Geometric growth of the normal human craniocervical junction from 0 to 18 years old

Abstract

The craniocervical junction (CCJ) forms the bridge between the skull and the spine, a highly mobile group of joints that allows the mobility of the head in every direction. The CCJ plays a major role in protecting the inferior brainstem (bulb) and spinal cord, therefore also requiring some stability. Children are subjected to multiple constitutive or acquired diseases involving the CCJ: primary bone diseases such as in FGFR‐related craniosynostoses or acquired conditions such as congenital torticollis, cervical spine luxation, and neurological disorders. To design efficient treatment plans, it is crucial to understand the relationship between abnormalities of the craniofacial region and abnormalities of the CCJ. This can be approached by the study of control and abnormal growth patterns. Here we report a model of normal skull base growth by compiling a collection of geometric models in control children. Focused analyses highlighted specific developmental patterns for each CCJ bone, emphasizing rapid growth during infancy, followed by varying rates of growth and maturation during childhood and adolescence until reaching stability by 18 years of age. The focus was on the closure patterns of synchondroses and sutures in the occipital bone, revealing distinct closure trajectories for the anterior intra‐occipital synchondroses and the occipitomastoid suture. The findings, although based on a limited dataset, showcased specific age‐related changes in width and closure percentages, providing valuable insights into growth dynamics within the first 2 years of life. Integration analyses revealed intricate relationships between skull and neck structures, emphasizing coordinated growth at different stages. Specific bone covariation patterns, as found between the first and second cervical vertebrae (C1 and C2), indicated synchronized morphological changes. Our results provide initial data for designing inclusive CCJ geometric models to predict normal and abnormal growth dynamics.

This research investigates the growth dynamics of the craniocervical junction (CCJ) in children. By compiling geometric models of normal skull base growth and analysing closure patterns of synchondroses and sutures, it reveals distinct developmental trajectories and covariations between CCJ bones.

1. INTRODUCTION

Approximately 35% of children are born with morphological anomalies affecting the skull, jaws, ears, and/or teeth (Tavares & Moody, 2022), including constitutional bone diseases and acquired conditions leading to deformations and growth impairment. In clinical examination of the cranio‐cervical junction (CCJ), it is thus essential to distinguish diseases causing primary malformations from diseases leading to secondary instability. For instance, constitutional bone diseases including FGFR‐related craniosynostoses can be associated with CCJ deformities with impaired growth (Huggare, 1995), without clear predictive factors or genotype–phenotype correlations. Similarly, acquired conditions may lead to osteoligamentary and neurological anomalies, with different clinical challenges than in structural deformities. For instance, cervical spine luxation resulting from traumatic injuries, such as whiplash during a car accident (Guéz et al., 2004). All CCJ disorders can be linked with functional issues, including central sleep apnoea, discharge disorders, cerebrospinal fluid (CSF) obstruction, and, in extreme cases, sudden death (Pauli et al., 1984), as for instance in achondroplasia (Smid et al., 2020). Without appropriate management, CCJ anomalies can lead to severe and definitive neurological damage (Chirossel et al., 2000; Flanagan, 2015a, 2015b).

The natural history of CCJ anomalies in FGFR‐related craniosynostoses is little understood and predictive parameters are critically missing to propose patient‐specific treatment plans. In fact, these patients have highly variable CCJ phenotypes, with trajectories varying from spontaneous stabilization to severe instability (Strahle et al., 2011). We know imaging plays an important role in diagnosis (Sargar et al., 2017). And while surgery is often intended to address the morphological anomalies, predicting the outcome of these treatments remains a significant challenge because very little is known about the growth dynamics of this region. Currently, anticipating the evolution of the CCJ in patients remains challenging as little is known about the factors determining its shape and as quantitative approaches remain scarce (but see Massimi et al., 2021; Urbančič et al., 2023).

Head and neck growth involves many bones, starting before birth and continuing through postnatal stages (Meyer‐Marcotty et al., 2018). The occipital bone ossifies from multiple centres, including the basioccipital, lateral parts, squamous part, and left and right condylar ossification centres, which gradually fuse, leading to its final shape, forming the posterior and lower aspects of the skull (Bernard et al., 2015; Srivastava, 1992). The temporal bones ossify from multiple centres, including the petro‐occipital synchondrosis, petrosquamous suture, tympanic part, and mastoid process ossification centres (Grzonkowska et al., 2023; Madeline & Elster, 1995). They contribute to critical skull components (Eby & Nadol, 1986; Yokoyama et al., 2009). The ethmoid and sphenoid bones also develop from multiple ossification centres. The ethmoid bone forms the eye sockets and facilitates olfactory nerve passage, with structures like the perpendicular plate and cribriform plates playing key roles (Krmpotić‐Nemanić et al., 1997, 1998). The sphenoid bone contributes in the construction of the central skull base (Jin et al., 2016; Nie, 2005). The main ossification centres of the sphenoid bone include the presphenoid, basisphenoid, and lateral parts. These centres gradually fuse to give rise to the sub‐parts of the sphenoid bone, such as the major wings, lesser wings, and the body. The ethmoid bone undergoes ossification through various centres, including those responsible for forming its main perpendicular plate, orbital regions, and cribriform plates, ultimately culminating in the formation of the fully developed bone (Abdumuminova et al., 2024).

Postnatally, cranial vault and base growth continues through bone remodelling and bone deposition at sutures and synchondroses, accompanying the underlying expansion of the brain and the eyeballs, determining the overall shape and size of the skull throughout childhood and adolescence (Zoetis et al., 2003). This intricate process ensures the proper protection and support for the developing brain while accommodating the sensory organs and facial structures.

