Fundamental functional differences between gyri and sulci: implications for brain function, cognition, and behavior

Abstract

Folding of the cerebral cortex is a prominent characteristic of mammalian brains. Alterations or deficits in cortical folding are strongly correlated with abnormal brain function, cognition, and behavior. Therefore, a precise mapping between the anatomy and function of the brain is critical to our understanding of the mechanisms of brain structural architecture in both health and diseases. Gyri and sulci, the standard nomenclature for cortical anatomy, serve as building blocks to make up complex folding patterns, providing a window to decipher cortical anatomy and its relation with brain functions. Huge efforts have been devoted to this research topic from a variety of disciplines including genetics, cell biology, anatomy, neuroimaging, and neurology, as well as involving computational approaches based on machine learning and artificial intelligence algorithms. However, despite increasing progress, our understanding of the functional anatomy of gyro-sulcal patterns is still in its infancy. In this review, we present the current state of this field and provide our perspectives of the methodologies and conclusions concerning functional differentiation between gyri and sulci, as well as the supporting information from genetic, cell biology, and brain structure research. In particular, we will further present a proposed framework for attempting to interpret the dynamic mechanisms of the functional interplay between gyri and sulci. Hopefully, this review will provide a comprehensive summary of anatomo-functional relationships in the cortical gyro-sulcal system together with a consideration of how these contribute to brain function, cognition, and behavior, as well as to mental disorders.

Introduction

Cortical folding is a hallmark of brain topography whose variability across individuals and groups is strongly coupled with inter-individual and inter-group differences in brain function, cognition, and behavior. Gyri and sulci are the standard nomenclature (Fig. 1) and are usually defined and applied to anatomical domains (Cachia et al., 2003a; Toro and Burnod, 2003). The anatomical and functional relationships between gyri and sulci have not been clearly elucidated. For example, at a coarser resolution of cortical region parcellation, primary sulci have relatively more constant relationships with cytoarchitecture or functional architecture, while the accuracy of this alignment drops for secondary/tertiary sulci on higher-order association cortices involved in complex cognitive functions (Zilles et al., 1997; Roland and Zilles, 1998; Hasnain et al., 2001; Morosan et al., 2001; Fischl et al., 2008; Hinds et al., 2008; Frost and Goebel, 2012). An increasing number of studies over the last decade have therefore investigated, at a finer resolution, whether these two elementary anatomical units could serve as building blocks of cortical morphology and brain function, and whether they could be functionally and cognitively differentiated (e.g. Uddin et al., 2010; Deng et al., 2014; Jiang et al., 2015a, 2018a).

Figure 1:

The gyral/sulcal parcellation of cortical surface based on the anatomical information (average convexity in this figure, adapted from Liu et al., 2019). Gyri has larger average convexity than sulci.

It has been demonstrated repeatedly that the local gyro-sulcal morphology of many cortical regions could predict the location of functional activity in adult human brains (Amiez et al., 2006, 2013; Benson et al., 2012; Amiez and Petrides, 2014, 2018; Li et al., 2015; Weiner et al., 2018; Troiani et al., 2020). For instance, the locations of visuomotor eye and hand conditional activity are localized in different sub-regions within the superior precentral sulcus (Amiez et al., 2006). The retinotopic organization of striate cortex can be predicted by surface topology (Benson et al., 2012) and feedback-related activity during exploration is differentially located in the cingulate sulcus and paracingulate sulcus of different individuals (Amiez et al., 2013). There are also three somatotopically organized cingulate motor areas along the cingulate sulcus, each of which is associated with specific morphological features of the cingulate/paracingulate cortex (Amiez and Petrides, 2014). The functional activities of different action controls are localized in different sulci in the posterior lateral frontal cortex (Amiez and Petrides, 2018) and different value signals are localized in different sulcal regions of the orbitofrontal cortex (Li et al., 2015; Troiani et al., 2020). Moreover, local gyro-sulcal morphology can predict the location of functional organization in adult human brains, with the presence/absence of the inferior rostral sulcus and the subgenual intralimbic sulcus influencing inter-participant variability in the ventromedial prefrontal cortex (vmPFC) sulcal morphology, and the vmPFC peak location in the default mode network (Lopez-Persem et al., 2019).

Associations between gyro-sulcal patterns and human behavior have also been demonstrated to investigate brain-behavior relationships as well as neuropsychological correlates of inter-participant cortical folding variations (Fornito et al., 2004; Whittle et al., 2009; Cachia et al., 2018; Yang et al., 2019). For instance, inter-participant differences in paracingulate sulcus asymmetry in healthy males have been shown to be related to individual differences in executive abilities in cognitively demanding tasks (Fornito et al., 2004). The leftward anterior cingulate cortex asymmetry pattern of male early adolescents is associated with individual differences in temperament (Whittle et al., 2009). Adults whose left lateral occipito-temporal sulcus were interrupted by the gyrus in the posterior portion showed better reading fluency than those who had a continuous one, and the greater the interruption, the better was their performance (Cachia et al., 2018). Another study reported more gyral than sulcal regions within the middle frontal cortex, inferior parietal lobe, and visual cortex, for which rs-fMRI signal temporal variability was positively correlated with fluid intelligence (Yang et al., 2019). In addition to studies on adults, one on newborns has reported that there were specific morphological characteristic differences between twins and singletons, as well as between normally developed newborns and those with intrauterine growth restriction. Indeed, these brain morphological measures at birth could predict infants’ cerebral volumes and neurobehavioral development at term-equivalent age (Dubois et al., 2008).

To measure brain functional activity, functional magnetic resonance imaging (fMRI), rather than other techniques including electroencephalography (EEG), magnetoencephalography (MEG), and functional near infrared spectroscopy (fNIRS), has been predominantly adopted for gyro-sulcal functional characteristic studies. This has been due to the development of advanced in vivo fMRI technologies for both high spatial and temporal resolution that can both guarantee accuracy of gyri/sulci localization and measure the bandwidth range of their blood oxygen level dependent (BOLD) signals (Biswal et al., 1995; Logothetis, 2008; Friston, 2009; Van Essen et al., 2013). Gyro-sulcal functional characteristics have been extensively studied during performance of different tasks based on task-based fMRI (t-fMRI) (Logothetis, 2008; Friston, 2009; Barch et al., 2013), or during task-free rest conditions based on resting state fMRI (rs-fMRI) (Biswal et al., 1995; Smith et al., 2013). In addition to the commonly adopted fMRI data in volume space, the Human Connectome Project (HCP) has also publicly released "grayordinate" fMRI data (Barch et al., 2013; Glasser et al., 2013; Smith et al., 2013; Van Essen et al., 2013). The HCP grayordinate data consists of both cortical surface vertices and subcortical voxels of gray matter in the standard MNI152 space, which correspond closely across different individual brains to facilitate group-wise analysis (Glasser et al., 2013). The grayordinate fMRI (including both t-fMRI and rs-fMRI) data is especially suitable for gyro-sulcal functional characteristic studies since the gyri/sulci as well as their fMRI signals are defined on the cortical surface vertices with satisfying accuracy in the grayordinate space (Glasser et al., 2013). In this review, our focus will be on fMRI-based studies since reported findings are substantially more extensive and better established.

