Skip to main content
NCBI home page
Journal of Anatomy logoLink to Journal of Anatomy
. 2019 Nov 6;236(1):105–116. doi: 10.1111/joa.13085

Cranial growth in isolated sagittal craniosynostosis compared with normal growth in the first 6 months of age

Ezgi Mercan 1, Richard A Hopper 1,2, A Murat Maga 3,4,✉
PMCID: PMC6904596  PMID: 31691965

Abstract

Sagittal craniosynostosis (SCS), the most common type of premature perinatal cranial suture fusion, results in abnormal head shape that requires extensive surgery to correct. It is important to find objective and repeatable measures of severity and surgical outcome to examine the effect of timing and technique on different SCS surgeries. The purpose of this study was to develop statistical models of infant (0–6 months old) skull growth in both normative and SCS subjects (prior to surgery). Our goal was to apply these models to the assessment of differences between these two groups in overall post‐natal growth patterns and sutural growth rates as a first step to develop methods for predictive models of surgical outcome. We identified 81 patients with isolated, non‐syndromic SCS from Seattle Children's Craniofacial Center patient database who had a preoperative CT exam before the age of 6 months. As a control group, we identified 117 CT exams without any craniofacial abnormalities or bone fractures in the same age group. We first created population‐level templates from the CT images of the SCS and normal groups. All CT images from both groups, as well as the canonical templates of both cohorts, were annotated with anatomical landmarks, which were used in a growth model that predicted the locations of these landmarks at a given age based on each population. Using the template images and the landmark positions predicted by the growth models, we created 3D meshes for each week of age up to 6 months for both populations. To analyze the growth patterns at the suture sites, we annotated both templates with additional semi‐landmarks equally spaced along the metopic, coronal, sagittal and lambdoidal cranial sutures. By transferring these semi‐landmarks to meshes produced from the growth model, we measured the displacement of the bone borders and suture closure rates. We found that the growth at the metopic and coronal sutures were more rapid in the SCS cohort than in the normal cohort. The antero‐posterior displacement of the semi‐landmarks also indicated a more rapid growth in the sagittal plane in the SCS model than in the normal model. Statistical templates and geometric morphometrics are promising tools for understanding the growth patterns in normal and synostotic populations and to produce objective and reproducible measurements of severity and outcome. Our study is the first of its kind to quantify the bone growth for the first 6 months of life in both normal and sagittal synostosis patients.

Keywords: 3D, craniosynostosis, geometric morphometrics, morphometrics, shape modeling

Normal growth vs. growth in sagittal craniosynostosis. (A) Mediolateral displacement: positive values indicate medial movement (can be interpreted as suture closure) and negative values indicate lateral movement (mostly associated with the overall growth of the skull). (B) Superoinferior displacement: positive values indicate a superior movement. (C) Anteroposterior displacement: positive values indicate anterior, and negative values indicate posterior movement with respect to basion.

graphic file with name JOA-236-105-g008.jpg

Introduction

Sagittal craniosynostosis (SCS) is the most common type of premature perinatal cranial suture fusion that results in progressive abnormalities in cranial shape. Although our understanding of the genetics of SCS has increased over the years (Cox et al. 2013; Heuzé et al. 2014; Flaherty et al. 2016; Clarke et al. 2018), clinical care of these patients has remained much the same, requiring substantial surgical corrections to address concerns regarding normal brain development due to restricted intracranial volume (ICV) or increased intracranial pressure. Surgeries can be divided into techniques that correct head shape in a single stage and those that require either internal devices or postoperative molding helmet treatment to achieve their goals (Governale, 2015). Common to all procedures is that the infant or young child must be anesthetized for 1–3‐h for a surgery that results in considerable blood loss (based on clinical data collected in SCH Craniofacial Clinic). The increased duration of anesthesia and the larger volume of blood transfusion during any type of craniofacial surgery have been associated with potentially poor neurodevelopmental outcomes in patients later in life (Tunçbilek et al. 2005; Stricker & Fiadjoe, 2014; Bonfield et al. 2016). Since these surgeries are typically performed in the first 6 months of age, it is important to identify the optimal time to intervene for best surgical outcome while having the least detrimental impact to the patient (Yarbrough et al. 2014; Sivakumar et al. 2015; Bergquist et al. 2016; Persing, 2016; Mann et al. 2017). An increased understanding of the shape changes and suture closure patterns that occur in SCS compared with normal children in the first 6 months of age would provide valuable information in pursuit of this goal.

Current challenges in CS outcome research

There are both biological (high rate of skull and brain growth in infants) and non‐biological (choice of surgical technique, age at surgery, socioeconomic status) variables that may impact the surgical outcome. Although craniosynostosis cases can be considered common for a structural birth defect, they are still relatively rare conditions (1/1000 for non‐syndromic CS, and less frequent for syndromic cases) (Garza & Khosla, 2012). These variables, as well as the paucity of large datasets, contribute to the wide range of results and contradictory opinions that exist in the craniosynostosis surgical literature. For example, the important metric of intracranial volume (ICV), which CS surgeries are designed to increase, have been reported to be significantly higher (Heller et al. 2008), lower (Sgouros et al. 1999; Seeberger et al. 2016) as well as not different in preoperative synostosis patients compared with controls (Netherway et al. 2005; Anderson et al. 2007; Tovetjärn et al. 2014; Fischer et al. 2016). Such variable results can be attributed to low sample sizes (only one study had 30 cases in any group – Fischer et al. 2016 and some studies had sample sizes as low as five per group), to differences in cohort selection (different types of CS might be lumped together), or to methodology. There is also no clear consensus on how to measure the surgical outcome objectively in a longitudinal fashion (Lloyd et al. 2016). Long‐term studies have indirectly evaluated surgical outcome by comparing the neurodevelopmental outcomes of CS patients who have undergone surgery with those of non‐CS children (Speltz et al. 2004, 2015; Kapp‐Simon et al. 2007; Starr et al. 2010). Although these studies tend to factor in surgical technique as a variable, they did not always control for the initial severity of the condition and the magnitude of surgical correction. Even when the severity was measured, the metrics used were simplistic and did not represent the full spectrum of disorder (Heller et al. 2008; Dvoracek et al. 2015; McEwan et al. 2016; Fearon et al. 2017). One study that used the facial appearance of children with CS who underwent surgery found that lay‐people did not perceive them as ‘normal’ compared with unaffected children, even after surgery (Collett et al. 2013).