Sutures, the fibrous joints between cranial bones, are instrumental in a process called intramembranous ossification, where mesenchymal cells differentiate into osteoblasts without transiting via a cartilaginous matrix. These osteoblasts at the borders of sutures deposit mineralized matrix, leading not only to the surface expansion of the skull vault bone and the gradual fusion of sutural regions but also to the expansion of the cranial base, for example the growth at the spheno‐frontal suture (Di Rocco et al., 2014; Opperman, 2000; White et al., 2021). Understanding the anatomy of sutures is crucial when studying cranial regions, as it aids in differentiating between normal developmental variations and potential trauma. Thorough examinations and descriptions of these sutures have already been done (Idriz et al., 2015; Leitch et al., 2020), emphasizing their importance in paediatric radiological assessments. On the other side, synchondroses are cartilaginous joints also playing an important role in the growth of the skull base. The spheno‐occipital synchondrosis is a crucial site where the sphenoid and occipital bones connect as it contributes to the development and fusion of ossification centres in the posterior part of the skull and plays an important role in defining the length of the skull base (Moss, 1975). In the spine, synchondroses are found in the vertebral epiphyses, where the growth plates facilitate longitudinal growth. Ossification in synchondroses involves endochondral processes, where a transient hyaline cartilage matrix is replaced by bone. The cartilage within the synchondrosis becomes calcified as osteoblasts invade the area, leading to the formation of bone tissue (McBratney‐Owen et al., 2008; Wei et al., 2017). Premature fusion of synchondroses mostly impairs sagittal skull base growth and can lead to midface hypoplasia (Funato et al., 2020; Melsen & Melsen, 1980) such as what is reported in FGFR‐related craniosynostoses and osteochondrodysplasia (Calandrelli et al., 2023). Synchondroses, although critical for understanding craniofacial growth, are often overlooked in quantitative studies. Further investigations are thus required to quantify the shapes of sutures and synchondroses, assess the impact of environmental factors and genetics, and understand the functional meaning of shape modifications such as the various degrees of radiological fusion (Di Ieva et al., 2013; Funato et al., 2020; Remesz et al., 2023).

The CCJ includes eight integrated elements: two vertebra and six bones from the skull base (Hallgrimsson et al., 2006; Goswami, 2006). Skull modularity refers to the concept that the skull is composed of distinct but interconnected regions, allowing for differential growth and development. This modular structure facilitates adaptability and specialization of skull parts for various functions. However, there is a significant level of integration across these modules to maintain the overall structural and functional cohesion of the skull (Zelditch & Goswami, 2021). The concept of integration refers to the degree of interaction between one or more anatomical structures (Olson & Miller, 1999). Covariation between specific regions of the vertebral column has been studied by focusing on the link between posture and locomotion (Manfreda et al., 2006; Meyer et al., 2018) as well as the relationship with the cranium (Nalley & Grider‐Potter, 2015). It was demonstrated that the brain and the skull are strongly integrated (Richtsmeier et al., 2006). But the integration within the craniocervical region screening for bone interactions and dynamics throughout growth remains to be investigated in details.

Several attempts have been made over the past few years to describe the relationships between growth impairment and morphological anomalies using growth models (Kreinces et al., 2022; and see Geoffroy et al., 2022 for a review), including statistical descriptions (geometric morphometrics) and biomechanical models (Malde et al., 2019). Geometric morphometrics extends beyond measurements of linear distances and angles, using various quantification techniques to provide a more detailed and multimodal understanding of changes in the size and shape of anatomical structures (Hallgrimsson et al., 2015; Strauss & Bookstein, 1982; Thompson, 1942).

In the study of the CCJ, geometric morphometrics helps capturing and analysing intricate morphological variations (Bapuraj et al., 2019; Karaaslan et al., 2019; Liang et al., 2023). This approach contributes to examine not only the overall dimensions and orientations of the CCJ but also focuses on specific regions, such as sutures and synchondroses, which are major zones in the growth process. In the literature, there is a noticeable grey zone in studies focusing on the statistical descriptions of cranial bones, with a predominant emphasis on facial growth and cranial vault development. Notably, the regions of the skull base and cranio‐cervical junction (CCJ), despite their critical importance, have received little attention so far. Additionally, quantitative data concerning cranial sutures and synchondroses are frequently underrepresented in geometric and biomechanical models.

To better understand the geometric growth of the CCJ in healthy children and in patients with craniofacial abnormalities, it is essential to establish a robust reference cohort and develop growth models (Russo & Smith, 2011; Sgouros et al., 1999). Here we attempted to better understand the growth of the CCJ by describing normal processes and by establishing a reference growth model for further comparisons with populations affected by craniofacial anomalies. The description includes the shape of bones, the patterns of sutures and synchondroses, the integration of bone structures, and the quantification of growth rates. We believe that this comparative approach will eventually facilitate the identification of markers associated with abnormal growth patterns.

2. MATERIALS AND METHODS

2.1. Clinical data

Computed tomography (CT) scans of healthy control children from both genders, performed at the paediatric Hôpital Necker – Enfants Malades (Paris, France), were included (Table 1). The inclusion criteria were an age range from 0 to 18 years, and an indication for CT‐scan in the context of a superficial infection or facial trauma/tumour without history of chronic disease. The stack of images had to encompass the two first cervical vertebra and skull base with complete cerebellar fossa and orbital roofs. The exclusion criteria included the presence of bone fractures or displacements. All scans were reviewed by two experienced surgeons (RHK and SB) and 37 patients were included. Because of the unequal quality of CT‐scans, one or several bones were discarded in several patients (Table S1).

TABLE 1.

Number of scans used for each bone analysis.

Bone	Number of scans
Occipital	37
Sphenoid	30
Temporal left	32
Temporal right	33
Orbitals	30
Ethmoid	34
C1	37
C2	37

2.2. Segmentation and post‐processing

The segmentation of the CT‐scan images was performed using BoneSplit (v.0.9.2 – Nyström et al., 2017) that allowed assisted manual segmentation of cranial bones separately, following the sutures. BoneSplit used a graph‐based segmentation algorithm to automatically segment bones from the painted regions. The algorithm constructed a graph of all the voxels in the CT image, and then assigned labels to the voxels based on their connectivity and similarity to known bone structures (Nysjö et al., 2015). The segmentation labels of all bones were then exported as 3D isosurfaces.

