Structuro‐Functional Differentiation and Coupling of Gyri and Sulci in the Neonatal Cortex

ABSTRACT

Brain structure and function undergo rapid development during the neonatal period. Investigating neonatal brain structure, function, and their relationship is therefore crucial for understanding how the brain matures into a complex structuro‐functional system. As fundamental anatomical units of the cerebral cortex, gyri and sulci provide novel and valuable insights for such investigations. However, gyro‐sulcal differentiation and their structuro‐functional developmental relationship in neonates remain poorly explored. To address this gap, we used multi‐modal MRI data (structural T2w, diffusion‐weighted, and resting state functional MRI) from 438 neonatal brains in the public dHCP dataset. We systematically examined differences in functional connectivity (FC) and structural connectivity (SC) between gyri and sulci from 38 to 44 weeks postmenstrual age, as well as their FC‐SC coupling characteristics. From 38 to 44 weeks, both FC and SC were consistently strongest between gyro‐gyral regions and weakest between sulco‐sulcal regions, demonstrating that gyri act as global information processing hubs while sulci serve as local functional units in the neonatal brain. FC‐SC coupling exhibited distinct patterns across cortical lobes and over time, with a characteristic shift from coupling to decoupling around 41 weeks in most regions. This study provides a foundation for understanding early developmental mechanisms of brain structure–function relationships and establishes a normative reference of gyro‐sulcal differentiation as well as FC‐SC coupling in the neonatal period. These findings may inform future investigations of atypical neurodevelopment and contribute to the identification of early biomarkers for neurodevelopmental disorders.

The cortical gyro/sulcal regions are defined in each neonatal brain based on multi‐modal MRI images, and both structural and functional connectivity matrices among gyro‐sulcal regions are constructed and adopted to unveil the connectivity strength differences and coupling characteristics between gyri and sulci in neonatal brains.

1. Introduction

Brain structure, function, and cognition in the neonatal period differ significantly from those in the fetal, adolescent, and adult stages of development (Dubois et al. 2014; Fletcher et al. 2018; Gilmore et al. 2018; Hüppi 2010; Lebel and Deoni 2018; Li et al. 2015). For instance, graph theory‐based studies of structural and functional connectivity have reported a less developed neonatal cortical connectome compared to the more complex infant cortical connectome (Batalle et al. 2017; Cao et al. 2017). At the same time, functional networks such as the visual, auditory, somatosensory, motor, frontoparietal, and executive control networks in the neonatal cortex, have been found to resemble those in the adult cortex (Doria et al. 2010; Fransson et al. 2009; van den Heuvel et al. 2015), yet their maturation rates differ across networks during the first year after birth (Gao et al. 2015, 2009). More importantly, the neonatal period represents a critical window of brain development, characterized by rapid growth and significant structural and functional reorganization (Kostović and Jovanov‐Milosević 2006; Makropoulos et al. 2016). Although most neurons have reached their final positions by this time (Fenchel et al. 2020), key processes such as synaptogenesis (Kostović et al. 1995; Petanjek et al. 2011), residual neuronal migration (Paredes et al. 2016), and the formation of short‐range white matter fiber connections (Kostović and Jovanov‐Milosević 2006; Kostović et al. 2019) remain actively ongoing. Furthermore, in populations with perinatal risk factors‐particularly preterm infants‐alterations in brain connectivity development can establish a substrate for atypical neurodevelopment outcomes (Batalle et al. 2018). Therefore, studying the neonatal cerebral cortex provides a valuable platform for investigating the origins of subsequent behavioral and cognitive abilities (Fenchel et al. 2020), understanding how the brain develops into a complex structural and functional system (Cao et al. 2016), and identifying early alterations and abnormalities that may precede neurodevelopmental disorders (Di Martino et al. 2014; Morgan et al. 2018).

Structural connectivity (SC) and functional connectivity (FC) are two fundamental metrics in neuroimaging research that characterize the physical wiring and dynamic interactions of brain networks, respectively. SC is typically defined as the number of white matter fiber tracts connecting two brain regions which are estimated by diffusion tractography (Zhu et al. 2013). FC, commonly defined as the Pearson correlation coefficient between fMRI BOLD signals of two brain regions (Fox and Raichle 2007), reflects the degree of functional integration or segregation among them (Friston 2011). Together, SC and FC offer complementary perspectives for investigating brain structure and function (Biswal et al. 1995; Fan et al. 2016; Matyi and Spielberg 2021; Saarela et al. 2007), as well as the neural basis of brain disorders and functional deficits (Caligiuri et al. 2017; Christodoulou et al. 2012; Fornito et al. 2012; Gibbard et al. 2018; Lynall et al. 2010; Maximo et al. 2014). Importantly, SC and FC are not independent; rather, SC shapes and constrains FC to some extent. A growing body of research has focused on elucidating this relationship, demonstrating that SC and FC can effectively predict one another (Goñi et al. 2014; Honey et al. 2009, 2010; Zhang et al. 2022). Both strong SC and strong FC have been observed in gyral regions (Deng et al. 2014), and several studies have characterized the spatial alignments between the two modalities (Hermundstad et al. 2013; Mišić et al. 2016; Shen et al. 2012; Uddin 2013), which has been formalized as FC‐SC coupling. Studies have revealed that FC‐SC coupling varies across brain regions and differs between healthy controls and clinical populations, including individuals with cognitive impairment (Wang et al. 2018), Alzheimer's disease (Dyrba et al. 2015), depression (Bi et al. 2016), and idiopathic generalized epilepsy (Zhang et al. 2011). Crucially, FC‐SC coupling is not uniformly strong across the whole brain (Gu et al. 2021) but follows a gradient along the cortical hierarchy. It is stronger in lower‐order sensory‐motor regions and weaker in higher‐order cognitive regions (Preti and Van De Ville 2019). That is, coupling predominantly characterizes unimodal areas, while decoupling is a hallmark of transmodal areas (Valk et al. 2022; Vázquez‐Rodríguez et al. 2019). By capturing this differential relationship between structure and function, the framework of FC‐SC coupling/decoupling provides a more comprehensive understanding of the neurobiological mechanisms linking brain structure, dynamics, and cognition than examining SC or FC in isolation. Consequently, characterizing FC‐SC coupling patterns during neurodevelopment is of critical importance.