This review article will include the following sections: (i) a review of gyral/sulcal definition and labeling scheme in individual brains; (ii) a review of structural substrates for gyro-sulcal functional differences; (iii) a review of the major findings of gyro-sulcal functional differences; (iv) a dynamic functional model of gyri and sulci; and (v) discussion of future research priorities.

Gyral/Sulcal Definition and Labeling Scheme in Individual Brains

Definition of gyri and sulci

In previous studies, gyri and sulci are usually treated as morphological and anatomical nomenclature, and usually defined based on morphological features (Li G. et al., 2009, 2010). For example, in FreeSurfer (Fischl, 2012), a mid-surface is defined such that the displacements of all vertices from their original locations to it are summed to zero. Usually, gyral vertices have negative values (larger average convexity) and sulcal vertices have positive values (Fig. 1). An alternative approach is labeling sulci and gyri via matching individual images to a gyro-sulcal atlas. This approach has caught the attention of neuroscientists since the pioneering work of Brodmann (1909), which demonstrates that architectonic/functional boundaries align with sulci on human brains (Watson et al., 1993; Roland and Zilles, 1994; Tzourio-Mazoyer et al., 2002; Fischl et al., 2008). Although this anatomy–function relation is less pronounced in secondary and tertiary cortices across participants due to great inter-individual variabilities in folding geometry (Fischl et al., 2008; Onoet al., 1990), major sulci and gyri, which develop earlier, have usually been used as anatomical proxies, or landmarks, to align macroscale functional regions. It is worth noting that biological definition and delineation of gyral and sulcal regions, as well as the transition area of gyral wall, warrants extensive studies in the future.

Methods for sulci/gyri labeling

In general, there are two main methods for sulci/gyri labeling. The first method uses deformable registration to warp the labeled sulci/gyri on template to a single subject (Behnke et al., 2003; Desikan et al., 2006). Most of these registration methods were traditionally designed to achieve automatic volumetric alignment but do not constrain cortices to align (Talairach and Szikla, 1967; Woods et al., 1998; Johnson and Christensen, 2002; Shen and Davatzikos, 2003; Joshi et al., 2007). With an increasing interest in cortex, many recent registration approaches have been performed on the entire cortical surface, rather than volumes, to align shape metrics such as curvature and sulcal depth (Fischl et al., 1999; Tosun, Rettmann and Prince, 2004; Goebelet al., 2006). These methods, aiming at the alignment of the entire volume or cortical surface, however, could lead to misalignment of cortical landmarks of interest such as sulci/gyri (Pantazis et al., 2010). Therefore, in a second method, manually or automatically identified landmarks, such as sulci/gyri, have been explicitly used to constrain the registrations of volumes and surfaces (Thompson and Toga, 1996; Davatzikos, 1997; Joshi and Miller, 1997; Collins et al., 1998; Van Essen et al., 1998; Thompson et al., 2000; Hellier and Barillot, 2003; Glaunes et al., 2004; Van Essen, 2005; Desikan et al., 2006; Joshi et al., 2007). These methods have thus demonstrated achievement of direct improvement of the cross-subject correspondence for landmarks (Desai et al., 2005; Pantazis et al., 2010).

In addition to the registration methods, there are studies that directly identify and label sulci/gyri in individual spaces from volumes and cortical surfaces (Mangin et al., 1995; Goualher et al., 1997; Lohmann, 1998; Lohmann and von Cramon, 2000; Zhou et al., 1999; Rivière et al., 2002; Rettmann et al., 2002; Kao et al., 2007; Li G et al., 2009, 2010). These methods preserve inter-individual variability to the best extent, especially for the secondary and tertiary sulci/gyri, but still might need additional steps (possibly registration methods) before or after the labeling in individual space to estimate cross-subject correspondence (Joshi et al., 2012).

In summary, these sulci/gyri labeling methods were developed on a prerequisite that these anatomic landmarks align well with brain cytoarchitecture and functions such that they can be used as a proxy. However, such a relationship is variable in more developed higher-order cortices (Ono et al., 1990; Cachia et al., 2003b; Fischl et al., 2008), which have abundantly more variable secondary and tertiary cortical folding that could often produce misalignment across participants or unsatisfactory labeling in individual spaces. Therefore, a fundamental improvement could possibly be achieved by exploring a more precise and higher-resolution mapping between folding patterns and brain cytoarchitecture or brain functions.

Structural Substrates for Gyro-Sulcal Functional Differentiation

Neuronal processes and gyro-sulcal pattern development

The developing brain undergoes a number of complicated and inter-related neuronal processes (Smart et al., 2002; Betizeau et al, 2013; Borrell and Götz, 2014; Florio and Huttner, 2014; Lewitus et al, 2014; Taverna et al., 2014). Many of these processes have been demonstrated to be relevant to cortical folding and even to specific gyral/sulcal patterns (Noctor et al., 2004; Englund et al., 2005; Fietz et al., 2010; Hansen et al., 2010; Reillo et al., 2011; Wang et al., 2011; Reillo and Borrell, 2012; Betizeau et al., 2013; Pilz et al., 2013; de Juan Romero and Borrell, 2015; Johnson et al., 2015). In particular, the abundance of basal radial glia cells (bRGCs) in the outer subventricular zone (oSVZ) is suggested to be tangentially heterogeneous (Borrell and Gotz, 2014), such that an increase in bRGC number in local areas gives rise to the fanning-out of a radial fiber scaffold and dispersion of radially migrating neurons, leading to a gyrus-like structure in the corresponding locations in the cortical plate while the germinal layers remain unfolded (Fietz et al., 2010; Hansen et al., 2010; Reillo et al., 2011; Wang et al., 2011; Pilz et al., 2013; Stahl et al., 2013; Borrell and Götz, 2014; Gertz and Kriegstein, 2015). The regulation of basal progenitor heterogeneity in the fetal phase persists in the mature stage. Gyri also have a relatively higher absolute cortical thickness and total neuron number, especially in the deep cortical layers (V + VI) (Hiletag and Barbas, 2005).