In summary, there is an acute need to find objective and repeatable measures of severity and surgical outcomes in SCS treatment and planning. As the first step towards this goal, we developed statistical models of infant (0–6 months old) skull growth both in normative and SCS patients and applied them to the assessment of differences in both overall post‐natal growth patterns and sutural growth rates. By studying the shape and suture changes in SCS patients during this critical period, we can better understand the implications of operative intervention at a given age on final surgical outcome.

Materials and methods

The study was reviewed and approved by the Institutional Review Board at Seattle Children's Hospital (STUDY00000167, STUDY00000424).

Dataset and statistical analysis

There were three distinct steps in our full analysis: estimation of canonical templates for both normal and SCS cohorts; statistical modeling and visualization of cranial growth; and estimation of bone growth and suture closure rates. Below we provide the specifics of each step. A summary flowchart is also provided in Supporting Information Fig. S1. Table 1 summarizes the demographics of both normal and SCS cohorts.

Table 1.

Demographics – sagittal craniosynostosis and control cohorts

Control Sagittal craniosynostosis
Age range (min–max) 3–183 days 11–179 days
Sex distribution
Female 25 19
Male 92 62
Anatomical template
Mean age 93 days 97 days
Number of samples 34 81
Open in a new tab

Normal cohort

We retrospectively reviewed all patients under the age of 6 months with a head CT exam performed at Seattle Children's Hospital between 2004 and 2017. The patients with skull abnormalities, bone fractures, head/face/ear deformities and congenital conditions (including growth restrictions) were excluded after a review of radiology reports and clinical histories. For the remaining patients, CT scans were 3D‐rendered in bone window using 3d slicer (Fedorov et al. 2012) and visually inspected to discard any low‐quality or substantially incomplete field of view of cranium.

Our final control cohort included 117 CT exams, of which 79 had a complete cranium. In the remaining 38 CT exams, the maxilla was cut below the nasal spine, yet all the pertinent anatomical landmarks were identifiable. All 117 exams were included in the control population. The most common reasons for a head CT exam with no abnormal findings were head injury, seizure and scalp swelling.

Sagittal craniosynostosis cohort

Seattle Children's Craniofacial Center maintains an independent database of patients seen in the clinic. We reviewed this database to identify patients under the age of 6 months with a non‐syndromic isolated sagittal craniosynostosis diagnosis with a preoperative CT scan performed at our institution between 2004 and 2017. We visually inspected CT scans by rendering them in bone window to discard any low‐quality or incomplete scans. Since preoperative CT scans were performed for surgical planning or diagnostic purposes, almost all scans provided a high‐quality and complete image of the cranium. Our sagittal craniosynostosis cohort consisted of CT scans from 81 patients.

Anatomical template building

We built both a normal and an SCS population‐level anatomical template for infants 0–6 months of age. Because we expected that major changes in the normal skull form would be driven entirely by biological growth, we wanted the normative canonical template to be unbiased with respect to age, as well as sex. Therefore, we selected 34 individuals from our normal cohort, such that CT exams from a male and a female patient approximately represent 1 week of age between 0 and 6 months.

For the SCS cohort, we had no such a priori expectation, and qualitatively the phenotypic variability of skull forms appeared to be independent of the age. Moreover, SCS tends to be more common in males than females (with an approximate ratio of 4 : 1, Garza & Khosla, 2012), making it harder to identify a sub‐cohort that is balanced both for age and sex. Therefore, we opted to use all of our SCS cohort to generate SCS.

All CT images in this study were manually edited to remove mandible and other non‐cranial bones and artifacts (e.g. tubes, pacifiers) using Segment Editor in 3d slicer (Fedorov et al. 2012). A global threshold of 143 HU was used to extract only the bony cranium from CT exams, and all samples were rigidly aligned using automated methods from Advanced Normalization Tools (ANTs) package (Avants et al. 2009, 2011) to a reference sample to remove positional differences. At this step, all exams were mirrored along the midsagittal plane, and these mirror images included in the anatomical template building process to generate bilaterally symmetric templates. Since SCS is not associated with any bilateral asymmetry in the skull, any asymmetry introduced in both normal and SCS templates would be individual variation. Although it was not necessary, by mirroring exams and creating a symmetrical template, we simplified the calculation in later steps.

We used the multivariate template construction pipeline that is provided with ANTs image registration and analysis ecosystem (Avants et al. 2009, 2011). ANTs template‐building script estimates the best average shape that represents the entire population by iteratively calculating an average image and registering each sample to the new average. We used the default registration parameters: the geodesic diffeomorphism transformation of symmetric normalization (SyN) and the similarity metric of cross‐correlation (Avants et al. 2011). We found that six iterations were sufficient to construct an anatomically refined template and that, after the 5th iteration, the changes in the template were negligible. The resultant template reviewed by a morphologist (A.M.M.) and a craniofacial surgeon (R.H.) for anatomical accuracy.

Landmarks and semi‐landmarks

We rendered all exams in bone window and annotated 38 anatomical landmarks (Supporting Information Appendix S1) using a 3d slicer program (Fedorov et al. 2012). A subset of five control and five sagittal craniosynostosis images were independently annotated by two observers (E.M. and A.M.M.) to measure inter‐rater variability, and the annotations were repeated after a wash‐out period of 1 month to measure intra‐rater variability. After obtaining low intra‐ and inter‐rater variances (< 3 mm overall all landmarks, < 2 mm when euryon, glabella, vertex and inion were excluded, see Supporting Information Appendix S2 for individual landmark variances) for all landmark points, all CT scans in both cohorts were annotated by E.M. Figure  1 illustrates the landmarks on the normative template. The cranial landmarks were annotated on all CT scans and used to model growth in two populations.

