Language is a complex brain function that involves two primary language regions, named Broca and Wernicke [1]. According to the pertinent studies on healthy individuals, 92.5%–97% of right- handed people show left-sided lateralization for language functions [2,3&4]. Pathological processes involving the left frontal and temporal lobes yield speech disorders in right-handed patients [2,5&6]. Neurologic deficits that disrupt typical brain architecture have been suggested to affect the functional connectivity between brain regions corresponding to the language area [7]. Functionally active areas during language task activities are shown to be dislocated to either the healthy ipsilateral or contralateral homotopic areas [8].

These interhemispheric patterns or compensation are explained by the reduction of collateral inhibition of specific centers on other structures within the language network, also known as brain plasticity, the ability of the brain to modify and reorganize itself. (10) This shift can be reversed after surgical resection of a tumor leading to improved speech performance [8]. The functional structure pattern is hypothesized to be altered based on brain degeneration and tumor areas' type, size, and location [9 &10].

This research has investigated the effect of tumor location on patients having tumor mass in the brain, closed to main language network areas, using a BOLD fMRI study. Patients with brain tumors who have no speech problems were divided into three groups, no disruptive involvement (non-invasion group), moderate effect (partial-invasion group), and severe eloquent cortex involvement (severe-invasion group). In the first part, we designed the study to check for plasticity in the cases of a destructive lingual network in the Broca area by relatively slow- growing glial tumors. We examined both the severe-invasion and partial-invasion groups in comparison to the non-invasive group. In the second step, the destructive effect of tumor invasion on functional language integrity was investigated using differences between the activation pattern of the non-invasive group compared to partial-invasion and severe-invasion groups. It should be taken into account that all the images and information of patients in this research were extracted from the available data.

Patients

Pre-operative functional MRI datasets for patients who were routinely referred to our lab located in Imam Khomeini Hospital Complex, Tehran, Iran, were retrieved for pre-surgical fMRI assessments between 2017 and 2019. Thirty-two subjects were selected based on being right- handed, having low -grade glioma in their left hemispheres, but without language deficits.

The subjects were then divided into three groups. A non-invasive group, referred to patients with left-brain tumors distant from Broca's area (11 cases), a moderate group, referred to patients with partially invasive left-brain tumors to Broca's area (11 cases) and finally, the severe group in which the left-brain tumors showed severe invasion of Broca's area (10 cases).

Table 1: The demographic information for each group is as follows (age in years (mean±SD), gender (no.F/no.M), and education in years (mean ± SD))

MRI acquisition

Neuroimaging data were acquired on a 3T Siemens Trio scanner at the Medical Imaging Center of Imam Khomeini Hospital, Tehran, Iran. Anatomical T1-weighted and post contrast T1- weighted images, in some cases, were obtained using magnetization-prepared rapid gradient echo (MP- RAGE) sequence (TR=1600ms; TE=2.8ms; Flip angle=8o; FOV=256 mm; Matrix size = 256 x 256; Voxel size = 1x1x 1 mm; Slice thickness = 1 mm and Slice gap = 0 mm). While functional T2* - weighted were collected using blood oxygen level dependent (BOLD) contrast (TR=3000ms; TE=30ms; Flip angle=8o; FOV=192 mm; Matrix size = 64 x 64; Voxel size = 3x3x3 mm; Slice thickness = 3 mm and Slice gap = 0 mm).

Data analysis

The language task utilized is Reverse Word Reading (RWR). It has been shown that this is one of the best language tasks for localization of language areas in Persian speakers and was described in detail elsewhere [1]. To represent the paradigm used in this study briefly, this is a block design, lasting for 6 min and 24s and consisting of 16 Activation, Neuter and Rest blocks in RARN fashion with 24s for each block 128 total volumes. In activation blocks, including 10-word trials, subjects saw five- letter Persian words and the letters in a reverse order in which they were asked to read them once silently. Alternatively, the neuter and rest blocks consisted of Chinese and blank white screen letters. Functional images were analyzed using FEAT (FMRI Expert Analysis Tool), which was a part of the FMRIB software library (FSL http://www.fmrib.ox.ac.uk/fsl) v6.0 [11], which used the powerful General Linear Model (GLM)-based fMRI analysis. Some pre-statistics processing steps had been carried out, including motion correction using MCFLIRT [2]; slice- timing correction using Fourier-space time-series phase- shifting; non-brain removal using BET [3]; spatial smoothing using a Gaussian kernel of FWHM = 5mm; grand-mean intensity normalization of the entire 4D dataset by using a single multiplicative factor; high-pass temporal filtering (Gaussian- weighted least-squares straight -line fitting, sigma=96 seconds). Functional MR images were rejected if they exceeded movement criteria 2.5mm displacement or three o rotations within a functional analysis. Statistical analysis performed on time-series using FILM pre-whitening (FMRIB Improved Linear Model) [4] to reach a valid and maximally efficient statistical and a "Z- score" (which is a transformation of t-statistics dividing by its standard error), devoted to the corresponding BOLD signals. Finally, by thresholding, clusters having higher Z- stat and 0.01 P-value were mainly considered as significantly activated area [8].