However, it is still unclear whether the timing of abundance of progenitors in oSVZ is concordant with that of gyrogenesis (Kroenke and Bayly, 2018), because the cerebrum is still lissencephalic when neurogenesis is complete around embryonic day (E) 90 (Rash et al., 2019). In this regard, other processes, such as gliogenesis, axonogenesis, and myelinogenesis are also correlated with the development and growth of cortical folding (Chi et al., 1977; Huang et al., 2009; Takahashi et al., 2010, 2012; White et al., 2010; Dubois et al., 2014; Hagemeyer et al., 2017; Rash et al., 2019). Taking gliogenesis as an example, as the oSVZ grows, more astrocytes and oligodendrocytes are produced than neurons until after E125, which coincides with the formation of white matter cortical connections and the onset of gyrification (∼E100–E125). Nevertheless, the contribution of gliogenesis alone may be insufficient to induce cortical folding and may additionally need to be considered jointly considered with axonogenesis (Rash et al., 2019). Axonogenesis has been relatively better studied and is more closely related to cortical folding, not only due to its concomitant occurrence with gyrification (Huang et al., 2009; White et al., 2010; Takahashi et al., 2012; Dubois et al., 2014), but also due to abundant direct support from observations that manipulations of axon growth lead to massive alterations in gyrification patterns (Rakic, 1988).

Neuronal processes and gyro-sulcal functional differentiation

The heterogeneity in neuron distribution (Hiletag and Barbas, 2005) and axon distribution (Huang et al., 2009; Takahashi et al., 2012; Ouyang et al., 2017), induced by the processes described in the section of 'Neuronal processes and gyro-sulcal pattern development', is related to gyro-sulcal functional differentiation. Within the cortex, somata and arbors of pyramidal neurons are stretched in gyri and compressed in sulci due to the heterogeneity in neuron numbers (Hiletag and Barbas, 2005), resulting in different lengths of apical segments. Identical neurons with different lengths of apical segments have been demonstrated to produce different types of action potentials (Fig. 2A, Hiletag and Barbas, 2005). Thus, differences in dendrite morphology between sulci and gyri result in putative functional differences in attenuation of excitatory postsynaptic potentials, and possibly provide a fundamental explanation for gyro-sulcal differences in fMRI signal time-frequency characteristics outlined in the section of 'Gyro-sulcal functional activity time-frequency differences'.

Figure 2:

(A) A low magnification overview of coronal sections through 8AD (area 8 of cortex, anterodorsal part) and 8B (area 8B of cortex) in the rhesus monkey prefrontal cortex. Tissue was processed for SMI-32, which labels largely pyramidal neurons in cortical layers II + III and V + VI. (A1)–(A3) High magnification views of layer V and II at the gyral crest and gyral wall in 8AD, and sulcal fundus in 8B. Arrowheads highlight the neurons of typical morphologies. Elongated dendritic arbors and cell bodies are observed in (A1) and (A2) while tangentially oriented dendrites and flattened cell bodies are observed in (A3). (B) A low magnification overview of coronal section through 8AD and 8B. Weil's stain was applied to the tissue, where the myelin sheaths are stained dark blue. (B1)–(B3) High magnification views of boundaries between white matters and gray matters at gyral crest and gyral wall in 8AD, and sulcal fundus in 8B. Arrowheads in (B1) and (B2) highlight the myelin sheaths that radially oriented across the boundaries. Such myelin sheaths are barely seen in (B3), where myelin sheaths tangentially course along the sulcal fundus. These data in (A) and (B) were available at http://brainmaps.org (Mikula et al., 2007). (C) White matter axonal wiring patterns between gyri and sulci. Left column: maximum principal curvature (CUR) and diffusion tensor imaging (DTI) derived white matter fiber density (FD) were mapped to the cortical surface as references. Middle: illustration of DTI fiber connection strength between precentral gyri (PCG, postcentral gyri (POG), central sulcus (CS), and postcentral sulcus (POS) on two hemispheres. Line thickness indicates the connection strength. Right: DTI fiber connections between PCG, POC, CS, and POS in detail. The red, white, green, and blue bubbles represent the regions of interest (ROI) on PCG, POG, CS, and POS, respectively. Lines between ROI as well as the colors indicate the presence of connections and their strength. These figures in the middle and the right panels were adapted from Deng et al., 2014. (D) Left: cortical thickness (CT) and graphic metrics, including the degree (DEG), strength (STR), efficiency (EFF), betweenness (BET), and participation (PAR) of a structural connective network mapped to a white matter cortical surface. Right: comparison between gyri and sulci regarding the metrics in (C) and (D). These figures are adapted from Zhang et al., 2020.

Axonal transmission of action potentials is a key component of communication between neurons and a variety of theories have been proposed to interpret the contribution of axons to gyro-sulcal patterns and their related functional architecture (Van Essen, 1997; Xu et al., 2010; Nie et al., 2012; Budde and Annese, 2013; Chen et al., 2013; Zhang et al., 2014). In Van Essen, 1997, axons were found to be under a form of tension and assumed to pull cortical regions closer to each other. The cortical regions being pulled became gyral walls or sulci, leaving the other interleaved cortical regions with weaker axonal connections as gyri. Although tension along axons has been demonstrated by microdissection, the axons, as shown in Fig. 2B, were aligned circumferentially beneath the sulcal depth and radially inside gyri (Xu et al., 2010). Such a spatial relation between axons and gyro-sulcal patterns, which has gained support from both microscale (Budde and Annese, 2013) and macroscale observations (Nie et al., 2012; Zhang et al., 2014), is, however, not consistent with the proposed caudal effects of axon tension on folding in Van Essen(1997). Despite the spatial relation argument, there might nevertheless be a consensus that axonal wiring patterns are different between gyri and sulci. Also, axonogenesis is unlikely to be the only driving force for cortical folding since other neuronal maturation and mechanical processes could make significant contributions (Bayly et al., 2013; Ronan et al., 2014; Razavi et al., 2015; Zhang et al., 2016, 2017).

More importantly, the different axonal wiring patterns between gyri and sulci suggest that they may play different functional roles (Van Essen, 1997; Deng et al., 2014; Zhang et al., 2020). According to parsimonious principles of brain development (Chklovskii et al., 2002; Laughlin and Sejnowski, 2003; Bullmore and Sporns, 2012), Van Essen (1997) has proposed that cortical regions connected by denser axons are pulled closer to each other, such that the information communication and wiring construction costs are reduced. However, the cortico-cortical axon wiring diagram in this theory seems to consist of repeated units: cortical pairs connected by axons. If the axon-tension solution for wiring optimization is applied to all units, however, there would only be short-range connections over the entire cortex. In fact, many studies have reported the presence of considerable number of long-range axons, and that the nature of the axon wiring diagram is a "small-world" or "rich-world" network rather than a composition of repeated units linked by local short-range axons (Watts and Strogatz, 1998; He et al., 2007; van den Heuvel and Sporns, 2011; van den Heuvel et al., 2012). Brain networks allow the presence of long-range connections between gyri (Xu et al., 2010; Nie et al., 2012; Budde and Annese, 2013; Chen et al., 2013; Deng et al., 2014; Zhang et al., 2014), which are distant from one other. Regarding cortical folding patterns, long-range gyro-gyral white matter connections have been demonstrated to be stronger than gyro-sulcal and sulco-sulcal ones (Fig. 2C, Nie et al., 2012; Deng et al., 2014). These long-range gyro-gyral connections introduce a higher wiring cost but this can be offset by their higher capacity as shortcuts in reducing communication steps and upgrading global integration in brain networks (Kaise and Hilgetag, 2006). In this sense, denser gyro-gyral long-range connections might endow gyri with a plausible role as "hubs" of a structural network, which has been demonstrated in Zhang et al. (2020) (Fig. 2D), and will be revealed by functional connectivity/network analyses in the section of 'Gyro-Sulcal Functional Differences from Various Perspectives'.