Figure 1.

Figure 1

Open in a new tab

Thirty‐eight cranial landmarks used in the study (the landmark vertex, ‘v’, was removed in control samples since it does not fall onto a bony structure in young infants with a patent sagittal suture).

Although cranial landmarks represented the standard anthropometric features used to assess overall skull shape and size in literature, they are not sufficient to measure precisely the growth and size changes along the major cranial sutures. We therefore annotated the major cranial sutures on both templates with semi‐landmarks equally spaced within regions (Fig. 2. Starting from the landmark nasion and ending at the landmark asterion, cranial sutures were landmarked in the following order: metopic, coronal, sagittal and lambdoidal. Both templates had a total of 80 matched semi‐landmark points to allow direct comparison. Since the normal and SCS templates were symmetrical, only the left side of the cranium was annotated with the semi‐landmarks. The semi‐landmarks were only annotated on the templates and were automatically transferred to the meshes produced with the landmark‐based growth model. The details of semi‐landmark annotation are given in Appendix S1.

Figure 2.

Figure 2

Open in a new tab

Semi‐landmarks placed along the metopic, coronal, sagittal and lambdoidal sutures to estimate the suture closure rates in control and sagittal craniosynostosis populations. Red points correspond to the borders of different regions based on sutures and fontanelles. Top row: normal template; bottom row: sagittal craniosynostosis.

Growth models

We modeled skull growth using the cranial landmarks and patient age. The models predicted the locations of the cranial landmarks over the age range of 0–6 months. The normal growth model included 117 samples (a total of 234 with mirrored samples) and the sagittal craniosynostosis growth model included 81 samples (a total of 162 with mirrored samples). Since the young infants in the control dataset have a patent sagittal suture, the vertex point (the highest point of the head) was not always on a bony structure and was thus excluded from the normal growth model. Similarly, the inion had a high variance in the sagittal craniosynostosis population and was excluded from the sagittal craniosynostosis growth model. Cranial landmarks were aligned using Generalized Procrustes analysis (GPA) with the geometric morphometrics package Morpho in R (Bookstein, 1991; Schlager, 2017). Because we wanted to model cranial growth in physical space, the uniform scaling step of the GPA analysis was skipped.

Principle component analysis (PCA) is a workhorse of the geometric morphometric methods (Bookstein, 1989). It converts a set of observations (landmark coordinates) into uncorrelated latent variables sorted in the order of decreasing variance. PCA is commonly used for feature selection by eliminating the components with small variances. In our models, we used the first 20 principle components (PCs), which explained ~ 90% of the variation while reducing the number of variables from 111 to 20. PCA also provides uncorrelated variables that can be modeled independently in contrast to the ‘raw’ coordinate values that are highly correlated in cranial growth (see Supporting Information Appendix S3 for comparison of Procrustes coordinates with PC scores in prediction of landmark positions). Several regression models (linear, local, polynomial) were evaluated in a leave‐one‐out cross‐validation (LOOCV) setting (Appendix S3) and the linear regression, which has comparable error to other models but was much simpler than others, was chosen. All statistical modeling was conducted in R.

Bone growth and suture closure

Using the growth models, we estimated landmark positions for each week of age up to 6 months for both the control population and the sagittal craniosynostosis population. Landmark positions provide control points that allow a surface mesh to be interpolated to estimate the shape and size changes in growing skulls. We extracted surface meshes from the volumetric template images using the marching cubes algorithm in VTK library (Lorensen & Cline, 1987; Schroeder et al. 2004) and warped the template meshes to the target landmark points at different ages using the thin plate splines (TPS) method from the package Morpho in R (Bookstein, 1989; Schlager, 2017).

The semi‐landmarks from the templates were transferred to the meshes at different ages produced by TPS. Since the orientation of the mesh determines the growth direction, meshes were aligned to Frankfort Horizontal using the landmarks porion and zygoorbitale and translated to have the basion as the origin. Finally, the displacement of each semi‐landmark was calculated between consecutive weeks and averaged to obtain average growth rates at the margins of the bone plates for both populations.

Results

Template meshes for normal and SCS cohorts, anatomical landmarks for samples and templates, and meshes generated based on growth model predictions are available online on the OSF project page (https://osf.io/kft8y). Landmark coordinates and cohort demographics are also provided as Supporting Information Data S1.

Principal component analysis of the landmarks

To better understand the shape changes associated with each PC, we visualized the eigenvectors associated with each PC. By changing the PC scores between two standard deviations around the mean of their respective populations, we created synthetic meshes that illustrate the shape changes captured by each PC.

For our study population, PC1 (visualized in Fig.  3) explained 35% of the variation in the control cohort and 30% of the variation in the sagittal craniosynostosis cohort. In both populations, PC1 modeled the overall skull size and the closure of the anterior fontanelle (as noted by red arrows in Fig.  3). Because change in skull size is strongly correlated with age in young infants, PC1 was also strongly correlated with the age in both populations (R 2 = 0.68 for normal cohort and R 2 = 0.59 for sagittal craniosynostosis cohort) (Supporting Information Fig. S2).

Figure 3.

Figure 3

Open in a new tab

The shape variation captured by the first principle component. The top row shows the reconstructed normative template and the bottom row shows the reconstructed sagittal craniosynostosis template using the first PC scores in two standard deviations around the mean of their respective populations.