The activation maps in Severe, Partial and non-invasion groups

To obtain the mean activation map within each group, we first performed a single-subject analysis at the cluster level, with a z- score greater than 2.3 and a p-value of 0.05, to obtain the mean activation map within each group. The Mixed Effects Module: FLAME1 was then used for group- level analysis. We performed a one-sample t-test with Z > 2.3 and a p-value of 0.01 and calculated activation maps representing the means of each group [7–9]. The same procedure was performed in all groups including severe, partial and non-invasive groups.

Tumor segmentation

By using the transformation matrix computed during the FEAT analysis, all T1-weighted and T1- weighted post-contrast images were linearly registered in the MNI152 standard space (degrees of freedom = 7), if available to the subject, and then the borders of the core of the tumor as well as the border of the whole tumor and edema mass were manually segmented by radiologists using ITK-SNAP tool from the website (http://www.itksnap.org).

The core ROIs and whole tumor ROIs for each group were then overlaid by 50% with the MNI152 standard space threshold (more than 50% of subjects in each group indicate tumors on specified regions).

Figure (1) Illustrates the activations for the single-group average in the sever group, overlaid on MNI152 standard space. Besides, tumor core ROIs (violet colour) and whole abnormal tissue ROIs including edema (yellow colour), are also represented. Similarly, Figure.2 and Figure.3 represent the same results for the partial group and the non-invasion group, respectively.

Figure 1: Average activation map overlaid on MNI152 standard space for the severe-invasion group (2.3 < Z < 6) where Tumor core and whole abnormal tissues are represented by violet and yellow in colour, respectively.

Figure 2: Average activation map overlaid on MNI152 standard space for the partial-invasion group (2.3 < Z < 6) where Tumor core and whole abnormal tissues are represented by violet and yellow in colour, respectively.

Figure 3: Average activation map overlaid on MNI152 standard space for the non-invasion group (2.3 < Z < 6) where Tumor core and whole abnormal tissues are represented by violet and yellow in colour, sequentially.

The Activation Maps for Two Group Differences

To obtain the difference activation maps, after performing a single-subject analysis using a Z-score greater than 2.3 and a p-value of 0.05 at the cluster level, a group-level analysis was performed using fixed-effect and two samples unpaired t-test using Z>2.3 and a p-value of 0.01 [7–9].

In this way, six activation maps for mean group differences were computed for six different contrasts in the three groups, including the non-invasion group minus partial-invasion group (Figure.4), non-invasion group minus severe-invasion group (Figure.5), partial-invasion group minus severe-invasion group (Figure.6), partial-invasion group minus non-invasion group (Figure.7), Severe-invasion group minus partial-invasion group (Figure.8), and severe-invasion group minus non-invasion group (Figure.9).

Figure 4: Difference activation map overlaid on MNI152 standard space for the non-invasion group minus partial- invasion group (2.3 < Z < 14.7).

Figure 5: Difference activation map overlaid on MNI152 standard space for the non-invasion group minus severe- invasion group (2.3 < Z < 11.5).

Figure 6. Difference activation map overlaid on MNI152 standard space for the partial-invasion group minus severe- invasion group (2.3 < Z < 14.7).

Figure 7: Difference activation map overlaid on MNI152 standard space for the partial-invasion group minus non- invasion group (2.3 < Z < 14.7).  

Figure 8: Difference activation map overlaid on MNI152 standard space for the severe-invasion group minus partial- invasion group (2.3 < Z < 14.7).

Figure 9: Difference activation map overlaid on MNI152 standard space for the severe-invasion group minus non- invasion group (2.3 < Z < 11.5).