Post‐processing of the surface objects was performed in Geomagic Wrap (Artec, 2013) and included artefact removal, holes patching, remeshing, decimation, and smoothing. Due to the incomplete upper parts of the skull in some patients, occipital and frontal bones were cut following a horizontal plane (Figure 1a). The occipital plane was determined by placing two landmarks at both distal ends of the transversal sinus separating the cerebellar fossa from the cerebral fossa (Figure 1b). The orbital plane was determined by placing two landmarks on the highest left and right supraorbital points (Figure 1c). This resulted in occipital surfaces encompassing the foramen magnum up to the upper ridge of the cerebellar fossa, and in frontal surfaces encompassing the orbital roofs and the orbital bandeau.

FIGURE 1.

(a) Lateral left view of the skull base bones in a 4‐month‐old subject. Occipital bone is represented in purple, the right temporal bone in dark blue, the left temporal bone in light blue, the sphenoid bone in orange, the orbitals in green and the ethmoid bone in yellow. 3D visualization after segmentation. (b, c) Cutting planes for the occipitals and the orbitals, respectively, that standardized the region of interest among scans.

2.3. Geometric morphometrics

2.3.1. Creation of the templates

Discontinuities in the 3D surfaces or missing data could prevent the manual or semi‐automatic positioning of landmarks (Brown et al., 2012). In children, because endochondral ossification was still ongoing, cartilaginous parts (particularly around the synchondroses of the cranial base and upper cervical vertebra) were not ossified yet, hence not visible on CT‐scans. To tackle this limitation and account for variations in these regions, we opted for a template‐based approach, consisting in designing standardized bone shapes, and their subsequent projection onto each target bone surface. The templates for the first two cervical vertebras were provided by Arts et Métiers – ENSAM, Paris.

Templates were considered for all bones except for the sphenoid and the ethmoid and were designed using Fusion360 (Autodesk, 2022) to roughly mimic the shape of bones, and triangular water‐tight meshes were generated using Blender (v.4.0.1 – Foundation Blender, 2023) (Figure 2). Due to the intricate three‐dimensional shape of the sphenoid, the template was created from a 3D model of a two‐year‐old patient included in our database (Figure 2.3). Due to its incomplete ossification and bone thinness (Belden et al., 1997) leading to unsatisfactory reconstruction, 11 landmarks were manually placed on the ethmoid using CheckPoint (Stratovan, 2023) (Figure 2.2). Three landmarks were placed on the midline, and 8 were placed on the skull base on the limits with the maxilla and the orbital region (Table S2).

FIGURE 2.

Initial templates. 1. Orbital region. 2. Location of the 11 landmarks on the ethmoid bone (Table S1 for a detailed description of their location). 3. Sphenoid bone. 4. Left temporal bone, 5. Right temporal bone. 6. Creation of discontinuities in the template shape at the location of synchondroses to mimic the pattern of synchondrosis closure of each real data bones on which the template will be projected. (a) General template of occipital bone with closed synchondroses. (b) Occipital with open anterior synchondroses. (c) Occipital with open anterior synchondroses and half‐open posterior synchondroses. (d) Occipital with open anterior and posterior synchondroses. 7.a. Template of C1 with closed synchondroses. (b) C1 template with open anterior arch (c) C1 template with open anterior and posterior arches. 8.a. Template of C2 with closed synchondroses. (b) C2 template with synchondrosis at the base of the dens, and at the posterior spine. (c) C2 template with synchondrosis at the junction between the base of C2 and the lateral masses, at the base of the dens, and at the posterior spine.

In younger patients (less than 1 year old), occipital and vertebral synchondroses were not always radiologically closed. Therefore, the templates applied to these patients were modified to account for synchondrosis opening without affecting the number of nodes. This resulted in additional versions of templates: 3 for the occipital (Figure 2.6), 2 for C1 (Figure 2.7), and 2 for C2 (Figure 2.8). First the global template with all closed sutures and synchondroses was made. Then, while keeping the exact same number and position of nodes, some surface triangles were cut to obtain an opening in the mesh at the anatomical location of sutures and synchondroses. This way, the open templates represent all the sutures and synchondroses combination while always ensuring the nodes are comparable for all.

2.3.2. Registration

The registration between target bone surfaces and corresponding template, further referred as wrapping, was achieved using R3DS Wrap v.2023.10.3 (Faceform, 2023). The wrapping process consisted in the mapping of the templates meshes onto the surface of the segmented bone meshes using Iterative Closest Point (ICP) rigid alignment followed by a Non‐rigid ICP (NICP). While the ICP aligned 3D surfaces in space with translation and rotation, the NICP iteratively deformed the template locally so that each of its node was moved closer to its closest counterpart on the target surface. Deformation was performed following Thin Plates Spline deformation that minimized bending energy. At convergence of the wrapping process, the template adopted the target shape. Ultimately, all wrapped bones were composed of the same number of nodes with the same topology. The 3D coordinates of the nodes were saved and further used as quantitative descriptors of bone shapes.

The wrapping process required the initial manual placement of landmarks for ICP and NICP: a set of 5–20 landmarks (Table S3) were placed in R3DS Wrap to initiate the wrapping process, and manual adjustments were performed to obtain alignment and coverage. For the younger patients, the ‘open’ templates were used on the first two vertebras and the occipital bone. An ‘open’ template represents a suture or synchondrosis fully open, so when it is applied to a half‐open bone configuration, the nodes of the template will place themselves in a half‐open configuration following the target surface. This way, only one ‘open’ template is needed for all the configurations of sutures and synchondroses instead of one for each age.

All 3D coordinates of the nodes of the wrapped bone meshes and of the landmarks placed manually on the ethmoid were saved, and a Generalized Procrustes Analysis (GPA) was performed, aiming at removing undesired variations in position, orientation, and scaling. The generated Procrustes coordinates were used as input shape data in all subsequent statistical analyses.

