroradiologist screened all images and no abnormalities were detected in any participant.

Laterality index

Lateralization of brain activity has been assessed for each participant by calculating the laterality index (LI). It was calculated as the proportion of active voxels in the left versus the right region of interest (ROI) (Table 2) averaged across multiple thresholds.35 Using the cluster tool in FSL, we test the effects of different thresholds on the second-level whole brain map by setting a threshold, and a range of z-values at (z=1.0, z=1.5, z=2.3). Then the total numbers of active voxels in all ROIs in both hemispheres were determined by using the Fslstats tool. Finally, we used the formula: LI = (left−right) / (left+right) to calculate LI. This yields a score for LI range from +1 (all left hemisphere activation only) to −1 (all right hemisphere activation only) and the intermediate values reflect varying degrees of laterality.35 This method was most appropriate for this study as tasks differed in the type of software used.