Beyond the dynamical evolution of brain structure and function, cortical folding, which is one of the most prominent features of the human cerebral cortex, also exhibits pronounced time‐dependent variation. Given the established relationship of cortical folding with brain structure, function, cognition, and behavior (Jiang et al. 2021), investigating its development in neonates provides critical insights into normative neurodevelopmental trajectories and the ways in which genetic, environmental, and pathological factors can influence brain maturation.

Previous studies have established that convex gyri and concave sulci, which are the two fundamental and distinct units of cortical folding, can serve as key determinants for characterizing the structural and functional architecture of the brain (Deng et al. 2014; Jiang et al. 2015; Liu et al. 2019). Indeed, gyri and sulci exhibit significant differentiation across multiple biological domains, including gene expression (de Juan Romero et al. 2015), cytoarchitecture (Hilgetag and Barbas 2005; Rakic 2009), white matter fiber connectivity (Deng et al. 2014; Nie et al. 2012; Xu et al. 2010), and functional organization (Jiang et al. 2015, 2018; Wang et al. 2023). For instance, even before cortical expansion and folding begin, discrete modules with distinct genetic expression profiles within germinal layers may help specify gyral and sulcal formations (de Juan Romero et al. 2015). Mechanistically, an FGF‐driven positive feedback loop that increases astrocyte proliferation has been identified as an important driver of cortical folding (Matsumoto et al. 2017; Shinmyo et al. 2022), and differential neurogenesis between gyri and sulci further contributes to the formation of cerebral folds (Borrell 2019; Chun et al. 2025; Lin et al. 2025). At the cellular level, a greater number of neurons populate gyri than sulci in deep cortical layers V and VI (Hilgetag and Barbas 2005). At the connectivity level, fiber densities are significantly higher in gyral compared to sulcal regions (Nie et al. 2012). FMRI BOLD signals in gyral regions exhibit less diverse temporal patterns and a lower frequency profile than those in sulcal regions (Liu et al. 2019). It is now widely accepted that the brain does not function through isolated modules but rather as an interconnected network of interacting regions (Dosenbach et al. 2007; Leitgeb et al. 2020; Raichle 2015) which seeks to explain the emergence of cognition and behavior (Bassett and Gazzaniga 2011; Cole and Schneider 2007; Power et al. 2011; Sporns 2013). Given the pronounced gyro‐sulcal contrast described above, it logically follows that these structures likely play distinct roles within brain networks. Indeed, previous studies have systematically reported SC and FC differences between gyral and sulcal regions. In one of the earliest investigations, Deng and colleagues reported that both FC and SC are strong among gyri, weak among sulci, and moderate between gyri and sulci, both within primary motor and somatosensory systems and across the entire cortex (Deng et al. 2014). Task‐based FC analysis has further demonstrated that FC is stronger between gyro‐gyral than between sulco‐sulcal regions across all seven tasks of the Human Connectome Project (HCP) (Liu et al. 2017). Moreover, time‐varying FC differs among gyro‐gyral, gyro‐sulcal, and sulco‐sulcal regions (Yang et al. 2019). Collectively, these findings suggest a model in which gyri act as functional hubs, exchanging information both with distant gyri and neighboring sulci, while sulci communicate directly with adjacent gyri and indirectly with other cortical regions through these gyral intermediaries.

However, while these insights into gyral and sulcal roles are profound, they have been derived almost exclusively from studies examining SC or FC in isolation. Despite clear evidence for SC and FC contrasts between gyri and sulci, significant ambiguities, uncertainties, and even disagreements persist. For example, gyri/sulci pairs with comparable FC strength exhibit marked differences in SC strength (Deng et al. 2014). Furthermore, structural asymmetries often diverge from functional asymmetries (Jung et al. 2009; Wei et al. 2022). Consequently, the precise relationship between SC and FC at the gyro‐sulcal level remains largely unknown. Elucidating this relationship is critical for understanding both how SC shapes and constrains FC and how gyro‐sulcal functional differences emerge from their underlying structural substrates. Crucially, human cortical gyri and sulci develop from a relatively smooth cortical plate during early fetal stages. While most existing findings come from the mature brain with fully developed gyro‐sulcal anatomy, recent studies (Hamouda et al. 2026; Mao et al. 2024) have begun to investigate gyral and sulcal SC/FC patterns in the neonatal brain. Nevertheless, the developmental origins of these patterns as well as the nature of their mutual relationship during this critical period remain poorly understood.