2.4. Shape analysis

All statistical analyses were performed in R (R Core Team, 2022). Alpha threshold index for rejecting null hypothesis was set at 0.05. GPA was performed using ProcSym in Morpho package (Schlager, 2017).

2.4.1. Growth trends

Two‐block Partial Least‐Squares (2b‐PLS) regressions were performed with the PLS2B function from the Morpho package to extract the covariance matrix between bone shape and log‐transformed age (log10 applied to age in days. Given the typical logarithmic distribution of shape component, this ensured a better fit of the model). Thin‐Plate Spline (TPS) with tps3d warping was applied to deform a reference 3D bone surface (scanlandmark) using the shape variations captured by the first and second components of the PLS regression (plsEffects). The resulting deformed bones were visualized using the shade3d function (Adams et al., 2024; Murdoch et al., 2024).

2.4.2. Growth rates

Twelve age categories were defined (Table 2). Intermediate bone shapes along the PLS model were generated for each age category. The meshdist function from the Morpho package was used to calculate the distances between bones at each age and compared with the theoretical shape corresponding to the age of 18 years. The distances represented the extent at which each stage spanned the remaining shape variation towards adult shape, hence providing growth rates. These rates were represented as gradients of colour and helped identifying areas of the bones where growth was more rapid, allowing to pinpoint regions that underwent the most significant changes with age.

TABLE 2.

Age categories.

Age category	In days	Year(s)
1	36–365	Birth to 1 year old
2	366–730	1–2 years old
3	731–1095	2–3 years old
4	1096–1460	3–4 years old
5	1461–1825	4–5 years old
6	1836–2190	5–6 years old
7	2191–2920	6–8 years old
8	2921–3650	8–10 years old
9	3651–4380	10–12 years old
10	4381–5110	12–14 years old
11	5111–5840	14–16 years old
12	5841–6545	16–18 years old

2.4.3. Sutures and synchondroses

Our template‐based approach accounted for the opening status of the synchondroses as the NICP enabled the projection of the open areas of the template onto the adjacent fragments of bone. To quantify opening along the synchondroses, the distance along the two wrapped fragments was calculated using meshdist function. The template nodes of interest involved in the synchondroses were retrieved in Blender. As synchondroses ossified, the distance between fragments decreased; the distance threshold for considering the synchondrosis locally closed was defined when the distance between adjacent fragments was lower than the distance typically found between two nodes in the surrounding mesh, considered homogenous (in Figure 3, blue lines, and distance between orange dots, respectively). The same procedure was applied to quantify suture closure between two adjacent bones.

FIGURE 3.

Closure condition for the sutures and synchondroses. The blue points represent the nodes on the extremity of the suture or synchondroses and the blue lines the distance between each side. The orange points represent the nodes surrounding the extremities of the suture or synchondrosis.

A second analysis was performed to quantify how closure occurred along the sutures and synchondroses. The calculated distances between mesh extremities at different vertices were filled in a distance matrix. The visualization of the suture in space was conducted using the rgl package (Murdoch et al., 2024), highlighting the extremities and their locations on both the template and the observed objects. Age categories were defined, and the average distance matrix for each category was calculated.

We performed regression analysis with the lm function for each age category, fitted b‐spline regression models (bs) and plotted the fitted curves (plot, barplot, lines functions). The percentage of closure was calculated for all age categories, and corresponding bar plots were created, including fitted regression lines. For each point along the suture, the distance between the nodes on each side of the separation was calculated. Of all these distances, the maximum one represents the location where the suture is the most open. We divided all the distances by this maximum distance and multiplied by 100 to obtain a relative percentage of the closure along the suture. This was done within each age category to see the evolution of closure. The bar that is at one is thus the bar representing the local node with highest distance.

2.4.4. Covariation patterns

We examined how the covariance between two scaled bones evolved with age. To do so, 2b‐PLS regressions (PLS2B function, Morpho package) were performed between pairs of scaled bones, in each age class. PLS regressions generate axes of maximum covariance between multidimensional matrices. Serving as a dimensionality reduction analysis, PLS generates multiple axes of covariation, each statistically tested to identify correlations between shape components. The coefficient of covariation quantifies the strength of these correlations, while the percentage of variance explains the proportion of total shape variance attributed to this interaction.

Because of the number of subjects in each age categories, we decided to create new categories for covariance analysis so that a minimum of 6 subjects are included in each category. The new distribution goes like this: 0–2 years old, 2–6 years old, 6–12 years old, and 12–18 years old.

3. RESULTS

3.1. Growth rates

The subjects were grouped by age categories (0–2 years, 2–4 years to 16–18 years). The PLS regressions between the shape and the log‐transformed age (log 10) indicated a covariation coefficient of 0.94 (p‐value of 0.001) for the occipital (Figure 4), 0.95 for the left temporal (Figure 5), 0.93 for the right temporal (Figure S2a), 0.90 for the orbital region (Figure 6), 0.94 for the sphenoid (Figure 7), 0.91 for the ethmoid (Figure 8), 0.93 for the atlas (Figure 9), and 0.95 for the axis (Figure 10).

FIGURE 4.

Occipital shape during growth. Growth of occipitals from given age to 18 years. Intermediate shapes with scaling, anterior view. Colour code: distance between each landmark of the occipital at each age category and the landmarks of the 18‐year‐old occipital, from ‐4 mm to +4 mm. Minimum and maximum mentioned on each occipital rendering for a better understanding of decrease of distances with age.

FIGURE 5.

Left temporal shape during growth. Growth of left temporals from given age to 18 years old. With scaling, lateral right view. Colour code: distance between each landmark of the left temporal at each age category and the landmarks of the 18‐year‐old temporal, from ‐10 mm to +6 mm. Minimum and maximum mentioned on each temporal bone rendering.