Age and Gender Effects

To inspect the possibility of significant differences related to age or gender among the three groups, we performed an independent sample t-test for age as a continuous variable and a chi- square test (Fisher's exact test) for gender categorical variable using IBM Statistics SPSS v.22.

According to the results comparing two by two groups, in both age and gender, the computed p- value was greater than 0.05, which means there is no significant difference related to these factors between specified groups.

Although the mechanisms by which brain gliomas alter brain functions are not entirely understood, it is clear that tumors infiltrate and disrupt the cortex and subcortical language areas, initiating a cortical functional impairment in the pertinent hemisphere. Comparing the functional pattern of language in patients with invasive tumors against those with non-invasive alterations of the language network (LN), we found that these patients are presenting an overall decrease in language activation within the main language network, subsequent to reorganization of LN within the ipsilesional hemisphere, or shifting language activity to the contralateral cortex. Some authors believe that activation of the right hemisphere's speech centres may provide a better behavioral advantage in patients with left-hemispheric lesion. However, the precise trigger for the functional reorganization of the language centres in brain tumor patients remains unknown.

When tumor growth starts to invade the functional activity of the brain, two main processes may happen. First, functional impairment initiates based on the size and position of the invasion. This is expected to be responsible for the type and the extent of dysfunction originated from the anatomical brain area's invasion pattern earlier than any functionally-correlated outcome. Consequently, no reorganization (i.e., plasticity) happens in this process. Second, the functional area's reorganization comes into effect; then, no or limited dysfunction is expected. Following, we discuss both different aspects of brain invasion and brain reorganization, considering various sizes and locations of invasive lesions.

Invasive effect of lesion on language disruption

As we expected, in the group with non-invasive tumors compared to the Partial-invasion group, there seems to be no significant change in the primary and secondary language areas due to the tumors' inoffensive effects on the main language network. However, some activation changes were observed in inter-connective reas, which seem primarily to be related to trivial changes in the activation pattern (i.e., minimal disruption) of the associated language areas.

Accordingly, in the cases of more anterior tumors in the frontal lobe, some areas such as right orbitofrontal gyrus (B11) which are responsible for communication between primary and secondary functional areas are activated to compensate for any likely network disruption. On the other hand, in posterior tumors, based on the same concept, areas such as the bilateral ventral posterior cingulate gyrus (B23) and the bilateral dorsal posterior cingulate cortex (B31) were shown to be mostly activated. In the non-invasive group compared to the severe-invasion group, more activation was observed in the primary and secondary areas of language, such as the left pars opercularis area (B44) and left angular gyrus (B39)) on the ipsilesional side. This might be due to the excessive disruption of language network areas in the severe-invasion group and their functional impairment caused by functional disruption of the related language areas.

In the partial-invasion group, compared to the severe-invasion group, the most active areas are the left pars triangularis (B45) and left angular gyrus (B39). This shows that in this group, the language network's primary and secondary areas are still active, whereas, in the severe-invasion group, no activity is seen on the left side of the main language network due to the high degradation in the primary language network.

Plasticity Phenomena Introduced By Lesion Incidence, Disturbing Main Language - Related Areas

In the partial-invasion group, compared to the non-invasive group, excessive activity was seen bilateral in pars opercula (BA44), pars triangular (BA45), and left dorsolateral prefrontal cortex (BA46), and primary motor areas. However, secondary motor area was activated only ipsilateral as it is not usually activated in a healthy brain. All these areas can be justified as a reorganized pattern of speech activation. It appears that due to the partially invaded eloquent cortex caused by slow tumor growth, the main functional network can adapt to the lesion involvement and is reorganized in the same hemisphere, but ultimately it might lead to the recovery of these centres in the contralesional hemisphere.

In the severe-invasion group compared with the partial-invasion group, excessive activity was also seen in the left insular cortex (BA13) and the left dorsal anterior cingulate gyrus (BA32), which may indicate functional reorganization due to the transfer of the language activity center to secondary areas of the language network in the ipsilesional hemisphere after severe disruption of the main speech area (i.e., Broca). This major type of reorganization is also evident from excess activation of the right anterior prefrontal area (BA10), which eventually occurs in the contralateral hemisphere. In the group with the severe-invasion involvement of Broca area, the brain cannot regenerate the primary area of the language network, and therefore primarily move dominant parts of the primary network to the secondary areas, either ipsilesional or contralesional.