Therefore, the present study aims to characterize SC and FC differentiation between gyral and sulcal regions, as well as SC‐FC coupling and decoupling patterns, in the developing neonatal brain. To this end, this study analyzes multimodal MRI data from the developing Human Connectome Project (dHCP), including T2‐weighted structural MRI (T2w‐MRI), diffusion‐weighted MRI (dMRI), and resting‐state functional fMRI (rs‐fMRI), to longitudinally examine gyro‐sulcal differences in FC, SC, and their coupling in neonates. This work will, for the first time, validate and extend the existing gyro‐sulcal functional differentiation hypothesis that is previously established in adult brains to the neonatal period, thus providing a foundation for understanding the early development of functional brain architecture.

2. Materials and Methods

2.1. Dataset and Preprocessing

We utilized 438 full‐term neonates from the public Developing Human Connectome Project (dHCP) (http://www.developingconnectome.org/) release 2–3 with appropriate authorization. The study was approved by the local Institutional Review Boards, and written informed parental consent was obtained prior to scanning. The principal inclusion criteria comprised pregnant women carrying fetuses at 20–42 weeks of gestational age, as well as live infants at 23–44 weeks of gestational age (https://www.developingconnectome.org/study‐inclusion‐and‐exclusion‐criteria/). The dataset for each subject includes T2w‐MRI, dMRI, and rs‐fMRI images. All included neonates were born at > 38 weeks of gestation and scanned between 38.14 and 44.71 weeks postmenstrual age. Birth and scan ages for the cohort are summarized in Table 1.

TABLE 1.

Distribution of birth and scan ages for the neonatal cohort (N = 438).

Group	Amount	Birth age (weeks)		Scan age (weeks)
		Range	Mean $\pm$ std	Range	Mean $\pm$ std
W38	25	38.00–38.71	38.31 $\pm$ 0.25	38.14–38.86	38.55 $\pm$ 0.23
W39	65	38.14–39.71	39.12 $\pm$ 0.34	39.00–39.86	39.41 $\pm$ 0.27
W40	96	38.00–40.86	40.00 $\pm$ 0.60	40.00–40.86	40.53 $\pm$ 0.28
W41	95	38.43–41.57	40.44 $\pm$ 0.95	41.00–41.86	41.44 $\pm$ 0.27
W42	68	38.29–42.14	40.71 $\pm$ 1.10	42.00–42.86	42.34 $\pm$ 0.30
W43	65	38.29–42.29	40.80 $\pm$ 0.84	43.00–43.86	43.41 $\pm$ 0.27
W44	24	38.86–42.00	40.70 $\pm$ 0.87	44.00–44.71	44.22 $\pm$ 0.19

Note: Neonates were assigned to weekly age groups based on their scan age: W38 (≥ 38 to < 39 weeks), W39 (≥ 39 to < 40 weeks), and so on up to W44 (≥ 44 to < 45 weeks).

T2w‐MRI images were acquired in sagittal and axial slice stacks with an in‐plane resolution of 0.8 × 0.8 mm² and 1.6 mm slices with 0.8 mm overlap. Acquisition parameters included TR/TE = 12,000/156 ms, with SENSE factors of 2.11 (axial) and 2.60 (sagittal). dMRI data were acquired across four shells (b0: 20, b400: 64, b1000: 88, and b2600: 128). Diffusion directions were temporally distributed to account for motion and duty cycle considerations. Additional parameters included SENSE factor 1.2, partial Fourier 0.86, acquired resolution of 1.5 × 1.5 mm², 3 mm slices with 1.5 mm overlap, and TR/TE = 3800/90 ms. Rs‐fMRI data were acquired with TR/TE = 392/38 ms, yielding 2300 volumes at an isotropic resolution of 2.15 mm. No in‐plane acceleration or partial Fourier was used. Cortical surfaces including curvature maps, region labels (32 cortical regions and 7 subcortical regions projected from T2w volumes), and other features are provided in the dHCP dataset following preprocessing steps that include skull stripping, tissue segmentation, and surface reconstruction. Detailed procedures are described in the dHCP T2w‐MRI pipeline (Makropoulos et al. 2018).

Preprocessed dMRI data (Bastiani et al. 2019) underwent eddy current distortion correction (Andersson and Sotiropoulos 2016) and mild erosion of the fractional anisotropy (FA) map using FSL (Andersson and Sotiropoulos 2016; Woolrich et al. 2009). Deterministic fiber tracking was performed using DSI Studio (Yeh et al. 2013) with the following parameters: FA threshold = 0.05 (Berman et al. 2009), turning angle = 55°, fiber length restricted to 10‐400 mm, and 6 × 10⁴ fibers per subject. T2w‐MRI volumes were registered to the dMRI FA map using FSL‐FLIRT (Jenkinson et al. 2002; Jenkinson and Smith 2001), generating a transformation matrix that was subsequently applied to align cortical surfaces to dMRI space. Rs‐fMRI preprocessing followed the pipeline described in Fitzgibbon et al. (2020). Processed rs‐fMRI volumes were then registered to T2w‐MRI space using FSL‐FLIRT (Jenkinson et al. 2002; Jenkinson and Smith 2001) and subsequently transformed to dMRI space using the same registration matrix that aligned T2w‐MRI to dMRI. BOLD signals were mapped to cortical surface vertices using nearest‐neighbor interpolation. Vertices without a nearest voxel were assigned signals via trilinear interpolation from neighboring vertices.