Bone growth

We quantified the bone growth and suture closure rates by measuring the displacements of semi‐landmarks that were transferred to surface meshes for each week of age from 11 days to 179 days. Although control cohort had younger (3 days old) and older (183 days old) patients, we did not have any samples to model the sagittal craniosynostosis growth for these extreme ages, hence we limited our growth analysis to a slightly narrower age range for the sagittal craniosynostosis cohort (11–179 days old). Figure 4 shows surface meshes and semi‐landmarks for five samples generated by the growth models for both cohorts. We analyzed the growth in three major axes: horizontal (mediolateral), anteroposterior and superoinferior. To further simplify the interpretation, we divided the semi‐landmarks into seven regions as described in Table 2.

Figure 4.

Figure 4

Open in a new tab

Visualizations of predicted growth from the statistical model along selected time points in normal (top row) and craniosynostosis populations (bottom row). An animation of continuous visualization is provided as Supporting Information Fig. S3.

Table 2.

Semi‐landmark regions along the metopic, coronal, sagittal and lambdoidal sutures

Region Description
Region 1 The metopic suture from the nasion to the anterior‐most point of anterior fontanelle at the mid‐sagittal plane.
Region 2 Anterior border of the anterior fontanelle on the frontal bone from the metopic suture to the coronal suture
Region 3–4 The coronal suture between parietal and frontal bones from the lateral apex of the anterior fontanelle towards the pterion constitutes region 3 and the reverse direction from the pterion towards the anterior fontanelle constitutes region 4.
Region 5 The posterior border of the anterior fontanelle on the parietal bone from the coronal to the sagittal suture.
Region 6 The sagittal suture (fused on the sagittal synostosis template and unfused on the normal template) between two parietal bones from the posterior‐most point of the anterior fontanelle on the mid‐sagittal plane to the landmark lambda.
Region 7 The lambdoidal suture from the landmark lambda to the edge of the mastoid fontanelle.
Open in a new tab

Mediolateral growth

Figure 5A shows the normal growth vs. growth in sagittal craniosynostosis in the mediolateral axis. The y‐axis shows the displacement (mm/week) at semi‐landmark positions where positive values indicate medial movement (can be interpreted as suture closure) and negative values indicate lateral movement (associated with overall growth of the skull). Displacement along the metopic suture (R1) and the sagittal suture (region 6) of both normal and SCS models was negligible. The displacement of frontal bone suture landmarks along the anterior border of the anterior fontanelle (R2) was in the lateral direction for the normal model. This movement away from the midsagittal plane is indicative of an overall regional growth in the normal model that was faster than the fontanelle closure (medial displacement) at the metopic and coronal sutures. In comparison, the overall medial movement of the corresponding R2 suture landmarks in the sagittal craniosynostosis model indicates that fontanelle closure was more rapid than overall lateral cranial growth. The displacement in R3 and R4, which corresponds to movement along the coronal suture, was in the lateral direction for both models, indicating lateral cranial growth, but it was smaller for the SCS model, indicating a restricted horizontal growth relative to normals. Parietal bone landmarks in R5 along the posterior border of the fontanelle moved in a medial direction, particularly as they approached the sagittal suture, indicating parietal bone growth towards the anterior fontanelle in both groups. This medial displacement rate was faster in the SCS model.

Figure 5.

Figure 5

Open in a new tab

Normal growth vs. growth in sagittal craniosynostosis. (A) Mediolateral displacement: positive values indicate medial movement (can be interpreted as suture closure) and negative values indicate lateral movement (mostly associated with the overall growth of the skull). (B) Superoinferior displacement: positive values indicate a superior movement. (C) Anteroposterior displacement: positive values indicate anterior, and negative values indicate posterior movement with respect to basion.

Superoinferior growth

Figure 5B shows the displacement of semi‐landmarks in the longitudinal axis for normal and sagittal craniosynostosis growth models. In all regions, the displacement was in the cranial direction due to the alignment of the model with basion at the origin. There was no vertical compression of the skull with age. The cranial displacement indicates overall skull growth that was faster in the normal model than in the SCS model. The slower vertical or superoinferior growth of the sagittal craniosynostosis model indicated relative growth restriction in the longitudinal axis. Interestingly, the pace of growth was considerably slower at the posterior cranium (R6) for the sagittal craniosynostosis model, which likely contributes to the decreased vertical posterior cranial height and occipital protuberance phenotype commonly observed at the time of surgery in these patients.

Anteroposterior growth

Figure 5C illustrates the displacement of semi‐landmarks in the anteroposterior axis for normal and SCS growth models. The positive values in the y‐axis indicate anterior movement, whereas the negative values indicate posterior movement. The consistent anterior displacement of semi‐landmarks in R1 through R4 indicate overall faster anterior skull growth compared with the bone growth of the frontal bones into the anterior fontanelle. The positive displacement of semi‐landmarks in R5 indicates greater growth of the parietal bone into the anterior fontanelle compared with posterior displacement from overall cranial growth. This anterior displacement rate was faster for the SCS model, indicating more rapid closure of the anterior fontanelle compared with the normal model from this posterior border. The anterior movement of the semi‐landmark at the border of R5 and R6, the point where the sagittal suture meets the coronal sutures, was much faster in the SCS model than in the normal model. These semi‐landmarks were posterior to the basion (the point of origin for alignment) at the second half of the R6, and they move posteriorly for both models, indicating skull growth. The posterior growth was faster for thebsagittal craniosynostosis model, which contributes to the scaphocephalic elongated skull shape of the sagittal craniosynostosis patient.

Frontal bone growth

We compared the semi‐landmark movements in R1 and R3, two borders of the frontal bone along the metopic and coronal sutures. Semi‐landmarks in both regions moved anteriorly but more rapidly in R1, suggesting an increase in the distance between the metopic and coronal sutures and an overall anteroposterior growth in the frontal bone in both models (Fig. 5C). Similarly, the semi‐landmarks at the metopic suture (R1) did not move mediolaterally in both models; however, the semi‐landmarks at the coronal sutures (R2) moved laterally in both models (Fig. 5A). This indicated a mediolateral growth of the cranium at the frontal bone but at a much more rapid pace in the normal model.