FIGURE 6.

Orbital region shape during growth. Growth of the orbital region from given age to 18 years old. With scaling, anterior view. Colour code: distance between each landmark of the orbital region at each age category and the landmarks of the 18‐year‐old orbital region, from ‐10 mm to +4 mm. Minimum and maximum mentioned on each orbital region rendering.

FIGURE 7.

Sphenoid shape during growth. Growth of sphenoid bone from given age to 18 years old. With scaling, posterior view. Colour code: distance between each landmark of the sphenoid bone at each age category and the landmarks of the 18‐year‐old sphenoid bone, from ‐11 mm to +5 mm. Minimum and maximum mentioned on each sphenoid bone rendering.

FIGURE 8.

Ethmoid shape during growth. Growth of ethmoid bone from given age to 18 years old. With scaling, left lateral view. Colour code: distance between each landmark of the ethmoid bone at each age category and the landmarks of the 18‐year‐old ethmoid, from ‐4 mm to +4 mm. Minimum and maximum mentioned on each ethmoid bone rendering.

FIGURE 9.

Atlas shape during growth. Growth of atlas from given age to 18 years old. With scaling, superior view. Colour code: distance between each landmark of the atlas at each age category and the landmarks of the 18‐year‐old atlas, from ‐5 mm to +5 mm. Minimum and maximum mentioned on each first cervical vertebra.

FIGURE 10.

Axis shape during growth. Growth of axis from given age to 18 years old. With scaling, posterior view. Colour code: distance between each landmark of the axis at each age category and the landmarks of the 18‐year‐old axis, from ‐5 mm to +5 mm. Minimum and maximum mentioned on each second cervical vertebra.

3.1.1. Occipital bone

By representing the distance between the 1‐month‐old and 18‐year‐old occipital bones, the shape of the fossa appeared in blue colour and the ridge between the fossa in dark orange colour. The same pattern was reported in the other age‐based comparisons in Figure 4 but with less and less importance because of the age getting closer to 18 years. The dynamics of shape modifications were slowing down with age: +/− 2 mm between 1 month old and 1 year vs. maximum change of 0.5 mm between 6 years and 10 years of age.

The global scale was defined based on the first comparison: distance between the first month of life and eighteen years of age. This scale was then used for all other distance comparisons, featuring the maximum and minimum of each independent comparison scale. The growth rates were defined similarly, decreasing with each yearly comparison.

The foramen magnum expanded, mostly within the first year of life. Concurrently, the clivus reached its final shape around the same time. The lateral parts of the occipital bone exhibited growth, particularly up to 2 years of age, after which the rate lowered, although persisting until around 6 years of age. From the age of 10 onward, the growth dynamics of the occipital bone became relatively uniform, with changes occurring at an approximate rate of 0.2 mm every 2 years until 18 years of age.

3.1.2. Temporal bones

In the first global comparison between 1 month of life and 18 years of age, the squamous part of the left temporal bone (Figure 5) increased its concavity while the petrous part expanded. The process was identical for the right temporal bone (Figure S2b). The zygomatic process underwent extension until the age of three, after which its length remained constant. The squamous part was characterized by the development of its distal ends that underwent expansion, while its central portion increased its concavity, mirroring the pattern observed in the mastoid process. The petrous part exhibited significant growth, expanding by 10 mm from 1 month of life to 18 years of age. During the first 2 years of life, one‐mm change occurred each year. Subsequently, between the ages of 3 and 10, the rate of change decreased to around 0.5 mm. From 10 to 18 years of age, the growth slowed down even more, with changes occurring at a rate of 0.2 mm every year.

3.1.3. Orbital region

The presence of the metopic suture was reported, delineating the left and right parts of the frontal bone during first few years of life (Figure 6). A visible transformation occurred in the nasal part of the orbital region, exhibiting growth of approximately 7 mm from the first month of life to 18 years of age. Substantial changes occurred during initial 6 years of life, characterized by a yearly reduction in scale by 1 mm. Subsequently, between the ages of 10 and 14, the rate of change diminished to 0.5 mm every 2 years. From 14 to 18, the growth further stabilized, with alterations occurring at a reduced rate of 0.3 mm every 2 years.

3.1.4. Sphenoid bone

In sphenoid bone, the expansion was mostly located in the major wings (Figure 7), which underwent significant development, expanding by 11.0 mm from the first month of life to 18 years of age. The pterygoid process experienced its most substantial development during the first month of age. From 2 years onward, the global scale of changes diminished. However, between 2 and 6 years of age, there remained a steady yearly increase of 1.0 mm for the whole sphenoid bone. Subsequently, from 6 to 12 years old, the growth rate decreased to 0.5 mm every 2 years. The changes further tapered off in adolescence, with alterations occurring at a reduced rate of 0.2 mm per year from 12 to 18 years old.

3.1.5. Atlas – First cervical vertebra

The anterior arch of the atlas widened while its superior articular face enlarged mostly in the first two years of life (Figure 9). The anterior arch experienced a considerable shortening of 5.0 mm from the first month of life to 18 years of age, coinciding with a symmetric widening of the articular faces. The transverse process underwent substantial growth predominantly within the first year of life. The articular faces, articulating with the axis and facilitating head movements, underwent the most significant changes between 1 and 8 years of age. Subsequently, from 10 to 12, the rate of change decreased to 0.3 mm, followed by a further reduction to 0.2 mm from 12 to 18.

3.1.6. Axis – Second cervical vertebra

The spinal process of the axis emerged in conjunction with the enlargement of the axis dent (Figure 10). The most significant alteration occurred in the vertebral arch, exhibiting a noteworthy change of 5.0 mm from the first month of life to 18 years of age. Subsequent changes in the axis dent involved a gradual thinning process, with the most substantial alterations occurring between 1 and 6 years of age. The pace of change reduced to 0.5 mm every 2 years until 10 years, followed by a further attenuation to 0.1 mm until 18 years. The spinal process underwent considerable transformations, with a notable 5.0‐mm change in the initial months compared to 18 years of age. This rate diminished to a 2.0‐mm difference at 4 compared to 18, followed by a slower, more gradual growth.