2.2. Definition of Gyri, Sulci, SC, and FC

A neonatal cortical atlas comprising 32 cortical regions (Tables S1 and S2), previously annotated by an expert neuroanatomist (Gousias et al. 2012), was registered to the T2w volume and projected onto the cortical surface (Figure 1A). For each cortical region, a curvature threshold $r$ was applied to delineate gyri and sulci. To assess the reproducibility of the results, multiple threshold values were tested. Vertices with curvature greater than $r$ were classified as gyral ROIs, while those with curvature less than $-r$ were classified as sulcal ROIs (Figure 1B). This procedure yielded 64 ROIs (32 regions × 2 gyral/sulcal types) across the cortical surface. Using these ROIs, a whole‐brain FC matrix (size: 64 × 64) was constructed, where each entry represented the average Pearson correlation coefficient of the rs‐fMRI signals between vertices of two ROIs (Figure 1D). An SC matrix was similarly constructed, with each entry defined as the proportion of total white matter fibers that connected two ROIs (Figure 1C). As half of the ROIs correspond to gyri and half to sulci, the FC and SC matrices were reorganized into gyro‐gyral, gyro‐sulcal, and sulco‐sulcal connection blocks, as illustrated in Figure 1E,F. The values in these matrices therefore represent FC and FC strengths between gyri and gyri, gyri and sulci, and sulci and sulci. In addition to these 64 ROIs, cortical lobes were also used as ROIs for constructing connectivity matrices.

FIGURE 1.

Flowchart of the study pipeline. The workflow consisted of three main stages. Left panel (A, B): structural parcellation. A white matter surface was generated from T2w‐MRI and parcellated into 32 cortical regions based on an established neonatal atlas. Each region was further subdivided into one gyral and one sulcal ROI, yielding 64 ROIs in total. An example from the superior temporal (ST) region is illustrated. Middle panel (C, D): connectivity mapping. (C) Structural connectivity (SC): Fibers were reconstructed from dMRI. SC strength between two ROIs was defined as the number of connecting fibers normalized by the total number of whole‐brain fibers. (D) Functional connectivity (FC): BOLD signals from rs‐fMRI were mapped onto the cortical surface. FC strength between two ROIs was defined as the mean Pearson correlation coefficient of the signals between the two ROIs. Right panel (E, F): matrix construction and analysis. 64 × 64 FC and SC matrices were constructed, with ROIs grouped by gyral and sulcal type. These matrices served as the basis for subsequent analyses, including statistical comparison of FC and SC strengths, modeling of developmental trajectories across 438 neonates, and FC‐SC coupling analysis.

2.3. Fitting and Coupling of FC and SC

Previous studies have demonstrated a close association between FC and SC, with each being predictive of the other (Goñi et al. 2014; Honey et al. 2009, 2010; Zhang et al. 2022). Furthermore, the developmental trajectory of FC exhibits a nonlinear relationship with postmenstrual age at scan (Sanders et al. 2023). To investigate group‐level developmental trends across the 38–44 week postnatal period, the Generalized Additive Mixed Model (GAMM) (Lin and Zhang 2002) was employed to fit FC and SC strengths as functions of scanning age for all 438 neonates. For each element of the connectivity matrices, FC or SC strength was modeled as the dependent variable, with postmenstrual age at scan as the independent variable. The 438 data points were used to estimate a developmental curve for FC or SC. The goodness of fit was assessed using the $R^2$ statistic, defined as:

$$Q={\sum }_i{\left(y_i-{\stackrel{`}{y}}_i\right)}^2$$ (1)

$$R^2=1-{\left(\frac{Q}{{\sum }_i{y}_i^2}\right)}^{0.5}$$ (2)

where $y_i$ is the observed FC or SC strength of the i‐th subject, and ${\stackrel{`}{y}}_i$ is the corresponding fitted value.

The simultaneous increase or decrease of FC and SC during development was conceptualized as FC‐SC coupling, which was quantified as the ratio of the derivatives of the fitted FC and SC curves:

$$\text{Coupling}=d{f}_{\mathrm{FC}}/{\mathit{df}}_{\mathrm{SC}}$$ (3)

where $f_{\mathrm{FC}}$ and $f_{\mathrm{SC}}$ are the fitted curves, and $d$ is the first derivative with respect to age. The coupling value represents the instantaneous relationship between FC and SC over time, with positive values indicating coupling (synchronized change) and negative values indicating decoupling (opposing change).