Parietal bone growth

The semi‐landmarks in R4, R6 and R7 marked the borders of the parietal bone along the coronal, sagittal and lambdoidal sutures. The coronal suture semi‐landmarks (R4) and lambdoidal suture semi‐landmarks (R7) moved laterally in both models, whereas the sagittal suture semi‐landmarks did not move mediolaterally, indicating mediolateral growth of the parietal bone in both models (Fig. 5A). Lateral movement of the semi‐landmarks in R4 and R7 was faster in normal model, indicating a restriction in mediolateral growth of the parietal bone in SCS model. All semi‐landmarks surrounding parietal bone (R4–R7) moved superiorly in both models (Fig. 5B). The superior movement was faster in normal models, again indicating a superoinferior growth restriction of the parietal bone in the SCS model. Finally, the coronal suture semi‐landmarks (R4) moved anteriorly and the lambdoidal suture semi‐landmarks moved posteriorly in both models (Fig. 5C). The movement was faster in both directions for the SCS model, indicating a faster anteroposterior movement of the parietal bone compared with the normal model, contributing to the scaphocephalic elongated skull shape of the sagittal craniosynostosis patient.

Closure pattern of the anterior fontanelle

The symmetry of the models allowed an assessment of the relative movement of the frontal and parietal bone plates in the region of the anterior fontanelle. Lateral displacement of the suture landmarks would indicate that horizontal growth perpendicular to the metopic and sagittal sutures was greater than the growth of the frontal and parietal bone plates into the anterior fontanelle. In comparison, medial displacement would indicate that anterior fontanelle closure was faster than overall cranial horizontal growth. Leveraging the bilateral symmetry of the model, we reflected the semi‐landmarks across the mid‐sagittal plane and measured distances between the right–left pairs of semi‐landmarks. Using the growth model, we calculated the average weekly change in the distance between bilateral semi‐landmarks (Fig.  6. In normal growth, the overall width of the anterior fontanelle increased at a rate up to 0.3 mm per week at the point where it met the coronal, but it decreased in width at the posterior border as it approached the sagittal suture. This would indicate that the fontanelle closure was primarily from a medial in‐growth of the opposing parietal bone plates. In comparison, in the SCS model, the width of the anterior fontanelle decreased at both its anterior (metopic) and posterior (sagittal) apices. This would indicate that compared with normal fontanelle closure, in SCS there is a greater compensatory ingrowth of the frontal bones, resulting in a circumferential rather than a posterior to anterior closure observed in the normal model.

Figure 6.

Figure 6

Open in a new tab

Average change in the distance between bilateral pairs of semi‐landmarks.

A similar analysis was conducted by calculating the distances between pairs of landmarks on opposing anterior‐posterior sides of the anterior fontanelle. Figure 7 shows the average change in the distance between the semi‐landmarks in a sagittal plane. The distance slightly increases in the normal growth model, likely secondary to overall skull growth. However, for the SCS growth model, the anterior‐posterior distances decreased. This indicated that the closure of the anterior fontanelle in this direction was greater than the overall parallel growth of the skull. The pace of the fontanelle closure was greatest towards the mid‐sagittal symmetry plane.

Figure 7.

Figure 7

Open in a new tab

Average change in the coronal suture width along seven semi‐landmark pairs.

Discussion

During the first months after the birth, the infant skull goes through a transformation that is unparalleled at any other point in human life: a newborn cranium is approximately 25% of its adult size, and it doubles in size by the first 6 months (Dekaban, 1977). This rapid brain development and intracranial volume expansion cause an infant's skull to expand through the patent sutures between the major cranial bones. The crucial role of major cranial sutures in skull shape and growth is accepted and molecular mechanisms underlying the bone differentiation at the suture sites are being identified by ongoing research (Flaherty et al. 2016).

Virchow's law and bone‐plate theory

Despite the advances made in animal models and molecular experiments, the characteristic shape deformities associated with the sagittal synostosis is still explained by the Virchow principle of 1851 (Persing et al. 1989) which states that ‘the growth is restricted in a plane perpendicular to the fused suture, and enhanced in the corresponding parallel plane’. We did observe a more rapid pace of growth in the antero‐posterior direction in the SCS model compared with normal which is consistent with this principle. Virchow's principle does explain the scaphocephalic elongated and narrow cranial shape associated with premature fusion of the sagittal suture, but it does not explain the subtler phenotypic changes such as occipital constriction, decreased posterior vertical height, and pterion constriction. In 1989, Delashaw et al. observed that asymmetrical bone deposition occurred mainly at perimeter sutures, with growth directed away from the fused bone plate (Delashaw et al. 1989). They also stated: ‘As new radiologic techniques develop to define the configuration of the skull in intricate detail, a skull pattern of growth explaining the pathogenesis of all deformities created by premature fusion of a cranial vault suture may become apparent’. We also found a higher rate of bone growth at the coronal sutures around the anterior fontanelle in the sagittal craniosynostosis model compared with the normal growth model.

Our preliminary analysis with the PCA showed that the first principle component, PC1, was correlated with skull size and patient age. When growth was visualized using only PC1, the dominant direction of change, it also showed a consistent pattern of parietal bone growth around the anterior fontanelle and along the sagittal suture in both cohorts (see Fig. 3). By annotating the ends of the sagittal suture on parietal bones, namely, bregma (bp) and lambda (l) landmarks (Fig. 1, Appendix S1), we were able to capture this dominant pattern across the population.

Premature closure of uninvolved sutures

The bone growth around unfused metopic and coronal sutures was faster in the sagittal craniosynostosis cohort than in the normal cohort. This explains the observation that older infants with sagittal synostosis present with a smaller anterior fontanelle. An interesting notion to consider is the phenomenon of delayed synostosis of uninvolved sutures or even pansynostosis (premature fusion of all sutures) after surgical intervention (Greene, 1998; Jenkins et al. 2013; Yarbrough et al. 2014; Sivakumar et al. 2015). With our findings indicating a tendency to early closure at metopic and coronal sutures in the sagittal synostosis cohort, it is worthwhile to investigate the effects of the surgery and the synostosis on the uninvolved sutures separately.