3.2. Sutures and synchondroses

3.2.1. Synchondroses

Two synchondroses were open at birth in the occipital bone, the anterior synchondroses (left and right) and the posterior synchondroses (left and right). The anterior synchondroses were open on the CT‐scans at 1 month of age, started closing at 11 months, were half closed at 18 months, were almost closed at 2 years, and were completely closed at 4 years (Figure S1). The posterior synchondroses were open at 1 month, half closed at 4 months, almost closed at 6 months, and completely closed at 18 months. Globally, most of changes occurred between birth and 3 years, then followed mostly by the growth of the fossa. Small numbers (10 patients) did not allow the statistical assessment of the closure of the posterior intra‐occipital synchondrosis in this age range. The following analysis thus focuses on the anterior intra‐occipital synchondrosis based on data from 14 patients.

When considering closure along the left anterior intra‐occipital synchondrosis (Figure 11), the width of the synchondrosis decreased from 4.0 to 2.0 mm between 0 and 200 days of age. Throughout growth, the reference curve grew from 2.0 to 3.0 mm, indicating that the nodes were evenly displaced further from each other. The intersection of the two curves was located at 180 days of age (that is 6 months). The intersection of the confidence intervals was between 120 and 215 days of age (that is between 4 and 7 months of age).

FIGURE 11.

Closure of the left and right anterior intra‐occipital synchondrosis.

For the right anterior intra‐occipital synchondrosis, the width decreases from 3.5 to 2.0 mm between 0 and 400 days of age. Globally, from birth to eighteen years old, the reference curve grew from 2.5 to 3.0 mm. The intersection of the two curves was located at 380 days of age (that is 12.5 months). The intersection of the confidence intervals was between 240 and 420 days (that is between 8 months old and 14 months).

3.2.2. Sutures

The spheno‐frontal suture started its closure already before 1 month of age, was almost closed at 3 years, and completely closed at 6 years. The spheno‐temporal suture was already partially closed before 1 month of age and was completely closed at 4 months.

The occipitomastoid suture has a left aspect (connection with left temporal bone) and a right aspect (connection with right temporal bone). The closure of this suture started in its middle portion, was almost effective at 8 months old, and was completed at 1 year of age (Figure 12).

FIGURE 12.

Percentage of closure for the occipitomastoid suture according to age.

At 6 months, the maximum distance was around 8.0 mm for the left side, then around 3.0 mm between 6 months and 1 year, and finally around 0.7 mm at 2 years, meaning that the suture was closed at this time.

3.3. Covariation patterns

The relationship between the first and second cervical vertebrae (C1 and C2) stood out with a covariance coefficient of r = 0.954, explaining 91.3% of the total variance between 0 and 18 years of age (Figure 13, Table S4). Specifically, C1 exhibited growth in its transverse process and posterior and anterior tuberculum, while C2 underwent changes in its corpus and spinal process.

FIGURE 13.

Covariation (PLS) between all bones from 0 to 18 years, then divided in four age categories. Plain line: link on the first axis; dotted line: link on another axis, with line thickness proportional to rPLS values, representing the percentage of variance explained by the covariance – thin line: 0%–50%; medium thin line: 50%–75%; thickest line: 75%–100%.

Another significant covariation was observed between C1 and the occipital bone, marked by a covariance coefficient of 0.907, contributing to 73.4% of the total variance. As the occipital bone exhibited changes in its lateral parts and foramen magnum, corresponding alterations were discerned in the inferior articular faces of C1.

Moreover, the occipital and sphenoid bones were marked by a covariance coefficient of 0.8, elucidating 49.44% of the total variance. Specifically, as the lateral parts of the occipital bone underwent development, concurrent changes manifested in the major wings of the sphenoid, which exhibited expansion. Additionally, the pterygoid process of the sphenoid demonstrated lengthening during this covariation.

The occipital bone showed covariation coefficients of 0.883 with the left temporal and 0.880 with the right temporal, explaining 53.42% and 40.93% of the total variance, respectively. This suggested a coordinated pattern where changes in the lateral parts of the occipital bone were associated with distinct alterations in the zygomatic process of both the left and right temporal bones. Specifically, as the lateral parts of the occipital expanded, the zygomatic processes of the temporals concurrently sharpened and elongated.

A relationship between the sphenoid bone and both the left and right temporal bones was also reported. The covariation coefficients of 0.899 with the left temporal and 0.903 with the right temporal unveiled substantial connections, elucidating 59.97% and 53.85% of the total variance, respectively. As the squamous part of the temporal became rounder and the zygomatic process sharper, the major wings of the sphenoid concurrently exhibited a more horizontal orientation, and the pterygoid process underwent vertical lengthening.

There was also a connection between the sphenoid and ethmoid bones, as indicated by a covariation coefficient of 0.786, elucidating 37.36% of the total variance. As the pterygoid process of the sphenoid bone lengthened vertically, the crista galli of the ethmoid bone underwent a distinct movement towards the superior aspect.

The covariation analysis between the sphenoid and orbitals had a coefficient of 0.886, elucidating 34.34% of the total variance. As the major wings of the sphenoid bone assumed a more horizontal and wider orientation, coupled with a reduction in the height of the pterygoid process, various alterations occurred in the orbital region. Specifically, the nasal spine exhibited a posterior displacement, while, on a global scale, other aspects of the orbital region experienced an overall enlargement.

The observed significant correlation coefficient of 0.976 contributed to explaining 17.55% of the total variance between the sphenoid bone and the right temporal between 0 and 2 years of age (Figure 13, Table S5). As the major wings of the sphenoid exhibited a broader morphology, a corresponding widening of the squamous temporal part ensued, accompanied by a sharper, and more developed zygomatic process in the right temporal bone.