3. Results

3.1. Gyro‐Sulcal FC and SC in Different Brain Regions

FC and SC strengths were computed for gyro‐gyral, gyro‐sulcal, and sulco‐sulcal pairs, both across the whole brain and within five cortical lobes. The t‐tests with FDR correction were performed to assess significance of differences in connectivity strengths among the three connection types. Results obtained using cortical surface curvature thresholds of 0.1, 0.15, and 0.20 were highly consistent. Therefore, a threshold of 0.15 was applied in all subsequent analyses. Because SC strength was defined as proportional to the number of white matter fibers connecting two ROIs, raw SC values were influenced by ROI size: ROIs with larger surface areas tended to possess higher fiber counts and thus higher SC strengths. To enable meaningful comparisons across connection types, SC strengths in Figure 2 were proportionally normalized. Specifically, for each ROI pair type (gyro‐gyral, gyro‐sulcal, and sulco‐sulcal), SC strengths were divided by the sum of strengths across all three types within the same ROI or lobe (Figure 2, SC row). This normalization preserves the relative proportions among connection types while removing absolute scale differences. In all subsequent analyses beyond Figure 2, SC strengths were used without the normalization.

FIGURE 2.

FC and SC strengths in the whole brain and five cortical lobes. Vertices with curvature greater than threshold r were classified as gyri, while those with curvature less than −r were classified as sulci. (A–C) show mean connectivity strengths across 438 neonates using curvature thresholds of 0.1, 0.15, and 0.2, respectively. For visualization across lobes, SC strengths were normalized by the proportional contribution of gyro‐gyral, gyro‐sulcal, and sulco‐sulcal connections within each ROI. Statistical significance of differences among the three connection types was assessed using a t‐test with FDR correction. FDR‐corrected *p < 0.05, **p < 0.01, and ***p < 0.001.

As shown in Figure 2B, both FC and SC in the whole brain exhibited a consistent hierarchy: strongest for gyro‐gyral connections, moderate for gyro‐sulcal connections, and weakest for sulco‐sulcal connections. This pattern was consistent in the temporal, limbic, frontal, and parietal lobes. The occipital lobe, however, showed a distinct FC pattern. While gyro‐gyral FC remained strongest, gyro‐sulcal FC was weakest, and sulco‐sulcal FC was slightly but consistently higher than gyro‐sulcal FC. In contrast, SC in the occipital lobe followed the whole‐brain hierarchy (i.e., gyro‐gyral > gyro‐sulcal > sulco‐sulcal). A two‐way ANOVA examining FC strengths across connection types and lobes revealed significant main effects of both folding type (gyro‐gyral, gyro‐sulcal, sulco‐sulcal; p < 0.001) and lobe (p < 0.05), indicating that both factors contribute independently to FC differentiation.

We next characterized the longitudinal trajectories of FC and SC strengths from 38 to 44 weeks postmenstrual age. Each neonate was assigned to a weekly age bin by rounding their scanning age down to the nearest integer week (e.g., neonates scanned at ≥ 38 to < 39 weeks were assigned to the W38 group). For each week, mean and standard deviation of FC and SC strengths were calculated separately for gyro‐gyral, gyro‐sulcal, and sulco‐sulcal connections across the whole brain and within each lobe (Figure 3). For the whole‐brain trajectories, both FC and SC consistently exhibited the same hierarchical pattern across the entire 38–44 week period: strongest for gyro‐gyral connections, moderate for gyro‐sulcal connections, and weakest for sulco‐sulcal connections (Figure 3A). This pattern remained stable across all weekly time points. For the lobar trajectories, the whole‐brain pattern was consistent in most lobes. Specifically, in the temporal, frontal, and parietal lobes (Figure 3B,E,F), both FC and SC maintained the gyro‐gyral > gyro‐sulcal > sulco‐sulcal hierarchy across the 38–44 week period. In the limbic lobe (Figure 3C), FC showed the expected hierarchy only from 38 to 43 weeks. In the occipital lobe (Figure 3D), SC remained consistent with the whole‐brain pattern only from 39 to 44 weeks. Notably, the differentiation among gyro‐gyral, gyro‐sulcal, and sulco‐sulcal connections was more pronounced in the frontal and parietal lobes compared to other regions. In the limbic lobe (SC) and occipital lobe (FC), no significant differences among the three connection types were observed at any individual week, despite overall patterns across the full period.

FIGURE 3.

Developmental trajectories of FC and SC strengths from 38 to 44 weeks in the whole brain and five lobes. For each weekly age bin (W38–W44), means ± SD of connectivity strengths are shown for gyro‐gyral (blue), gyro‐sulcal (gray), and sulco‐sulcal (red) connections. (A) Whole brain. (B) Temporal lobe. (C) Limbic lobe. (D) Occipital lobe. (E) Frontal lobe. (F) Parietal lobe. Statistical significance among connection types was assessed using a t‐test with FDR correction. FDR‐corrected * p < 0.05, **p < 0.01, and ***p < 0.001.

3.2. Coupling and Decoupling of FC and SC

Given the well‐established relationship between brain structure and function (Honey et al. 2010) and our observation of both convergent and divergent developmental trajectories of FC and SC (Figure 3), we next quantified FC‐SC coupling strengths across the 38–44 week period. For this analysis, FC and SC strengths represent the average of gyro‐gyral, gyro‐sulcal, and sulco‐sulcal connections within each region.