Statistical growth models vs. finite element analysis

We modeled the calvarial bone growth during the first 6 months of life in infants with isolated sagittal craniosynostosis and infants without any craniofacial abnormalities using a cross‐sectional population approach in a data‐driven way. We were able to create two growth models that confirm the well‐established observations, as well as reveal novel findings about the bone growth at the suture sites.

Human skull growth was also been modeled as a mechanical process that is driven by the expansion of the brain and consequently neurocranium using finite‐element (FE) methods (Margulies & Thibault, 2000; Li et al. 2015; Libby et al. 2017). This approach focuses on material properties of the calvarial bones and sutures, whereas our approach relies on the data that CT exams provide. Furthermore, it models the bones as they are in the template used and does not incorporate the bone growth at the suture sites. The driving force in FE models is the brain growth as a physical force determined via in vivo experiments. The population‐level variance in shape and growth is not part of FE analysis, whereas it is an important piece of data‐driven statistical models. The FE approach allows an analysis of extreme details on one sample, whereas our statistical models enable an analysis at the population‐level.

Manually annotated landmarks vs. dense correspondence

We used cranial landmark positions as the response of our growth models based on age. We included landmarks around the anterior fontanelle, such as four bregma points on the corners of frontal and parietal bones, to better model the overall bone growth along the patent sutures. Although admittedly sparse, selected cranial landmarks provided us with a sufficient topographic coverage of the skull surface to model the overall cranial growth. This approach can be expanded to use dense correspondence between 3D models of skull. The challenge, however, is that infants, and especially non‐craniosynostosis cases, display a wide range of variability in suture shape and closure. This is a challenging case for the automated registration algorithms to find the optimum correspondence without expert guidance. In this study, we established a baseline growth based on expert knowledge, which we plan to refine using a combination of anatomical landmarks and semi‐landmarks automatically detected on all surfaces, which could provide a denser correspondence among samples and help us capture subtle differences in surface areas that are not annotated in the current study.

Frontal bossing and occipital protuberance are common deformities associated with SCS and believed to change over the first 6 months of age with rapid growth of the brain. Our growth models were based on anatomical landmarks that do not cover the regions of the cranial bones that exhibit these deformities. For this reason, we were unable to capture the shape changes that occur on the bone surfaces without anatomical landmarks. We plan to develop a landmark‐free growth model based on dense correspondences between samples. Such models will allow us to interpret growth at sutures and on bone surfaces.

Future directions, strengths and limitations

Our study is the first of its kind to model and compare the skull growth of large normative (n = 117) and sagittal craniosynostosis (n = 81) populations. It is of clinical importance to establish normative templates and growth models that can be used to evaluate the degree of deformity and the surgical outcomes.

Although our dataset is one of the largest in this type of study, it has limitations. The SCS scans we used were for diagnosis and surgical planning as part of a treatment protocol. As in most studies using clinical images, there might be a bias towards more severe cases in early scans, since these patients may have presented earlier. However, the timing of the scan does not depend on the age of presentation or need for surgery and it is essentially randomly distributed between the clinic visit and the surgery.

The bone growth rates and trends we report are population‐level for SCS and normal cohorts, and do not indicate how the cranial bones would have grown in an individual with or without SCS. The growth may be affected by the initial severity of SCS as well as other factors. As in the SCS cohort, there might also be multiple different growth trajectories among unaffected individuals in the normal cohort. Our models show average growth curves for both populations to compare overall growth trends. With larger datasets, it may be possible to detect different growth patterns in SCS and normal populations.

Although our growth curves are population‐level, they provide an important baseline for quantifying the individual shape by accounting for growth. Often the severity is calculated in comparison with normal, yet our work may inform the caregivers about the individual's deviation from the normal and relative position in the SCS population with respect to their age.

We plan to extend our study to other isolated craniosynostoses (such as metopic and unicoronal craniosynostosis) and older age groups. Comparing skull growth and bone deposition patterns at the sutures can provide valuable insight into understanding the etiology of single suture craniosynostosis and guide the surgical treatment.

Conclusions

Sagittal craniosynostosis is the most common single‐suture isolated form of craniosynostosis. The surgical intervention is usually planned in the first year of life, most commonly between 3 and 6 months of age. The brain development and skull growth are at their fastest rate during this period, yet the literature lacks studies with large sample sizes focusing on the shape and growth of the infant skull. In this study, we first generated two anatomical templates, a normative and a sagittal craniosynostosis template, to represent the two distinct populations. Then, using anatomical and geometric landmarks, we modeled the growth of the two populations from birth to 6 months. Finally, with the help of additional semi‐landmarks along the sutures, we compared the bone growth rates between two populations.

We found that the dominant growth pattern in both cases is the overall increase in the skull size and the decrease in the size of the anterior fontanelle in the case of sagittal craniosynostosis population. Our measurements showed that the sagittal craniosynostosis model grows faster in the antero‐posterior direction compared with the control cohort. Additionally, the coronal and metopic sutures close faster in the sagittal craniosynostosis model compared with normative growth.

Conflict of interest

Authors declare no conflict of interest.

Supporting information

Fig. S1. Summary flowchart.

Click here for additional data file. (108.7KB, pdf)

Fig. S2. PC1 scores are highly correlated with age at the time of exam.

Click here for additional data file. (126.5KB, pdf)

Fig. S3. Growth animations for normal and SCS models.

Click here for additional data file. (49.2KB, pdf)

Appendix S1. Landmarks and semi‐landmarks.

Click here for additional data file. (175.2KB, pdf)

Appendix S2. Landmark intra‐rater variances.

Click here for additional data file. (185KB, csv)

Appendix S3. Comparison of growth models.

