Functional Gradient Of The Fusiform Cortex For Chinese Character Recognition Part 2

Behavioral analysis

The ACC and RT were calculated for the four conditions. The main effects of stimulus categories were analyzed by one-way repeated ANOVA. Paired t-tests with post hoc Bonferroni correction (p, 0.05) were conducted across conditions.

There is a close relationship between stimulation and memory. The more interesting, novel and challenging something is, the easier it is for us to remember it. This is why we often use various methods in schools to stimulate students' interest and curiosity to improve their memory. So, what stimuli can improve our memory?

First of all, imagination and association skills are important factors in improving memory. We need to let the brain further process the information and connect it with our own experience, knowledge, emotions, etc., to remember it better. For example, when learning a new word, we can associate it with a visual scene or an existing word with similar pronunciation, which can help us understand and remember it better.

Secondly, using multiple senses can also improve memory. We can obtain information through various channels such as vision, hearing, touch, and smell, which can create more connections and associations in the brain and help long-term memory. So when studying, we can try various ways to obtain information, such as listening to recordings while reading, or focusing while taking photos, to better record the information on the spot.

Finally, consistent exercise for your body and brain can also improve your memory. Physical health can improve the flexibility and mobility of the brain and better improve our memory. Regular exercise can promote blood flow, increase oxygen supply, and promote brain metabolism, thereby improving brain function. In addition, keeping the brain active through a lot of reading, thinking, studying, etc. can also improve our memory.

In short, stimulation has a positive effect on memory. We can improve our memory in various ways, such as using more imagination and association skills, using multiple senses to obtain information, and insisting on exercising the body and brain. As long as we persevere, I believe our memory will continue to improve, helping us learn and grow better. It can be seen that we need to improve memory, and Cistanche deserticola can significantly improve memory, because Cistanche deserticola can also regulate the balance of neurotransmitters, such as increasing the levels of acetylcholine and growth factors. These substances are very important for memory and learning. In addition, Meat can also improve blood flow and promote oxygen delivery, which can ensure that the brain receives sufficient nutrients and energy, thereby improving brain vitality and endurance.

Univariate activation analysis

In single-subject level analysis, a general linear model (GLM) was conducted, with the convolution of stimuli onset time (SOT) and hemodynamic response function (HRF) as independent variables, the time series of fMRI signals as dependent variables, and six realignment parameters as regressors. 

In group-level analysis, one sample t-tests were used to analyze each voxel to acquire activation maps for each condition [p, 0.05, FDR correction (q, 0.05), cluster size. 10].

To investigate different functional levels of FG activation during Chinese word recognition, we determined five types of brain activation maps: (1) RWs versus fixation minus PWs versus fixation for lexical effects, (2) PWs versus fixation minus FWs versus fixation for word form effects, (3) PWs versus fixation minus RWs versus fixation for abstract orthographic processing, (4) FWs versus fixation minus RWs versus fixation for low-level orthographic processing, and (5) SCs versus fixation minus RWs versus fixation for basic visual processing. 

Specifically, PWs have the same orthographic regularity as RWs but fail to access lexical phonology and meaning. FWs have regular radicals or logo graphemes but no legal Chinese orthography while SCs were spatially interleaved. 

Together, the functional level is incremental from the first to the fifth contrasts. Besides, fewer processing stages but more activation were expected for the later three contrasts because of the prediction errors because of the last failed stage, i.e., the stronger activation for more attempts to map global orthography onto word phonology and meaning or to integrate local radicals into a whole character (Price and Devlin, 2011).

RSA

RSA is powerful for integrating different levels/scales/modalities (e.g., neural, behavioral, physical, theoretical) activities to identify cognitive manipulation (Fischer-Baum et al., 2017; Wang et al., 2018; Deniz et al., 2019). The current study aimed to investigate the precise functional roles of the FG during Chinese word recognition. 

This goal was achieved by relating the theoretical representational dissimilarity matrix (RDM) of different levels of Chinese orthography and neural RDM in the FG. Quantifying dissimilarities between abstract and lexical orthography is the key question. We achieved this result by calculating the logo-grapheme representations of RWs, PWs, and FWs.

Theoretical RDMs

The logo grapheme is the basic representational unit of Chinese characters (Han et al., 2007). The logo-grapheme RDM was constructed by calculating one minus the ratio of shared basic units between any two stimuli within RWs, PWs, and FWs. 

Note that SCs consist of random strokes, but not all strokes are logo graphemes. Thus, logo-grapheme RDMs can only be constructed for RWs, PWs, and FWs. Logo-grapheme representations indicate internal manipulations treating the logo-grapheme as the minimum unit. 

During character recognition, internal cognitive processes contain lexical orthography (i.e., orthographic legality and mapping word form onto phonology and semantics), word-form orthography (i.e., radical position and orthographic legality), radical orthography (i.e., stroke position), and general visual information composed of light and dark patches. 

During PW recognition, the logo-grapheme representations indicate processing orthographic legality and general visual properties. For FW recognition, the logo-grapheme representations indicate radical and general visual processing.

Semantic representations were calculated for RWs, as PWs and FWs were meaningless. Semantic dissimilarity was calculated as one minus the cosine similarity between word vectors of any pair of RW stimuli. Skip-gram algorithms (window size = 5, subsampling rate = 10 4, negative sample number = 5, learning rate = 0.025, dimension number = 300) were used to calculate word vectors based on the open-source Wikipedia Chinese Corpus.