Figure 4 displays the fitted developmental trajectories of FC and SC along with their coupling curves for the whole brain and each lobe. The whole brain (Figure 4A) exhibited a pattern that was broadly consistent in the temporal, frontal, and parietal lobes (Figure 4B,E,F): FC increased from 38 weeks onward, peaked around 42 weeks, and then declined, while SC increased monotonically throughout the entire period. Consequently, FC and SC were coupled during the initial phase (38–42 weeks) and became decoupled after 42 weeks. The limbic lobe (Figure 4C) showed a distinct pattern: FC initially increased then decreased, while SC initially decreased then increased, resulting in predominantly decoupled development. In contrast, the occipital lobe (Figure 4D) exhibited relatively concordant FC and SC trajectories, with coupling sustained throughout most of the period. To quantitatively characterize these patterns, we defined the coupling rate as the proportion of time during which FC and SC were coupled (i.e., coupling value > 0) relative to the total period. Coupling rates confirmed predominantly coupled development in the whole brain (0.56), temporal lobe (0.59), occipital lobe (0.81), frontal lobe (0.52), and parietal lobe (0.54), while the limbic lobe showed predominantly decoupled development (0.13).

FIGURE 4.

FC and SC developmental trajectories and their coupling in the whole brain and five lobes. For each panel, the left and middle show fitted FC and SC developmental curves from 38 to 44 weeks. The right shows the coupling curve, defined as the ratio of FC and SC derivatives (see Methods). Positive coupling values indicate synchronized change (coupled development), while negative values indicate opposing change (decoupled development). The horizontal line at zero marks the transition between coupling and decoupling. (A) Whole brain. (B) Temporal lobe. (C) Limbic lobe. (D) Occipital lobe. (E) Frontal lobe. (F) Parietal lobe.

To examine spatial heterogeneity in FC‐SC coupling, we further computed coupling rate ${\text{Rate}}_c$ for each of the 32 cortical regions and visualized them on the cortical surface (Figure 5). Based on their coupling rates, regions were classified into three categories: coupling < decoupling (${\text{Rate}}_c<$ 0.45), 11 regions; coupling $\approx$ decoupling ($0.45\le {\text{Rate}}_c<$ 0.55), 8 regions; coupling > decoupling (${\text{Rate}}_c\ge$ 0.55), 13 regions. These results demonstrate substantial regional heterogeneity in FC‐SC developmental patterns, indicating that the relationship between structure and function during the neonatal period is not uniform across the cortex.

FIGURE 5.

Regional FC‐SC coupling rates across the cortical surface. Coupling rates, defined as the proportion of time (38–44 weeks) during which FC and SC were coupled, are projected onto the white matter surface. Warmer colors indicate higher coupling rates (predominantly coupled development), while cooler colors indicate lower coupling rates (predominantly decoupled development).

4. Discussion

4.1. Structuro‐Functional Differentiation of Gyri and Sulci in Neonates

Gyri and sulci have been proposed to serve as global and local functional communication centers, respectively, with established structural and functional differentiation in the adult brain. Previous studies have demonstrated that in adults, both FC and SC are strongest among gyri, weakest among sulci, and intermediate between gyri and sulci (Deng et al. 2014). Task‐based FC (Liu et al. 2017) and time‐varying FC (Yang et al. 2019) analyses have similarly reported stronger FC in gyral than sulcal regions, while fiber density is significantly higher in gyral regions (Li et al. 2015). Our findings demonstrate that this fundamental gyro‐sulcal differentiation pattern is already present in the neonatal brain. Across the whole brain and within individual lobes, FC and SC strengths consistently followed the adult‐like hierarchy: gyro‐gyral > gyro‐sulcal > sulco‐sulcal. This pattern remained stable across each weekly time point from 38 to 44 weeks postmenstrual age. These results support the view that gyri function as global communication hubs, exchanging information directly with remote gyri and neighboring sulci, while sulci communicate directly with adjacent sulci and indirectly with distant gyri via long‐range fiber connections.

Consistent with adult studies reporting regional variation in gyro‐sulcal differentiation (Zhang et al. 2019), we observed significant heterogeneity across lobes and regions in the neonatal cortex. However, our results also revealed important developmental differences between neonates and adults. Notably, although the limbic and occipital lobes numerically follow the gyro‐gyral > gyro‐sulcal > sulco‐sulcal pattern, the differences among the three connection types did not reach statistical significance (Figure 3). This finding aligns with a previous study reporting significantly higher fiber density in gyral compared to sulcal regions within frontal, parietal, and temporal lobes, but not in limbic and occipital lobes (Li et al. 2015).

Longitudinally, SC strength exhibited sustained increases across most lobes from 38 to 44 weeks, with the notable exception of the limbic lobe (Figure 3C). This regional variation likely reflects differential maturation rates across cortical areas. Previous studies have identified several resting‐state networks (RSNs) that exhibit adult‐like topology by term‐equivalent age (Chugani et al. 1987), with rapid weekly maturation observed in primary RSNs (medial motor, somatosensory, visual) and association RSNs (motor association, posterior parietal, frontal parietal, visual association). The more pronounced gyro‐sulcal differentiation we observed in frontal and parietal lobes similar as in adult brains is consistent with the relatively advanced maturation of these regions by the neonatal period. Conversely, the limbic lobe, which showed transient or non‐significant gyro‐sulcal differentiation, may follow a distinct developmental trajectory that extends beyond the neonatal window examined here.