Support from cross-species comparative studies on brain structure

The complexity of cortical convolution increases generally along the phylogenetic tree and is usually related to species intelligence (Roth and Dicke, 2005). Thus, identifying species-preserved and -specific anatomy and function is a unique and important approach to understanding the development of cortical folding, as well as its relation to structural and functional architectures (Orban et al., 2004; Rilling et al., 2008; de Schotten et al., 2012; Chen et al., 2013; Zhang et al., 2013; Zhang et al., 2014; Li et al., 2017; de Schotten et al., 2019; Eichert et al., 2020).

By limiting our interests to gyro-sulcal patterns, we have found that denser long-range axons between gyri in the section of 'Neuronal processes and gyro-sulcal functional differentiation' are preserved across humans and other primate species such as chimpanzees, macaques, and marmosets (Nie et al., 2012; Chen et al., 2013; Zhang et al., 2014; Li et al., 2020a), as well as in other mammalian species such as ferrets (Xu et al., 2010) and even mice (cerebellum, Zeng et al., 2015). In addition, cross-species comparative studies have revealed that U-shaped short-range fibers coursing along sulcal fundi comprise a larger portion of the human brain relative to that of chimpanzees and macaques (Zhang et al., 2014). In Chen et al.(2013) the elaboration and complexity of gyral folding increases from macaque to human and is positively correlated with an increase in the complexity of axons connected to these gyri. These observations have led to a hypothesis that gyral folding and fiber connection patterns "co-evolve" in primate brainsl (Chen et al., 2013). This co-evolution of gyral folding and fiber connection patterns is also related to the evolution of brain function. For example, de Schotten et al. (2012) and Rilling et al. (2008) found major differences between human and non-human primates in the arcuate fasciculus that may underlie language function. A graph analysis in Zhang et al. (2020) has also revealed that, in contrast to sulci, gyri play an increasingly important role in core cortico-cortical networks across non-human and human primates. Taken together, these cross-species studies, although in their infancy, suggest that gyro-sulcal patterns are critical to cortical adaption along the phylogenetic tree and strongly related to the evolution of brain structural/functional organization, cognition, and behavior (Roth and Dicke, 2005; Eichert et al., 2020).

Gyro-Sulcal Functional Differences from Various Perspectives

By adopting the methods of sulci/gyri definition and labeling in individual brains in the section of 'Gyral/Sulcal Definition and Labeling Scheme in Individual Brains' and further co-registering the individual volumetric fMRI to the structural data, the fMRI signals of gyri/sulci in the volume space were mapped and separated (Deng et al., 2014). The fMRI signals of gyri/sulci in the cortical space were also separated via mapping the co-registered fMRI signals onto the cortical surface or directly adopting the HCP grayordinate fMRI data.

Gyro-sulcal functional activity time-frequency differences

The human brain is considered to be a multi-frequency oscillation system (Deco et al., 2017), in which high-frequency activity may reflect local domain cortical processing, and low-frequency activity synchronization across distributed brain regions (von Stein and Sarnthein, 2000; Buzsáki and Draguhn, 2004; Canolty and Knight, 2010; Siegel et al., 2012; Buzsáki et al., 2013). While the conventional temporal resolution of fMRI with a repetition time (TR) of 2 s corresponding to a sampling frequency of 0.5 Hz has warranted the gyro-sulcal functional activity time-frequency analysis without bias (Zhang et al., 2018a; Ge et al., 2019), the development of in vivo fMRI techniques has significantly improved fMRI signal temporal resolution and further facilitated the analysis in wider frequency bands of brain activity. For example, the HCP fMRI data (Barch et al., 2013; Smith et al., 2013) has a TR of 0.72 s corresponding to a sampling frequency of 1.39 Hz, which covers multiple frequency bands of brain activity (0.01–0.027, 0.027–0.073, 0.073–0.198, 0.198–0.5, and 0.5–0.69 Hz) based on previous electrophysiological studies (Penttonen and Buzsáki, 2003; Buzsáki et al., 2013). Investigating the potential time-frequency differences between gyro-sulcal fMRI signals has therefore received increasing interest.

Fast Fourier transformation has commonly been used in previous studies for calculating the power spectrum of gyral/sulcal signals (Zhang et al., 2018a; Ge et al., 2019; Liu et al., 2019; Jiang et al., 2020). The power spectrum distribution characteristics across different frequency bands, as well as other measures (e.g. "1/f" characteristic of the power spectral density plot) have been analyzed and compared between gyro-sulcal signals (Zhang et al., 2018a; Ge et al., 2019; Liu et al., 2019; Jiang et al., 2020). Wavelet entropy (Mallat, 1989; Coifman and Wickerhauser, 1992; Daubechies, 1992; Sang et al., 2011) analysis has also been used (Liu et al., 2019), which quantifies the degree of temporal signal complexity, with higher entropy indicating a more complex signal temporal pattern (Sang et al., 2011).

With the rapid development of deep learning techniques (LeCun et al., 2015) in medical image analysis studies (Litjens et al., 2017; Shen et al., 2017), the convolutional neural network (CNN) has also been adopted for gyral/sulcal fMRI signal analysis given its ability in learning meaningful and discriminative features from raw data as well as class-specific and nonlinearly distributed features of signals in different frequency scales (Golik et al., 2015; Zhou et al., 2016; Huang et al., 2017). Liu et al. (2019) performed one of the earliest studies using a CNN-based classifier to reveal differences in gyro-sulcal functional activities based on HCP grayordinate rs-fMRI and t-fMRI BOLD signals. In another study (Zhang et al., 2018a), the CNN-based classification model was modified by adding more convolution layers to improve the classification accuracy of volumetric fMRI data. Macaque rs-fMRI data was also used for cross-species studies (Zhang et al., 2018a). In a recent study (Ge et al., 2019), the CNN-based model was adopted on both healthy people and autism spectrum disorder (ASD) patients. Finally, a more recent study (Jiang et al., 2020) adopted a region-specific one-dimensional CNN model to investigate the characteristics of such gyro-sulcal functional activity differences within each brain region instead of in the whole brain.