Neural RDMs and searchlight RSA

A GLM was performed at the first level for each of the 120 trials, with 6 head motion parameters regressed. In each condition (RWs, PWs, and FWs) and for each subject, voxel-wise neuronal similarities between any pair of 40 trials were calculated as significant correlations between b -b-values extracted from a self-centered sphere with a 6- -mm radius. 

A one-minus correlation between any two stimuli was set as the dissimilarity. The centered voxel of the sphere completed transversally within cortical regions of interest (ROIs), such as a searchlight, and voxel-wise neural RDMs were obtained for each subject in each condition. The ROIs in the current study were defined as the bilateral fusiform areas (55#, 56#) in the Automated Anatomical Labeling 3 (AAL3) template. 

Bilateral inferior occipital cortices (53#, 54#) in AAL3 were also included. Spearman's correlations were calculated between neural RDMs and logo-grapheme/ semantic RDMs at the voxel level. Spearman's r-transformed Z values were logo-grapheme/semantic representation values and were used to perform a one-tailed, one-sample t-test across subjects at the voxel level. 

Significant voxels (p, 0.05, uncorrected, cluster size . 10) in the t-test were identified as involved in logo-grapheme/semantic representation. The analysis scripts and the summary data are available on GitHub (http://github.com/miaocao88/ Functional-Gradient-in-vOT).

Validation analysis

To examine whether behavioral performance (ACC) affects brain activity during lexical decision tasks, validation analysis was conducted by excluding trials in which participants inaccurately judged lexicality. Particularly, for PWs condition, 6 participants whose ACC is,50% were excluded to ensure statistical effect of RSA results.

Results

Behavioral results

The ACC and RT of button pressing for the lexical decision task were analyzed. The main effects of ACC and RT among RWs, PWs, FWs, and SCs calculated by one-way repeated ANOVA were both significant, as shown in Figure 1B (Allen et al., 2019). Significant main effects measured by one-way repeated ANOVA were observed for both ACC (F(3,150) = 27.12, p, 0.001) and RT (F(3,150) = 16.68, p, 0.001). 

The ACC of PWs (0.80 6 0.21) was significantly lower than that of RWs (0.95 6 0.07, t(50) = 5.29, p, 0.001, Bonferroni corrected), FWs (0.96 6 0.06, t(50) = 6.12, p, 0.001, Bonferroni corrected), and SCs (0.98 6 0.05, t(50) = 6.23, p, 0.001, Bonferroni corrected), whereas the RT of PWs (938.81 6 15.60 ms) was significantly higher than that of RWs (793.78 6 170.21 ms, t(50) = 9.04, p, 0.001, Bonferroni corrected), FWs (780.416 149.84ms, t(50) = 10.28, p, 0.001, Bonferroni corrected), and SCs (728.686 152.54ms, t(50) = 12.84, p, 0.001, Bonferroni corrected). 

The ACC of SCs was greater than that of RWs (t(50) =2.89, p, 0.05, Bonferroni corrected). The RT of SCs was shorter than that of FWs (t(50) = 4.85, p, 0.001, Bonferroni corrected) and RWs (t(50) = 5.30, p, 0.001, Bonferroni corrected). 

Together, subjects showed the poorest performance in PW recognition compared with the other three conditions but better performance for SCs in the lexical decision task.

Functional activation results

In the current study, the word form effect was defined as activation of PWs versus fixation minus FWs versus fixation, whereas the lexical effect was defined as RWs versus fixation minus PWs versus fixation. As shown in Figure 2A, the word form effect activated the bilateral ventral occipitotemporal cortices and left middle occipital gyrus [p, 0.05, FDR correction (q, 0.05), cluster size. 10]. 

Left word form effect areas were located in a large cluster (cluster size = 472) spanning the middle part of the left lateral occipitotemporal sulcus, including the left inferior temporal gyrus, middle and anterior parts of the left FG, They left inferior occipital gyrus [p, 0.05, FDR correction (q, 0.05), cluster size. 10]. Right word form effect areas involved the contralateral homotopic cortices, including the right inferior temporal gyrus and middle FG. 

The lexical effect activated extensive brain regions, including the bilateral middle occipital gyrus, bilateral occipitotemporal cortices (consisting of the inferior temporal gyrus and middle FG), right FG, and anterior part of the left inferior temporal gyrus [p, 0.05, FDR correction (q, 0.05), cluster size. 10]. 

Massively activated brain regions might be derived from top-down modulation of lexical responses. Note that more anterior activations of lexical effects were found in the anterior part of the left inferior temporal gyrus than in the anterior part of the left FG. For more details, please see Table 1.

Based on the prediction error hypothesis, PWs versus fixation minus RWs versus fixation, FWs versus fixation minus RWs versus fixation, and SCs versus fixation minus RWs versus fixation corresponded to abstract orthographic processing, radical processing, and visual properties extraction, respectively, which belong to higher-to-lower levels of orthographic structure. 

As shown in Figure 2B, PWs versus fixation minus RWs versus fixation activated the bilateral ventral occipitotemporal cortices and bilateral middle occipital gyrus [p, 0.05, FDR correction (q, 0.05), cluster size. 10]. Brain regions for FWs versus fixation minus RWs versus fixation were found in the bilateral inferior temporal gyrus and left middle occipital gyrus [p, 0.05, FDR correction (q, 0.05), cluster size. 10]. 

SCs versus fixation minus RWs versus fixation only activated the left middle and inferior occipital gyrus [p, 0.05, FDR correction (q, 0.05), cluster size. 10].

Gradually changed and intermingled activations along the y-axis in the posterior part of the left inferior temporal gyrus are shown in the lowest panel of Figure 2B and confirmed functional gradients of the left FG. For more details, please see Table 2.

For more information:1950477648nn@gmail.com