Between the occipital and both temporals, the calculated coefficients of 0.963 (right) and 0.966 (left) underscored the strong correlations between these bone pairs, collectively elucidating 82.74% and 80.05% of the total variance. Notably, as the occipital bone assumed a rounder contour, the mastoid process in the temporal bones exhibited a more expanded morphology.

With a calculated coefficient of 0.971 between the sphenoid and the right temporal, covariation between 2 and 6 years explained 54.49% of the total variance (Figure 13, Table S6). As the major wings of the sphenoid widened, a rounder and more compact configuration was observed in the temporal. Moreover, an elongated pterygoid process in the sphenoid coincided with a vertically extended petrous part in the temporal.

With a computed coefficient of 0.96, the covariation between the sphenoid bone and the occipital accounted for 20.27% of the total variance. As the sphenoid underwent morphological shifts marked by horizontally broader major wings and a vertically shortened pterygoid process, we observed a discernibly tighter foramen magnum.

The occipital and the sphenoid bone were marked by a coefficient of 0.971, elucidating 29.58% of the total variance. As the posterior arch of C1 thickened, a wider clivus in the occipital region ensued. Also, when the transverse processes of the atlas underwent horizontal expansion, a concomitant tightening of the foramen magnum was observed.

The covariation between the occipital and atlas bones unfolded with a coefficient of 0.943, unravelling 45.6% of the total variance between 6 and 12 years of age (Figure 13, Table S7). When the posterior arch of the atlas opened, a notable enlargement of the clivus took place. In contrast, as the arch closed, a fully formed anterior tuberculum came into prominence, accompanied by a more compact clivus, and widened cerebellar fossa.

The atlas and axis were marked by a coefficient of 0.936 and elucidating 51.99% of the total variance. The atlas showed a wide posterior arch and robust lateral masses, while the dens axis of C2 was higher, the vertebra corpus wider, and the spinal process bigger. As the posterior arch of C1 assumed a thinner profile, the subsequent act revealed a more compact manifestation of the axis.

The interplay between the orbital region and the sphenoid bones showed a coefficient of 0.968 elucidating 24.26% of the total variance. As the intracranial aspect of the sphenoid adopted a more hollowed shape, the supraorbital incision got clearer, and the ethmoid incision extended.

The covariation analysis between the occipital and sphenoid bones reveals a coefficient of 0.983, explaining 50.96% of the total variance between 12 and 18 years of age (Figure 13, Table S8). An expanded basilar part in the occipital corresponded to a vertical extension of the major wings of the sphenoid and a lengthening of the pterygoid process. Additionally, the rounder and more defined lateral parts of the occipital aligned with larger and more rounded intracranial aspects of the sphenoid bone.

The covariation analysis between the atlas and axis vertebrae showed a coefficient of 0.923, elucidating 57.4% of the total variance. A wider posterior arch of the atlas corresponded to a broader spinal process in the axis.

4. DISCUSSION

Sutures and synchondroses are essential contributors to craniofacial growth (Opperman, 2000). Their functional significance lies in accommodating the rapid growth of the developing skull, providing flexibility during delivery, and facilitating incremental expansion (Celik et al., 2021). Precise closure timing is critical, as premature fusion can result in malformations such as craniosynostoses, affecting cranial growth and leading to functional consequences such as increased intracranial pressure and obstructive sleep apnoea (Paliga et al., 2014; Vu et al., 2021).

Using 3D approaches, several authors have studied suture and synchondrosis closures individually by scoring them as open, partially fused, mostly fused, and completely fused (Alhazmi et al., 2017). Here we introduced a quantification of suture closure using a morphometric approach. This quantitative assessment provides data on the specific periods and patterns associated with the closure of these structures, enhancing our understanding of craniofacial growth dynamics. Notably, the age of closure was predicted and could lead the development of clinically relevant patient‐specific models. Additionally, not only the time of closure was predicted but also the geometric pattern of closure. Data on the percentage of closure at each age and on the dynamics of closure would enhance the level of detail of these models, potentially facilitating early detection of deviations from the normal growth patterns. This could be achieved by assessing high‐resolution data such as synchrotron microtomography (Khonsari et al., 2012).

The understanding of covariations in craniofacial growth encompasses the link between vertebral structures, particularly the cervical vertebrae, and the cranial base or face (Arlegi et al., 2022; McCane & Kean, 2011). Several studies have highlighted the correlation between head posture and vertebral position (Kylämarkula and Huggare, 1985). Our study extends the scope of covariation analyses to finer and age‐specific patterns within the CCJ. Notably, the strong covariance between the occipital and first cervical vertebra (OCC and C1), and the covariance between the first and second cervical vertebrae (C1 and C2) had not been previously reported.

On the full age range of the dataset (0–18 years old), the covariation between the occipital and C1 (r = 0.907) indicates the strong connection between these two bones. The covariation between C1 and C2 (r = 0.954) is the one holding the biggest percentage of variance represented. More precisely, the covariation between the occipital and C1 is only represented between 2 and 12 years and the covariation between C1 and C2 only between 6 to 18 years. This is confirmed by literature as the spine exhibits a regionalized structure, where segments within the same region share more similarities than those in different regions (Zelditch & Goswami, 2021). During early development, the coordination and growth of C1 and C2 are intricately linked, contributing to the establishment of the CCJ. The identified covariation patterns within this junction may provide valuable insights, hinting at their potential relevance to pathological conditions like atlas occipitalization and ossiculum terminale of Bergmann, implying a correlation with disorders involving the fusion of neighbouring bones.

On the other hand, other cranial base bones, with less straightforward functional relationships, may exhibit more independent growth patterns. The lack of pronounced covariation between these bones within specific age ranges suggests that their individual developmental trajectories may be influenced by distinct factors. However, the observed covariation at the global scale from 0 to 18 years old indicates that, despite distinct growth patterns within certain periods, there is an overarching coordination in cranial development throughout the entire paediatric age spectrum. For example, there is a strong correlation between the occipital and both temporal bones between birth and 2 years of age, that disappears between two and eighteen years of age.