The gyral and sulcal t-fMRI/rs-fMRI signals (TR ranges from 0.72 to 2 s) can be effectively differentiated based on the proposed time-frequency-based analysis as well as deep-learning-based approaches (Zhang et al., 2018a; Ge et al., 2019; Liu et al., 2019; Jiang et al., 2020). Moreover, as shown in Fig. 3A, the sulcal t-fMRI/rs-fMRI signals have been found to have more diverse temporal patterns as well as higher frequency than gyral signals at the whole-brain level (Zhang et al., 2018a; Ge et al., 2019; Liu et al., 2019), which has been validated by an independent wavelet entropy analysis (Liu et al., 2019). A recent study (Jiang et al., 2020) further found that gyral/sulcal signals could be differentiated within each brain region, and that gyral/sulcal temporal patterns have different frequency scale characteristics across different brain regions.

Figure 3:

Summary of gyro-sulcal functional differences. (A) Gyro-sulcal functional activity time-frequency difference. The average power spectrum of gyral and sulcal temporal patterns is shown in the seven t-fMRI data sets from the HCP, which is adapted from Liu et al. (2019). (B) Gyro-sulcal functional connectivity/interaction difference. The left sub-figure shows the gyro-sulcal functional connectivity (black edges) within the primary motor and somatosensory system. Thicker edges represent stronger connections. The middle sub-figure shows the gyro-sulcal functional connectivity patterns of one example participant. The red,  white,  green, and blue bubbles represent the ROI on PCG, POG, CS, and POS, respectively. The edges are colored based on the functional connectivity strength. The left and middle sub-figures are adapted from Deng et al. (2014). The right sub-figure shows the graph-theory-based property difference of gyro-sulcal functional connectivity, which is adapted from Liu et al. (2017). (C) Gyro-sulcal functional network spatial overlap pattern difference. The static gyro-sulcal overlap pattern difference across multiple tasks is adapted from Jiang et al. (2015a). The temporal dynamics of gyro-sulcal overlap pattern difference in Emotion task from the HCP is adapted from Jiang et al. (2018a). (D) Gyro-sulcal functional signal representation accuracy difference. The gyro-sulcal signal representation residual is shown across the eight fMRI from the HCP, which is adapted from Zhao et al. (2020).

Gyro-sulcal functional connectivity/interaction differences

Functional connectivity (FC) is used to characterize functional integration and is defined as statistical dependencies among remote brain regions (Friston, 2011). One of the most commonly used measures of FC is the Pearson correlation coefficient (PCC), which calculates the temporal correlation of fMRI BOLD signals between voxels or brain regions (Fox and Raichle, 2007). A larger coefficient value represents stronger FC, and vice versa.

Deng et al. (2014) performed one of the earliest FC studies to investigate gyro-sulcal FC differences. They first focused on the primary motor and somatosensory systems as a test bed example, since these systems are proven to be structurally and functionally connected (Kandel et al., 2000). It has been found that within the primary motor and somatosensory systems, including four gyral regions (bilateral precentral gyrus and postcentral gyrus) and four sulcal regions (bilateral central sulcus and postcentral sulcus), the rs-FC strength is strong between gyro-gyral regions, moderate between gyro-sulcal regions, and weak between sulco-sulcal regions (Fig. 3B). Deng et al. (2014) further performed whole-brain analysis based on the A-DICCCOL (anatomy-guided dense individualized and common connectivity-based cortical landmarks) system (Jiang et al., 2015b), which defined 555 cortical landmarks on 68 major gyral/sulcal regions in the whole cerebral cortex, and obtained a similar gyro-sulcal rs-FC difference. In another study (Liu et al., 2017), the task-based FC was calculated using the PCC of t-fMRI temporal patterns between any pair of gyral/sulcal components, and found to be stronger in gyro-gyral than in sulco-sulcal regions in the whole brain across the seven different tasks of the HCP. In addition to investigating static FC by averaging scan-length rs-fMRI/t-fMRI time series (Deng et al., 2014; Liu et al., 2017), the temporal-varying dynamic FC differences between gyri and sulci have also been explored (Yang et al., 2019). The FC dynamic temporal variability within or between brain regions has been extensively demonstrated (Gilbert and Sigman, 2007; Chang and Glover, 2010; Smith et al., 2011; Calhoun et al., 2014; Li et al., 2014; Ou et al., 2014; Yuan et al., 2018). In Yang et al., 2019, the rs-fMRI BOLD signals were divided into non-overlapping time windows, and the temporal-varying FC dynamics of each gyral/sulcal region assessed by means of calculating the variance of the PCC of BOLD signals between any pair of time windows, to allow differences between gyri and sulci to be examined. It has been found that the dynamic temporal variance of the rs-fMRI signal was larger in gyral than in sulcal regions in the whole brain (Yang et al., 2019). Instead of assessing the FC between gyro-sulcal regions directly (Deng et al., 2014; Liu et al., 2017; Yang et al., 2019), another study explored the differential FC of gyri/sulci with other brain regions (Uddin et al., 2010). It has been found that the angular gyrus and the intraparietal sulcus, which are a part of the inferior parietal lobule, have differential rs-FCs with other brain regions (Uddin et al., 2010). Other measures have also been used to characterize functional interaction by quantifying the temporal correlation of fMRI signals. (Li et al., 2013) adopted a cross wavelet transform to measure the functional interaction between two regions. The functional interaction strength was defined as the locations where both signals shared high co-power in the time-frequency domain, and the gyro-sulcal difference was then obtained (Li et al., 2013).

Based on the FC/functional interaction network among gyral/sulcal regions, some studies have further adopted graph theory-based properties (Bullmore and Sporns, 2009; Rubinov and Sporns, 2010) to investigate the functional integration/segregation of such FC/functional interaction networks as well as their differences between gyri and sulci (Li et al., 2013; Deng et al., 2014; Liu et al., 2017). First, the FC/functional interaction networks were constructed within gyral/sulcal/mixed (both gyral and sulcal) brain regions, respectively. Second, certain graph theory-based properties such as edge degree, global efficiency, modularity, small-worldness, etc. were measured and compared among those networks. It has been found that the resting state functional interaction network composed of gyral regions has higher global and local economical properties, and stronger small-worldness and functional interaction strength than the sulcal network (Li et al., 2013). Within the rs-FC network, the graph edge degree of the gyral regions is significantly higher than that of the sulcal regions (Deng et al., 2014). Liu et al.(2017) have found that the FC network composed of gyral regions has higher average cluster coefficients and global efficiency than those of sulcal regions in both rs-fMRI and the seven t-fMRI data (Fig. 3B). In terms of modularity property, the sulcal network is significantly higher than the gyral one in t-fMRI but not rs-fMRI data (Liu et al., 2017).