Click here for additional data file. (88.3KB, pdf)

Data S1. Landmark coordinates and demographics (age and sex) of the samples used in the growth models.

Click here for additional data file. (3.3MB, gif)

Acknowledgements

This research was partially funded by Seattle Children's Research Institute (A.M.M.) and Seattle Children's Guild (R.H.). We thank two anonymous reviewers for their constructive and careful review of our work.

Data availability statement

All landmark data for the study are included in Supporting Information Data S1.

References

  1. Anderson PJ, Netherway DJ, McGlaughlin K, et al. (2007) Intracranial volume measurement of sagittal craniosynostosis. J Clin Neurosci 14, 455–458. [DOI] [PubMed] [Google Scholar]
  2. Avants BB, Tustison NH, Song G (2009) Advanced Normalization Tools: V1.0. Insight J 681 http://hdl.handle.net/10380/3113. [Google Scholar]
  3. Avants BB, Tustison NJ, Song G, et al. (2011) A reproducible evaluation of ANTs similarity metric performance in brain image registration. NeuroImage 54, 2033–2044. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Bergquist CS, Nauta AC, Selden NR, et al. (2016) Age at the time of surgery and maintenance of head size in nonsyndromic sagittal craniosynostosis. Plast Reconstr Surg 137, 1557–1565. [DOI] [PubMed] [Google Scholar]
  5. Bonfield CM, Sharma J, Cochrane DD, et al. (2016) Minimizing blood transfusions in the surgical correction of craniosynostosis: a 10‐year single‐center experience. Childs Nerv Syst 32, 143–151. [DOI] [PubMed] [Google Scholar]
  6. Bookstein FL (1989) Principal warps: thin‐plate splines and the decomposition of deformations. IEEE Trans Pattern Anal Mach Intell 11, 567–585. [Google Scholar]
  7. Bookstein FL (1991) Morphometric Tools for Landmark Data: Geometry and Biology. Cambridge: Cambridge University Press. [Google Scholar]
  8. Clarke CM, Fok VT, Gustafson JA (2018) Single suture craniosynostosis: identification of rare variants in genes associated with syndromic forms. Am J Med Genet A 176, 290–300. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Collett BR, Gray KE, Kapp‐Simon KA, et al. (2013) Laypersons’ ratings of appearance in children with and without single‐suture craniosynostosis. J Craniofac Surg 24, 1331–1335. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Cox TC, Luquetti DV, Cunningham ML (2013) Perspectives and challenges in advancing research into craniofacial anomalies. Am J Med Genet C Semin Med Genet 163, 213–217. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Dekaban AS (1977) Tables of cranial and orbital measurements, cranial volume, and derived indexes in males and females from 7 days to 20 years of age. Ann Neurol 2, 485–491. [DOI] [PubMed] [Google Scholar]
  12. Delashaw JB, Persing JA, Broaddus WC, et al. (1989) Cranial vault growth in craniosynostosis. J Neurosurg 70, 159–165. [DOI] [PubMed] [Google Scholar]
  13. Dvoracek LA, Skolnick GB, Nguyen DC, et al. (2015) Comparison of traditional versus normative cephalic index in patients with sagittal synostosis: measure of scaphocephaly and postoperative outcome. Plast Reconstr Surg 136, 541–548. [DOI] [PubMed] [Google Scholar]
  14. Fearon JA, Ditthakasem K, Herbert M, et al. (2017) An appraisal of the cephalic index in sagittal craniosynostosis, and the unseen third dimension. Plast Reconstr Surg 140, 138–145. [DOI] [PubMed] [Google Scholar]
  15. Fedorov A, Beichel R, Kalpathy‐Cramer J, et al. (2012) 3D Slicer as an image computing platform for the Quantitative Imaging Network. Magn Reson Imaging 30, 1323–1341. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Fischer S, Maltese G, Tarnow P, et al. (2016) Comparison of intracranial volume and cephalic index after correction of sagittal synostosis with spring‐assisted surgery or pi‐plasty. J Craniofac Surg 27, 410–413. [DOI] [PubMed] [Google Scholar]
  17. Flaherty K, Singh N, Richtsmeier JT (2016) Understanding craniosynostosis as a growth disorder. Wiley Interdiscip Rev Dev Biol 5, 429–459. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Garza RM, Khosla RK (2012) Nonsyndromic craniosynostosis. Semin Plast Surg 26, 53–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Governale LS (2015) Craniosynostosis. Pediatr Neurol 53, 394–401. [DOI] [PubMed] [Google Scholar]
  20. Greene CS Jr (1998) Pancraniosynostosis after surgery for single sutural craniosynostosis. Pediatr Neurosurg 29, 127–132. [DOI] [PubMed] [Google Scholar]
  21. Heller JB, Heller MM, Knoll B, et al. (2008) Intracranial volume and cephalic index outcomes for total calvarial reconstruction among nonsyndromic sagittal synostosis patients. Plast Reconstr Surg 121(1, 187–195. [DOI] [PubMed] [Google Scholar]
  22. Heuzé Y, Holmes G, Peter I, et al. (2014) Closing the gap: genetic and genomic continuum from syndromic to nonsyndromic craniosynostoses. Curr Genet Med Rep 2, 135–145. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Jenkins GH, Smith NR, McNeely PD (2013) Pancraniosynostosis following endoscope‐assisted strip craniectomy and helmet orthosis for sagittal suture craniosynostosis in a nonsyndromic patient. J Neurosurg Pediatr 12, 77–79. [DOI] [PubMed] [Google Scholar]
  24. Kapp‐Simon KA, Speltz ML, Cunningham ML, et al. (2007) Neurodevelopment of children with single suture craniosynostosis: a review. Childs Nerv Syst 23, 269–281. [DOI] [PubMed] [Google Scholar]