4.2. Structuro‐Functional Coupling and Decoupling in Neonates

Brain function is closely linked to its structural architecture, with white matter fibers facilitating inter‐regional communication that underpins cognition. In the adult brain, the alignment between FC and SC, which is referred to as FC‐SC coupling, is consistently reported (Gu et al. 2021; Hermundstad et al. 2013; Mišić et al. 2016; Shen et al. 2012; Uddin 2013). Given the rapid development of the neonatal brain, we conceptualized the co‐development of FC and SC as a form of developmental coupling and investigated its trajectory from 38 to 44 weeks postmenstrual age.

We observed a progressive shift from structuro‐functional coupling to decoupling between 38 and 44 weeks, evident globally in the whole brain and regionally in the temporal, frontal, occipital, and parietal lobes. During this period, FC followed an inverted‐U trajectory which increased initially and then decreased, while SC increased monotonically. This asynchronous developmental timing resulted in an initial phase of coupling (i.e., synchronized change) followed by a later phase of decoupling (i.e., divergent trajectories). Due to cortical immaturity, both functional and structural connectivity patterns undergo continuous reorganization during this period, lacking the dynamic stability characteristic of adult brains.

Several complementary mechanisms may explain the decoupling observed after approximately 41 weeks. First, axonal arborization after birth occurs primarily within cortical plate layers, accompanied by preferential growth of short‐range cortico‐cortical connections that predominantly link sulci, rather than expansion of long‐range fiber tracts (Kostović and Jovanov‐Milosević 2006; Kostović et al. 2019). This pattern of axonal growth may drive the sustained increase in SC strength without immediately enhancing brain activity. Indeed, significant increases in cortical activation only emerge around 3 months postnatally, when most cortical regions demonstrate substantial functional maturation (Chugani et al. 1987). Second, while several primary and association RSNs already exhibit adult‐like topological organization by term‐equivalent age, this period also coincides with the termination of long‐range fiber growth (Cao et al. 2017; Dubois et al. 2014; Lebel et al. 2019) and the disappearance of axonal growth‐associated molecules. Interhemispheric fibers in the corpus callosum cease developing, and abundant active callosal axons begin to retract (Kostović and Jovanov‐Milosević 2006). These structural reorganizations may temporarily interrupt the increase in FC strength, contributing to the observed decoupling. Third, adult studies demonstrate that stronger FC‐SC coupling typically characterizes unimodal regions, likely reflecting their need for rapid, stimulus‐driven response mediated by direct structural connections (Margulies et al. 2016; Huntenburg et al. 2017; Cioli et al. 2014; Hawrylycz et al. 2012). During the neonatal period, cognitive and perceptual abilities remain immature (Johnson 2001); for example, infants typically cannot reach a target object until approximately 9 months of age. At this stage, developing brain functions may rely less on structural fibers, resulting in decoupling. Fourth, previous work has reported that SC couples with lower‐frequency brain activity but decouples with higher‐frequency activity (Preti and Van De Ville 2019). Notably, the dHCP fMRI acquisition used a TR of 392 ms, enabling capture of higher‐frequency neural activity compared to the conventional 0.01–0.1 Hz band typically studied. This methodological consideration may have enhanced our sensitivity to detect decoupling.

The transition from coupling to decoupling occurred at approximately 41 weeks in multiple brain regions. This timing aligns with independent evidence from neonatal functional imaging studies. Gao and colleagues reported that functional individual variability, which was measured by amplitude of low‐frequency fluctuations (ALFF), fractional ALFF, and regional homogeneity, increased with age and then decreased from around 41 weeks (Gao et al. 2022). Similarly, studies in term‐born neonates have shown that global dynamic features of functional connectivity remain relatively stable between 37 and 44 weeks, with lower metastability observed with increasing postnatal days, suggesting that extra‐uterine life promotes more stable connectivity patterns (França et al. 2024). These convergent findings suggest a potential shift in functional development trajectories around 41 weeks, coinciding with the decoupling we observed. Concurrently, local short‐range white matter fibers continue to strengthen, supported by diffusion MRI studies showing that microstructural properties of cortico‐cortical tracts, such as fractional anisotropy and mean diffusivity, exhibit clear increases toward term‐equivalent age, reflecting enhanced white matter maturation and improved efficacy of long‐range neuronal signal transmission (Berman et al. 2005; Bonifacio et al. 2010; Hüppi et al. 1998; Neil et al. 1998; van den Heuvel et al. 2015). These microstructural changes influence the topology of macroscopic structural brain networks, with studies demonstrating increased integration capacity between 30 and 40 weeks (Dosenbach et al. 2007; Fair et al. 2008; van den Heuvel et al. 2015).