All partial least squares (PLS) analyses conducted in this study were performed on scaled data. This procedure neutralizes the influence of size‐related changes in growth, allowing for a focused examination of shape‐related covariations across different age categories. By eliminating the confounding effects of size variation, the scaled analyses accentuate the significance of allometries in the observed covariation patterns.

Craniofacial bones arise from the neural crest, the mesoderm and/or the occipital somites, without any known functional significance of this origin during post‐natal growth (Couly et al., 1993; Khonsari et al., 2007; McBratney‐Owen et al., 2008). The mammalian embryological development of the sphenoid bone involves two components: the basi‐post‐sphenoid and the orbito‐sphenoid, derived from cephalic mesoderm, and the basi‐pre‐sphenoid and the ali‐sphenoid, originating from neural crest cells (Catala, 2003). Temporal bones result from the combination of neural crest cells and mesoderm (Som et al., 2016). The occipital bone is derived from both neural crest cells and paraxial mesoderm (Shapiro & Robinson, 1976). The frontal bone, forming the orbital region, results from a combination of neural crest cells and mesoderm (Tyler, 1983). On the other hand, the first and second cervical vertebrae, the atlas and axis, belong to the axial skeleton and derive from somites (Huang et al., 2000). Checking if our correlations match with embryological origins of each bone may provide insights into the developmental pathways and genetic signals influencing the coordinated growth within the CCJ. However, the observed covariations do not seem to particularly align with embryological origins. Here, the occipital and the temporal bones often covary but have distinct embryological origins. On the other hand, C1 and C2 go through the same embryogenesis process but do not show consistent covariations through growth. It is possible that post‐natal functional adaptations take precedence on potential intrinsic embryological programs, as the CCJ responds dynamically to biomechanical forces associated with head movements and other functional inputs. Environmental influences, including muscular interactions and gravitational effects, also play a role in shaping skull bone growth postnatally. Individual genetic variability and unique responses to environmental stimuli introduce diversity in growth trajectories, influencing bone morphology beyond embryonic setup. In this context, two major clues linking genetic factors and post‐natal function are the position of muscle insertions and the mechanosensation and mechanotransduction mechanisms in the craniofacial region (Khonsari et al., 2013; Matsuoka et al., 2005; Tiberio et al., 2021).

The current study identifies key zones of interest in paediatric craniospinal development. This knowledge is crucial for defining morphological criteria that can be used to quantify inter‐individual variability. For instance, the lateral parts of the occipital bone and the region around the foramen magnum, as highlighted in our study, present morphological criteria sensitive to age‐related changes. Moreover, the study scrutinizes the major wings of the sphenoid bone and the development of the pterygoid process, revealing the growth dynamics of these components within the CCJ.

The strengths of our study lie in its ability to quantify closure patterns of sutures and synchondroses using a morphometric approach, offering a more precise timeline and pattern analysis. Existing research on the anterior intra‐occipital synchondrosis indicates it typically reaches a midpoint at 4 years old and closes around 10 years old (Vu et al., 2021). Our study specifically identifies the midpoint at eighteen months and closure at the earlier age of 4 years, suggesting a potentially earlier timeline for this developmental milestone in the cranial base. The integration analysis on all CCJ bones provides a holistic understanding of the coordinated growth within this critical junction. These strengths support the applicability of our study in both research and clinical contexts, providing valuable benchmarks for studying normal craniofacial development and contributing to the comparisons with models of CCJ anomalies by identifying deviations from healthy trajectories.

However, certain limitations should be acknowledged. First, our investigation is primarily a cross‐sectional study, with independent subjects rather than a longitudinal design, hindering our ability to follow individual participants over time. Furthermore, the sample size within each age category, although providing insightful results, is relatively small. We know the most changes occur in the first two years of life and a focus on this range would be more useful. This limitation restricts the generalization of our findings and calls for future studies with larger cohorts. Such expanded cohorts would facilitate more in‐depth analyses on additional regions, allowing for a more precise quantification of growth regions. Particularly, a larger sample size would enhance the examination of sutures and synchondroses that close at early stages, providing a more comprehensive understanding of craniofacial development in the paediatric spectrum. Nevertheless, paediatric CT‐scan data is not abundant and MRI imaging is generally preferred to investigate this region to decrease the exposure to radiations. Recent developments in MRI‐based bone segmentation could help to increase our database (Paholpak et al., 2021).

5. CONCLUSION

This study contributes to the understanding of paediatric craniofacial growth by introducing new methods applied to a reference control cohort. The incorporation of growth rates, a closure model for sutures and synchondroses, and an integration analysis on all CCJ bones are new contributions. The combination of these components offers complementary pipeline towards a more comprehensive exploration of the intricate patterns of cranial base bone growth, providing a detailed reference model.

Future research perspectives should involve the expansion of the cohort, enabling a more thorough examination of cranial growth, especially by allowing the examination of other sutures and synchondroses closing early pre‐ or post‐natally. Comparative studies, particularly with conditions such as achondroplasia and FGFR‐related craniosynostoses, will be crucial to delineate differences in shape‐ and age‐specific patterns. These endeavours aim to further refine our understanding of craniofacial growth and contribute to the development of targeted diagnostic and therapeutic approaches for abnormal cranial development.

Supporting information

Tables S1–S8.

Figures S1–S2.

Raoul‐Duval, J. , Ganet, A. , Benichi, S. , Baixe, P. , Cornillon, C. , Eschapasse, L. et al. (2024) Geometric growth of the normal human craniocervical junction from 0 to 18 years old. Journal of Anatomy, 245, 842–863. Available from: 10.1111/joa.14067

DATA AVAILABILITY STATEMENT

The authors commit to make the raw data of this study available on an online repository upon acceptance of the manuscript.