  25. Li Z, Park BK, Liu W, et al. (2015) A statistical skull geometry model for children 0‐3 years old. PLoS One 10, e0127322. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Libby J, Marghoub A, Johnson D, et al. (2017) Modelling human skull growth: a validated computational model. J R Soc Interface 14, pii: 20170202. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Lloyd MS, Buchanan EP, Khechoyan DY (2016) Review of quantitative outcome analysis of cranial morphology in craniosynostosis. J Plast Reconstr Aesthet Surg 69, 1464–1468. [DOI] [PubMed] [Google Scholar]
  28. Lorensen WE, Cline HE (1987) Marching cubes: a high resolution 3D surface construction algorithm. Comput Graph 21, 163–169. [Google Scholar]
  29. Mann RJ, Fahrenkopf MP, Burton M, et al. (2017) The directive growth approach for nonsyndromic, unicoronal craniosynostosis: patient and clinical outcomes. J Craniofac Surg 28, 2108–2112. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Margulies SS, Thibault KL (2000) Infant skull and suture properties: measurements and implications for mechanisms of pediatric brain injury. J Biomech Eng 122, 364–371. [DOI] [PubMed] [Google Scholar]
  31. McEwan TW, Martin AL, Tanaka T, et al. (2016) Evaluating children with metopic craniosynostosis: the cephalic width‐intercoronal distance ratio. Cleft Palate Craniofac J 53, e95–e100. [DOI] [PubMed] [Google Scholar]
  32. Netherway DJ, Abbott AH, Anderson PJ, et al. (2005) Intracranial volume in patients with nonsyndromal craniosynostosis. J Neurosurg 103(2 Suppl), 137–141. [DOI] [PubMed] [Google Scholar]
  33. Persing JA (2016) Discussion: age at the time of surgery and maintenance of head size in nonsyndromic sagittal craniosynostosis. Plast Reconstr Surg 137, 1566–1567. [DOI] [PubMed] [Google Scholar]
  34. Persing JA, Jane JA, Shaffrey M (1989) Virchow and the pathogenesis of craniosynostosis: a translation of his original work. Plast Reconstr Surg 83, 738–742. [DOI] [PubMed] [Google Scholar]
  35. Schlager S 2017. Morpho and Rvcg–Shape Analysis in R: R‐Packages for geometric morphometrics, shape analysis and surface manipulations In: Statistical Shape and Deformation Analysis (eds Guoyan Z, Shuo L, Gabor S.), pp. 217–256, Cambridge: Academic Press. [Google Scholar]
  36. Schroeder WJ, Lorensen B, Martin K (2004) The Visualization Toolkit: An Object‐Oriented Approach to 3D Graphics. New York: Kitware. [Google Scholar]
  37. Seeberger R, Hoffmann J, Freudlsperger C, et al. (2016) Intracranial volume (ICV) in isolated sagittal craniosynostosis measured by 3D photocephalometry: a new perspective on a controversial issue. J Craniomaxillofac Surg 44, 626–631. [DOI] [PubMed] [Google Scholar]
  38. Sgouros S, Hockley AD, Goldin JH, et al. (1999) Intracranial volume change in craniosynostosis. J Neurosurg 91, 617–625. [DOI] [PubMed] [Google Scholar]
  39. Sivakumar W, Goodwin I, Blagg R, et al. (2015) Pancraniosynostosis following endoscopic‐assisted strip craniectomy for sagittal suture craniosynostosis in the setting of poor compliance with follow‐up: a case report. J Med Case Rep 24, 9–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Speltz ML, Kapp‐Simon KA, Cunningham M, et al. (2004) Single‐suture craniosynostosis: a review of neurobehavioral research and theory. J Pediatr Psychol 29, 651–668. [DOI] [PubMed] [Google Scholar]
  41. Speltz ML, Collett BR, Wallace ER, et al. (2015) Intellectual and academic functioning of school‐age children with single‐suture craniosynostosis. Pediatrics 135, e615–e623. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Starr JR, Lin HJ, Ruiz‐Correa S, et al. (2010) Little evidence of association between severity of trigonocephaly and cognitive development in infants with single‐suture metopic synostosis. Neurosurgery 67, 408–415. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Stricker PA, Fiadjoe JE (2014) Anesthesia for craniofacial surgery in infancy. Anesthesiol Clin 32, 215–235. [DOI] [PubMed] [Google Scholar]
  44. Tovetjärn RC, Maltese G, Wikberg E, et al. (2014) Intracranial volume in 15 children with bilateral coronal craniosynostosis. Plast Reconstr Surg Glob Open 2, e243. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Tunçbilek G, Vargel I, Erdem A, et al. (2005) Blood loss and transfusion rates during repair of craniofacial deformities. J Craniofac Surg 16, 59–62. [DOI] [PubMed] [Google Scholar]
  46. Yarbrough CK, Smyth MD, Holekamp TF, et al. (2014) Delayed synostoses of uninvolved sutures after surgical treatment of nonsyndromic craniosynostosis. J Craniofac Surg 25(1), 119–123. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Fig. S1. Summary flowchart.

Click here for additional data file. (108.7KB, pdf)

Fig. S2. PC1 scores are highly correlated with age at the time of exam.

Click here for additional data file. (126.5KB, pdf)

Fig. S3. Growth animations for normal and SCS models.

Click here for additional data file. (49.2KB, pdf)

Appendix S1. Landmarks and semi‐landmarks.

Click here for additional data file. (175.2KB, pdf)

Appendix S2. Landmark intra‐rater variances.

Click here for additional data file. (185KB, csv)

Appendix S3. Comparison of growth models.

Click here for additional data file. (88.3KB, pdf)

Data S1. Landmark coordinates and demographics (age and sex) of the samples used in the growth models.

Click here for additional data file. (3.3MB, gif)

Data Availability Statement

All landmark data for the study are included in Supporting Information Data S1.

Articles from Journal of Anatomy are provided here courtesy of Anatomical Society of Great Britain and Ireland

RESOURCES