Our regional analysis (Figure 5) revealed substantial variation in coupling rates across the cortex, reflecting that functional activity, fiber connections, and cortical expansion differ between gyri and sulci with significant regional specificity. The limbic lobe presented a notable exception to the global pattern, with FC and SC remaining coupled throughout most of the period and exhibiting only a brief decoupling. This distinct trajectory arose because both FC and SC in the limbic lobe initially increase before subsequently decreasing, rather than following the monotonic SC increase and inverted‐U FC pattern observed elsewhere. This finding aligns with reports that limbic lobe surface area and cortical thickness show weaker associations with age compared to other lobes, which exhibit positive age‐related associations (Fenchel et al. 2020). This may explain why SC growth is not sustained in the limbic lobe. Furthermore, limbic lobe fibers develop early, with tract formation beginning before the second trimester (Huang et al. 2006; Takahashi et al. 2012), potentially explaining why its SC plateaus or decreases during the neonatal period rather than continuing to increase.

4.3. Limitations and Future Work

Several limitations of the present study should be acknowledged, along with corresponding directions for future research. First, the neonatal cortical surface was segmented into 32 regions based on the dHCP atlas labels. This parcellation may not be sufficiently fine‐grained, particularly in the occipital, frontal, and parietal lobes, which are simply divided into left and right hemispheres. Future studies employing higher‐resolution parcellation schemes could enable more precise characterization of regional developmental trajectories and gyro‐sulcal differentiation. Second, defining an optimal curvature threshold r for delineating gyri and sulci remains challenging, given individual differences and the complexity of cortical folding patterns particularly in the neonatal brain. Although we employed data‐driven sensitivity analysis to evaluate robustness across multiple thresholds, future work could explore optimization‐based approaches to automate threshold selection in a more principled manner. Third, an important extension of this work will be to investigate associations between gyro‐sulcal connectivity patterns and subsequent cognitive or behavioral outcomes. While the present study characterizes normative developmental trajectories, linking these features to later neurodevelopment could establish their predictive validity and clinical relevance. Fourth, recent advances in cellular and molecular neuroscience have produced high‐resolution, region‐ and age‐specific transcriptomic atlases of the developing human neocortex (Dear et al. 2024; Jiang et al. 2023). Parallel longitudinal MRI studies have begun linking rapid neonatal cortical microstructural maturation and emerging functional organization to underlying gene expression profiles (Natu et al. 2021; Wang et al. 2025; Zhao et al. 2024). However, the present work remains limited to neuroimaging measures and does not directly incorporate cellular or molecular data, constraining mechanistic interpretation of the observed patterns. Future studies integrating multimodal neuroimaging with genetic and transcriptomic information could provide biologically grounded interpretations of early structural and functional brain organization and their developmental trajectories.

Lastly, longitudinal neonatal MRI studies have demonstrated that brain imaging measures at early or term‐equivalent age possess significant predictive value for later adverse neurodevelopmental outcomes. Specifically, moderate‐to‐severe white matter abnormalities, diffusion‐derived microstructural alterations in major white matter pathways, and early cortical morphometric features (e.g., cortical thickness, sulcal depth, gray matter volume) have been consistently associated with increased risk of later cognitive and motor impairments (Pagnozzi et al. 2023; Rose et al. 2015; Setänen et al. 2013; Woodward et al. 2006). Quantitative radiomics features extracted from conventional T1‐ and T2‐weighted images have demonstrated prognostic utility even in the absence of overt white matter injury, highlighting the potential relevance of subtle neonatal imaging markers (Shin et al. 2021; Wagner et al. 2022). Despite this progress, the predictive capacity of neonatal gyro‐sulcal structural and functional characteristics for specific psychiatric or neurodevelopmental disorders remains insufficiently explored (Jiang et al. 2024; Hagan et al. 2025). This gap is noteworthy given that substantial alterations in cortical folding patterns have been widely documented in children, adolescents, and adults diagnosed with autism spectrum disorder, schizophrenia, and depression (Csernansky et al. 2008; Depping et al. 2018; Dierker et al. 2015; Hardan et al. 2004; Kikinis et al. 1994; Peng et al. 2015; White et al. 2003). With the increasing availability of longitudinal datasets spanning from the neonatal period into later life, future studies will be well positioned to directly test whether the neonatal cortical connectivity features identified in the present work are predictive of disease‐related cortical abnormalities observed in adulthood. Such investigations could establish early‐life markers of neurodevelopmental vulnerability and inform targeted intervention strategies.

Author Contributions

Wei Mao: methodology, data curation, formal analysis, software, visualization, writing – original draft. Zhibin He: methodology, data curation, investigation, writing – original draft. Xuewei Jin: data curation, visualization, writing – review and editing. Elmehdi Hamouda: writing – review and editing. Xun Ou: writing – review and editing. Keith M. Kendrick: writing – review and editing. Tuo Zhang: conceptualization, funding acquisition, project administration, supervision, writing – review and editing. Xi Jiang: conceptualization, funding acquisition, project administration, supervision, writing – review and editing.

Funding

This work was supported by National Natural Science Foundation of China (62276050, 62576077, 62476222, 62131009) and Noncommunicable Chronic Diseases National Science and Technology Major Project (2023ZD0500901).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Index and name of 32 neonatal brain regions.

Table S2: Categorization of 32 brain regions by lobes.

Data Availability Statement

The original data used in this study is obtained from the Developing Human Connectome Project (dHCP). Access to the data requires registration via the dHCP website (http://www.developingconnectome.org/project/) and approval by the dHCP administrators. Part of the preprocessed data and results is available upon request.