Functional rearrangement of the speech centers to the contralateral language network areas is also revealed from the language activation differences in the severe-invasion group compared with the non-invasive group. Here, significant activity was observed in the left anterior prefrontal cortex (b10) and right dorsolateral prefrontal cortex (BA46), suggesting that the functional reorganization of activation areas occurred in the contralesional hemisphere. Moreover, in this group difference, excessive activity in the left anterior prefrontal cortex (BA10) was seen, which again proves the use of the secondary areas of the network even in the same hemisphere to compensate for the damage due to the high volume of destruction of the primary language cortex.

When a tumor starts to invade the functional activity of the brain, two main processes may occur; as destructive or neoplastic effects. It appears to be no significant change in the language functional activity in the cases where tumors are outside the main language area (Broca). However, some activation changes were observed in the language network correlated areas, which seems to be primarily related to a small change in the activation pattern (i.e., minimal disruption) of the associated language areas. In the non-invasion and partial-invasion groups, the main language areas are still active and therefor minimal destructive effect are observed, whereas in the severe-invasion group, no activity are seen on the dominant side of the main language network due to the high degradation in the primary language network. In the cases of tumor invasion, the functional area's reorganization can come into effect due to a destructive lingual network in the Broca area affected by relatively slow-growing glial tumors; therefore, still no or limited dysfunction is expected.

It appears that due to the partially invaded eloquent cortex caused by tumor growth, initially the main functional network is adapted to the lesion involvement and moves mainly to the ipsilateral hemisphere, but finally more invasion leads to the recovery of these centers in the contralesional hemisphere. Moreover, in the group with more severe-invasion involvement of the Broca, brain cannot regenerate the primary areas of the language network, and its functional areas move to the secondary areas, either to ipsilesional or to contralesional hemispheres.

1. Broca PM. Remarques sur le siege de la faculte du langage articule, suivies dune observation daphemie (perte de la parole). Soc. Anat. 1861; 6: 330–357.
2. Knecht S, et , Handedness and hemispheric language dominancein healthy humans. Brain. 2000; 123: 2512–2518.
3. Sommer IE, Aleman A, Somers M, Boks MP, Kahn RS. Sex differences in handedness, asymmetry of the planum temporale and functional language Brain Res. 2008; 1206: 76–88.
4. Szaflarski JP, et al., Language lateralization in left-handed and ambidextrous people: fMRI Neurology. 2002; 59: 238–244.
5. Rutten GJ, Ramsey NF, van Rijen PC, van Veelen Reproducibility of fMRI-determined language lateralization in individual subjects. Brain Lang. 2002; 80: 421–437.
6. Stippich C, Rapps N, Dreyhaupt J, et Localizing and lateralizing language in patients with brain tumors: feasibility of routine preoperative functional MR imaging in 81 consecutive patients. Radiology. 2007; 243: 828–836.
7. Dronkers Nina F, et al. “What Do Language Disorders Reveal about Brain-Language Relationships? From Classic Models to Network ” Journal of the International Neuropsychological Society. JINS. 2017; 23: 741-754.
8. Katarzyna K, et “Reorganization of Language Areas in Patient with a Frontal Lobe Low Grade Glioma - fMRI Case Study”. Polish journal of radiology. 2015; 80: 290-295.
9. Duffau H, Capelle L, Denvil D, et al. Functional recovery after surgical resection of low-grade gliomas in eloquent brain: hypothesis of brain compensation. J Neurol Neurosurg Psychiatry. 2003; 74: 901–907.
10. Fakhri M, Oghabian MA, Vedaei F, et Atypical language lateralization: an fMRI study in patients with cerebral lesions. Funct Neurol. 2013; 28: 55–61.
11. Wernicke C, Der Aphasische Symptomencomplex—Eine Psychologische Studie auf Anatomischer Basis (Cohn & Weigert, Breslau, Germany 1874).
12. Kapsalakis IZ, Kapsalaki EZ, Gotsis ED, et Preoperative Evaluation with fMRI of Patients with Intracranial Gliomas. RadiolRes Pract. 2012; 2012.
13. Tononi G, Sporns O, Edelman GM. Measures of degeneracy and redundancy in biological networks. Proc Natl Acad Sci.1999; 96: 3257–3262.
14. Li Q, Jian DW, et al. “Functional Translocation of Broca's Area in a Low-Grade Left Frontal Glioma: Graph Theory Reveals the Novel, Adaptive Network ” Front. Neurol., 2019, Sec. Applied Neuroimaging.