Gyro-sulcal functional network spatial overlap pattern differences

The functional network spatial overlap pattern is defined as the set of brain regions that are spatially overlapped and involved in multiple brain networks. Previous studies have demonstrated extensively that there are multiple spatially overlapped brain functional networks that are interacting with each other (Bullmore and Sporns, 2009; Lv et al., 2014, 2015). The functional network spatial overlap pattern thus represents the higher-order brain regions that are involved in multiple cognitive tasks.

Sparse coding has been widely used for rs-fMRI and t-fMRI analysis to infer brain functional characteristics including the functional network spatial overlap pattern (Abolghasemi et al., 2015; Lv et al., 2014, 2015). The whole-brain fMRI BOLD signals are denoted as a matrix X = [x₁, x₂, …, x_n] ∈ ℝ^{t × n}, in which x_n is the fMRI BOLD signal of voxel n normalized to zero mean and standard deviation of 1 (Mairal et al., 2010), t is the length of the fMRI signal. By applying the online dictionary learning/sparse coding approach (Mairal et al., 2010), X is represented as

(1)

where D = [d₁, d₂, …, d_m] ∈ ℝ^{t × m} (m denotes dictionary size, m > t and m<<n) is an over-complete dictionary matrix, α = [α₁, α₂, …, α_n] ∈ ℝ^{m × n} is a sparse coefficient matrix, and ε is error term. Each fMRI signal x_i is represented as = D × α_i + ε. Note that each column of D (d_m) is a basis functional temporal pattern which is representative of whole-brain fMRI BOLD signals, and each row of α represents the spatial pattern of the functional network that corresponds to the basis temporal pattern. The meaningful functional network spatial overlap patterns can then be effectively identified (Zhang W. et al., 2019).

Using sparse coding approaches, Jiang et al. (2015a) characterized task-based heterogeneous functional regions (THFRs), which represented those higher-order brain regions involved in multiple cognitive tasks, based on all identified task-evoked and intrinsic functional network patterns. It has been found that both task-evoked and intrinsic connectivity functional networks can be effectively and robustly identified across the seven different t-fMRI of HCP grayordinate data (Jiang et al., 2015a). Moreover, the THFRs are located more on gyral than on sulcal regions within the bilateral parietal lobe, frontal lobe, and visual association cortices across different t-fMRI data (Fig. 3C). To further investigate the temporal dynamics of THFRs [i.e. spatial overlap patterns of functional networks (SOPFNs)] on gyri/sulci, SOPFNs and their spatial distribution proportion on gyri/sulci were identifed based on t-fMRI data within each sliding time window (Jiang et al., 2018a). It has been shown that the SOPFNs are also located more on gyral than on sulcal regions within bilateral parietal lobe, frontal lobe, and visual association cortices within each time window, and it is of interest that there is variability in the SOPFNs distribution percentage between gyri and sulci across different time windows (Fig. 3C). Specifically, the SOPFNs distribution percentage value of gyral regions was increased and achieved a peak value during the task blocks of block design t-fMRI, while that of sulcal regions was decreased and reached a trough value during the task blocks. When the task paradigm was changed from a task block to a resting state block, the SOPFNs distribution percentage was decreased from the peak for gyral regions and increased from the trough for sulcal regions, and achieved relatively the same percentage before starting the next task block (Jiang et al., 2018a). These findings were also consistent across the seven different block design t-fMRI of HCP grayordinate data (Jiang et al., 2018a).

Gyro-sulcal functional signal representation accuracy differences

The fMRI signal representation accuracy is defined as the PCC between the original signal x_i and the represented signal based on Equation (1) (Jiang et al., 2018b). Since sparse coding approaches can compactly represent the original fMRI BOLD signals using the identified small number of basis functional temporal patterns and corresponding functional network patterns, a larger PCC value indicates better representation of x_i as well as better participation of x_i in the neural function during a specific task.

Using nine intrinsic functional networks including default mode, visual, motor, auditory, executive control, and bilateral frontal/parietal networks (Smith et al., 2009), Jiang et al. (2018b) assessed and compared the signal representation of emotion processing t-fMRI data between gyral and sulcal regions within each of the nine networks. It has been found that during the emotion processing, gyral regions have better signal representation accuracy than sulcal regions within the nine intrinsic functional networks. In another recent study (Zhao et al., 2020), the signal representation accuracy was defined as the residual between the original signal x_i and the represented signal . Inversely, a smaller value of the residual indicates better representation of x_i. The signal representation accuracy was further assessed and compared between gyral and sulcal regions within the nine intrinsic functional networks under both the resting state and seven different assessment tasks including assessment of emotions, gambling, language, motor skills, relational skills, social skills, and working memory from the HCP. The variability magnitude of such gyro-sulcal differences was also assessed across different tasks/resting state and in intrinsic functional networks (Fig. 3D). It has been found that gyral regions consistently have better signal representation accuracy than sulcal regions within all nine intrinsic functional networks across all seven different t-fMRI of HCP grayordinate data. Moreover, the magnitudes of such gyro-sulcal differences vary across resting state/tasks and intrinsic functional networks. Specifically, in the resting state and "low-demand" tasks such as motor tasks, the gyro-sulcal signal representation accuracy contrast values are lowest and more consistent within the nine intrinsic networks as well as the whole-brain level across different participants. On the contrary, the gyro-sulcal contrast is higher in “high-demand" tasks, such as language and social and working memory, and less consistent across different participants. Moreover, the gyro-sulcal signal representation accuracy contrast is more pronounced within the default mode and executive control networks compared to the other intrinsic functional networks in all seven tasks, while it is the lowest within the sensorimotor network across all tasks (Zhao et al., 2020).

A Dynamic Functional Model of Gyri and Sulci

We have previously proposed a functional model of gyri and sulci at the whole-brain level (Deng et al., 2014) (Fig. 4B) and within functional networks (Jiang et al., 2018b) (Fig. 4D): individual gyri exchange information with other distant gyri through long-range fibers and with neighboring sulci through short-range fibers, while individual sulci exchange information directly with neighboring gyri through short-range fibers and indirectly with other regions through the neighboring gyri.

Figure 4:

Proposed functional models of gyri and sulci in previous studies. (A) The gyri/sulci defined in the whole-brain; (B) the proposed functional model of gyri and sulci at the whole-brain scale; and (C) an example of a motor functional network with two ROI. Each ROI consists of gyri and sulci. (D) The proposed functional model of gyri and sulci within functional networks.

However, this model does not explicitly explain how the interplay between gyri and sulci or their resulting dynamic interactions occurs. In this review, we would therefore like to propose a dynamic functional model of gyri and sulci as illustrated in Fig. 5, in which the involvement of gyri and sulci in global functional networks is dynamically balanced analogous to an activator–inhibitor system (Turing, 1952; Meinhardt and Gierer, 1974; Nakamasu et al., 2009; Yagi, 2009). The gyro-sulcal functional differences summarized in the section of 'Gyro-Sulcal Functional Differences from Various Perspectives' and Figs 3 and 4 from various perspectives are the foundation of this dynamic model. Compared with sulci, gyri have less diverse temporal patterns as well as lower frequency of fMRI signals, higher static or dynamic resting state/task-based FC, different graph-theory-based properties (higher global/local economical properties, graph edge degree, average cluster coefficient, and global efficiency; stronger small-worldness, and functional interaction), a greater distribution of THFRs and SOPFNs, and a higher signal representation accuracy. In the dynamic model in Fig. 5, when the task in a block design t-fMRI is loaded onto brain networks the gyri are activated. The SOPFNs distribution percentage of gyri begins to increase and achieves a peak value analogous to an activator since gyri are hypothesized to be the global functional center, while that of sulci begins to be inhibited to decrease and reach a trough value (Fig. 5B). When the task load is removed and the next resting state block starts, the excitatory and inhibitory effects vanish. The SOPFNs distribution percentage of gyri begins to drop from the peak point, while that of sulci begins to increase from the trough, until achieving relatively the same SOPFNs distribution percentage by the end of the resting state block (Fig. 5A). Such a gyro-sulcal dynamic balancing procedure then repeats when the next task block starts (Jiang et al., 2018a). The excitatory and inhibitory effects can be regulated by strong gyro-gyral functional connections and weak gyro-sulcal ones, analogous to an activator-inhibitor regulation system (Turing, 1952; Meinhardt and Gierer, 1974; Nakamasu et al., 2009; Yagi, 2009), underpinned by long-range and short-range axons, respectively, as illustrated in the model in Fig. 4.

Figure 5:

(A) and (B) An illustration of the dynamic functional model of gyri and sulci, at resting state (A) and during task performance (B). Two ROI as well as their connections are used to represent the whole-brain network. Areas of the gyri and sulci within each ROI illustrate the degree of involvement in global networks. The thickness of the double-head arrows illustrate the excitatory (thick) and inhibitory (thin) effects, putatively regulated by functional connections. (C) The degrees of involvement in global networks of gyri (red) and sulci (blue) during a task performance (Jiang et al., 2018a).

As a substrate for the dynamic model, long-range connections and short-range connections are needed to coincide in the axonal system. We have discussed the presence of long-range association projections in the section of 'Neuronal processes and gyro-sulcal functional differentiation', which are usually radial to the cortical surface of two remote gyri. The presence of a short-range axonal system, the connections of which are between two neighboring gyri and superficial to and parallel with the cortical surface, has also been demonstrated (Bullmore and Sporns, 2009; Sepulcre et al., 2010; Reveley, et al., 2015; Yoshino et al., 2020). Although the regulation mechanisms of long- and short-range axons in the dynamic model are still far from being clear, some studies have shown some progress toward achieving this goal. The functional network segregation is suggested to be driven by the abundance of short-range axons (Cao et al., 2017) while the long-range ones are involved in global information integration (Gao et al., 2009). A precise balance between the decreased short-range axon numbers and increased long-range axon numbers from 2 to 7 years in humans (Ouyang et al., 2017) has been shown to be positively associated with the reconfiguration of a developmental functional connectome from segregation to integration (Fair et al., 2009). A disturbance of this balance could result in brain disorders, such as hyperconnectivity for ASD and hypoconnectivity for schizophrenia (SZ) (Ouyang et al., 2017).

It is noteworthy that the excitatory and inhibitory effects might be implemented by the diffusion of functional gradients, as suggested in activator-inhibitor models (Turing, 1952; Meinhardt and Gierer, 1974; Nakamasu et al., 2009; Yagi, 2009). In fact, a global gradient has been observed in the cortex from a variety of perspectives ranging from genes to behavior (Huntenburg et al., 2018). Similarly, a gyro-sulcal gradient could also exist, although it might be overwhelmed by the global one. Indeed, a gradient between gyro-sulcal functional differences has been evidenced. During the dynamic balancing procedure of gyro-sulcal functions, the SOPFNs distribution percentage of gyri/sulci is increased or decreased in a gradient manner along with the task/resting state blocks instead of an abrupt change (Jiang et al., 2018a). Moreover, the gyro-sulcal pattern is associated with the functional localization gradient (Li et al., 2015; Troiani et al., 2020). For instance, it was found that the theoretically proposed rostral-caudal gradient of secondary-primary rewards is consistent with the spatial location of different rewards along a rostral-caudal gradient in the medial orbital sulcus (Li et al., 2015).

The gyro-sulcal gradient could be deeply rooted by a genetic protomap. In fact, a global scale predominant direction of gene expression gradients in the cortex has been widely reported (Miller et al., 2014; Burt et al., 2018; Huntenburg et al., 2018; Fulcher et al., 2019; Li et al., 2020b). In particular, distant cortical regions could also show similar gene expression if they have long-range connections (Richiardi et al., 2015; Krienen et al., 2016). In this sense, it could be inferred that cortical areas sharing a similar structural property could show similar gene expression, which would result in a gene expression gradient along the given structural property. Provided that gyri and sulci are two basic anatomical units of cortical morphology, the existence of a gyro-sulcal gradient is predicted, although overwhelmed by the global diffusion. This inference has gained partial support from several studies where differential expression of a great number of genes between gyri and sulci have been observed (de Juan Romero and Borrell, 2015; Zeng et al., 2015; Li et al., 2020). Some of gene differences already exist in the germinal layers, providing a protomap with abrupt gradients to locations of gyri and sulci that develop later (de Juan Romero and Borrell, 2015).

Future Research Directions

Definitions of gyri and sulci

As demonstrated in the section of 'Definition of gyri and sulci', gyri and sulci are usually treated as morphological and anatomical nomenclature. However, it is not clear how to define their morphological boundary. For example, if the zero level of convexity in FreeSurfer is used as a strict gyro-sulcal boundary, gyri below it would be defined as sulci. In this respect, it is difficult to define a clear boundary, if any, between gyri and sulci. Also, a group of gyri, termed pli de paset a (Gratiolet, 1854; Parent, 2014), were found to be buried in the depth of sulci and interrupting the continuity of the sulci, challenging the conventional definition of gyri, as being in continuity with each other across the entire cortex. Because of their specialty, pli de passage are suggested as benchmarks for cortical folding development modeling (Regis et al., 2005; Mangin et al., 2019) and a linkage between gyral and sulcal systems.

In this review, we have discussed the possibility of a mapping between gyro-sulcal patterns and functions and suggest that these two anatomical patterns could serve as basic blocks for a complex building of brain functions. Although conventional definitions of gyri and sulci are used in these studies, it is thought provoking that definitions could be revised by integrating axonal wiring and function information (Zhu et al., 2013; Jiang et al., 2015b; Lv et al., 2014, 2015; Zhang et al., 2018b, 2019a, b). For example, Jiang et al. (2015b) proposed 555 cortical landmarks on major cortical regions with gyro-sulcal anatomical and fiber connection correspondences across different participants. Zhang et al. (2018b, 2019a, b) further combined functional, structural, and anatomical information to identify cortical landmarks across participants. More effort needs to be devoted in the future to systematically investigate the gyro-sulcal definitions as well as their influences on functional models of gyri and sulci.

Gyro-sulcal framework is far from being sufficient

Another important question that needs to be asked is, are gyri and sulci the ultimate basic units for deciphering cortical morphology and anatomo-function relation, or can they be sub-divided? In fact, a group of studies (Lohmann et al., 2008; Im et al., 2010, 2011; Meng et al., 2014) have particularly focused on the deepest local regions of sulci, termed sulcal pits, the spatial placements of which remain invariant during brain growth. This invariance is closely related to intellectual ability (Im et al., 2011) and a protomap of functional areas (Lohmann et al., 2008; Im et al., 2010). In our studies, we have defined a novel gyral folding pattern (Li K. et al., 2010), termed the gyral hinge, which is the conjunction of multiple gyri from different directions. Gyral hinges are distinguished from other gyral regions because they have the thickest cortices, the strongest long-range axonal connections, the most aggregative functional profiles (Li K. et al., 2010; Yu et al., 2013; Chen et al., 2014; Jiang et al., 2015a, 2018a; Li et al., 2017; Ge et al., 2018), and comprise a majority portion of the cortico-cortical network's "core" (Zhang et al., 2020). Taken together, these studies indicate that classic gyro-sulcal patterns could be further sub-divided to meaningful units, such as ordinary sulci, ordinary gyri, sulcal pits, and gyral hinges.

Importance and necessity of considering gyro-sulcal functional differences in brain disorder studies

Size, morphology, and spatial patterns of cortical folding have been related to various brain functions (Barkovich et al., 2012; Fernández et al., 2016). Abnormalities and changes in cortical folding have been reported to lead to severe brain malfunctions and problems in intellectual and cognitive functioning (Barkovich et al., 2012; Fernándezet al., 2016). Some congenital cortical abnormalities, such as lissencephaly, microcephaly, and polymicrogyria, provide extreme examples of associations with severe anatomical pathology and mental deficits (Richman et al., 1974; Richman et al., 1975; Stutterd and Leventer, 2014; Di Donato et al., 2017: for a full review of these findings, see Fernández et al., 2016). Moreover, mild folding abnormalities, in developmental psychiatric disorders such as ASD and SZ, are also associated with cognition and behavioral impairments (White et al., 2003; Nordahl et al., 2007; Csernansky et al., 2008; Blanken et al., 2015; Dierker et al., 2015; Ecker et al., 2015). In ASD, these include deficits in social communication, social reciprocity, and repetitive and stereotyped behaviors and interests (Wing, 1997), and in attention, memory, social withdrawal and executive functions in SZ (Gur et al., 2000; Goldberg, 2003). Patients with ASD have also been reported to exhibit polymicrogyria, schizencephaly, and microgyria in the cortex (Piven et al., 1990). In particular, some sulci, such as the Sylvian fissure, are elongated (Levitt et al., 2003) and the gyrification of the frontal lobe is increased (Hardan et al, 2004). Similarly, in SZ patients, widespread abnormalities in cortical folding patterns have also been reported (Narr et al., 2004; Arnold et al., 2005) including in the temporal lobes (Kikinis et al., 1994) and frontal, occipital, and parietal cortices (Shenton et al., 2001; Narr et al., 2004).

A number of studies have focused on gyro-sulcal patterns with the aim of providing a more precise anatomo-functional mapping of potential contributions to mental disorders (reviewed by Nakamura et al., 2020). These studies have developed an approach for extracting gyro-sulcal patterns according to their spatial topology. In these studies, a set of H-shaped sulci in the orbito-frontal cortex (OFC) was of major interest because of its linkage with cognitive and social deficits in a variety of brain disorders (Ongur and Price, 2000). The OFC gyro-sulcal pattern has been classified into three types (Chiavaras and Petrides, 2000—type I, type II, and type III in order of frequency). In general, in type I the lateral orbital sulci (LOS) are continuous while the medial orbital sulci (MOS) are interrupted; in type II, both LOS and MOS are continuous; and in type III, both LOS and MOS are interrupted. It was found that the most common type I expression in healthy controls was decreased while the least common type III expression in healthy controls was increased in SZ (Nakamura et al., 2007). Similar atypical distributions of the gyro-sulcal patterns have been reported in ASD (Watanabe et al., 2014), attention deficit/hyperactivity disorder (ADHD), and bipolar disorder (BD) (Patti and Troiani, 2018).

However, most of those studies only focus on general morphological measures such as the gyrification index, cortical area, etc., while gyro-sulcal functional difference characteristics have received far less attention in existing human behavior and brain disorder studies. Conventional cognitive, psychological, and psychiatric studies usually select a large brain region of interest (ROI) (e.g. prefrontal cortex) based on an available brain atlas that neglects gyro-sulcal functional differences within it. As a consequence, subsequent ROI-based analyses (e.g. activation detection, FC analysis) are therefore inaccurate and may lose useful findings or generate false positive ones, especially when performing between-group comparisons. Since a number of mental disorder studies have demonstrated the possibility that gyro-sulcal topological patterns could be used as a proxy for brain functions and cognition (Nakamura et al., 2020; Troiani et al., 2020), it may be more important to use gyro-sulcal functional patterns as neurodevelopmental markers for future brain disorder diagnosis, cognitive, psychological, and psychiatric studies to obtain more accurate and reliable findings.

Author contributions

X.J.: conceptualization, visualization, writing of the original draft, and funding acquisition. T.Z.: conceptualization, visualization, writing of the original draft, and funding acquisition; S.Z.: conceptualization, writing of the original draft, and funding acquisition; K.M.K.: writing; review and editing, and funding acquisition; T.L.: supervision, conceptualization, and writing, review and rditing.

Conflict of interests

One of the authors, K.K., is also the editor-in-chief of Psychoradiology. He was blinded from reviewing or making decisions on the manuscript. The authors declare no other conflict of interests.