A Fractal Analysis of Human Cranial Sutures

JACK C. YU, M.D., D.M.D.

RONALD L. WRIGHT
MATTHEW A. WILLIAMSON, PH.D.
JAMES P. BRASELTON III, M.S.
MARTHA L. ABELL, PH.D.

Objectives: Many biological structures are products of repeated iteration

functions. As such, they demonstrate characteristic, scale-invariant features.
Fractal analysis of these features elucidates the mechanism of their formation.
The objectives of this project were to determine whether human cranial sutures
demonstrate self-similarity and measure their exponents of similarity (fractal
dimensions).
Design: One hundred three documented human skulls from the Terry Col-
lection of the Smithsonian Institution were used. Their sagittal sutures were
digitized and the data converted to bitmap images for analysis using box-
counting method of fractal software.
Results: The log-log plots of the number of boxes containing the sutural
pattern, Nr, and the size of the boxes, r, were all linear, indicating that human
sagittal sutures possess scale-invariant features and thus are fractals. The
linear portion of these log-log plots has limits because of the finite resolution
used for data acquisition. The mean box dimension, Db, was 1.29289 6
0.078457 with a 95% confidence interval of 1.27634 to 1.30944.
Conclusions: Human sagittal sutures are self-similar and have a fractal di-
mension of 1.29 by the box-counting method. The significance of these find-
ings includes: sutural morphogenesis can be described as a repeated iteration
function, and mathematical models can be constructed to produce self-similar
curves with such Db. This elucidates the mechanism of actual pattern forma-
tion. Whatever the mechanisms at the cellular and molecular levels, human
sagittal suture follows the equation log Nr 5 1.29 log 1/r, where Nr is the number
of square boxes with sides r that are needed to contain the sutural pattern and
r equals the length of the sides of the boxes.

KEY WORDS: fractal analysis, fractal dimension, human sagittal suture

This report presents the result of a fractal analysis of human
sagittal sutures using documented dry skull specimens. In 1999
a theoretical model of cranial growth and adaptation was pro-
posed by Yu et al. This crude but effective model explained
how growth of different parts of the cranium can be synchro-
nized solely by the expansion of the brain. Such coordinated,
synchronized growth of the cranial bones has been recognized
Dr. Yu is the Director of the Craniofacial Center at the Medical College of
Georgia, Augusta, Georgia. Mr. Wright is currently a graduate student in the
Department of Anthropology at the University of Florida, Gainesville, Florida.
Dr. Williamson is an Assistant Professor in the Department of Health and Ki-
nesiology, Mr. Braselton is an Assistant Professor of Mathematics, and Dr.
Abell is a Professor of Mathematics at Georgia Southern University, Statesboro,
Georgia.
This work was presented at the 58th Annual Meeting of the American Cleft
Palate–Craniofacial Association; Minneapolis, Minnesota; April 23 to 28, 2001.
Submitted March 2002; Accepted August 2002.
Address correspondence to: Dr. Jack Yu, Section of Plastic Surgery, De-
partment of Surgery, Medical College of Georgia, Augusta, Georgia 30912-
4080. E-mail jyu@mail.mcg.edu.
for many decades. Researchers have introduced different terms
to describe this growth. For example, Enlow (1975) used the
phrase architectonic equilibrium, and Hanken and Hall (1993)
called it the epigenetic cascade. At the heart of this cranial
growth model is the repeated iteration function, also known as
the recursive function, a special form of mathematical opera-
tions in which a rule or set of rules (iterators) are repeatedly
applied such that each output becomes the input for the next
iteration (Yu et al., 2001b). When a system with many inter-
acting elements undergoes repeated iterations of a set of rules,
the results are optimization of some parameters pertaining to
these elements as defined by the rules. Very often there is
generation of fractals, an example of such being the branching
morphogenesis of the tracheal-bronchial tree (Bassingthwaite
et al., 1994). The products of these recursive functions can
manifest as either temporal or spatial fractal patterns.
The word fractal, first coined by Benoit Mandelbrot in the
1970s and referred to by Kauffman (1993), means break or
fragment. Fractal objects have self-similar features, that is, re-
peating details emerge as the scales change, with parts resem-

409
410 Cleft Palate–Craniofacial Journal, July 2003, Vol. 40 No. 4
FIGURE 1 Example of a Von Koch curve. This curve is self-similar and
has a fractal dimension of 1.26. Magnifying any part of the curve will
produce the same features, which is the essence of scale invariance.
bling the whole, such as Sierpinski’s triangle or Von Koch
curve (Fig. 1). Mathematically, fractal objects are defined as
objects having a greater Hausdorff-Besicovitch dimension, or
box-counting dimension, than their topological or Euclidean
dimensions. Fractal objects are often found in circumstances
that demand fitting a larger total surface area within a smaller
constrained boundary perimeter. An excellent example of this
is the convolution of the cerebral cortex of the human brain.
The universe itself is fractal; from the distribution of the gal-
axies to the microporosities of a speck of limestone, fractal
features such as self-similarity, scale-invariance, and self-or-
ganized criticality are found everywhere (Mandelbrot, 1983).
Simply stated, self-similarity is the resemblance of a small
fraction of the whole to the entire structure or process. Those
characteristics that persist across different scales of observa-
tion are known as scale invariant (e.g., a small piece of broc-
coli resembling a larger piece of broccoli). Another important
property of a system with many interacting elements repeat-
edly iterating a set of rules is that whole system behavior is
more than the sum total of all the parts; in other words, there
exists nonlinearity. Because of the nonlinear nature of these
processes, they spontaneously drive themselves toward some
critical points (in which sudden transition in systems behavior
occur) in a deterministic but unpredictable manner, a phenom-
enon referred to as self-organized criticality.
Euclidean dimension forces on the analysis of natural ob-
jects an artificially applied approximation, as Mandelbrot
(1983) put it: ‘‘Clouds are not spheres and mountains are not
cones. Fractal objects should be analyzed using fractal analysis
. . . .’’ Fractal analysis of mammalian cranial sutures first ap-
peared in the 1980s (Long, 1985) and continued over the en-
suing decades. However, most of these previous studies relied
on hand calculation, which limits the number of data points
on the log-log curve plotting the logarithm of the number of
boxes needed to contain the sutural pattern against the loga-
rithm of the side length of the boxes (Long, 1985; Long and
Long, 1992). More recent work by Russell and Thomason
(1993) proposed that the fractal dimension is a function of the
local stress pattern. They hypothesized that the higher the
stress, the more complex the sutural morphology. For example,
the frontoparietal suture in a male white-tailed deer (Odoco-
ileus virginianus) is more elaborate, compared with a female
white-tailed deer because of the increase in stress from the
male deer antlers’ hypermorphosis. This proposed positive cor-
relation between local stress and the fractal dimension of the
cranial suture was independently supported in a comparative
analysis of cranial suture complexity among different species
of the South American Caimen. Caiman latirostris, the species
with the hardest diet, was found to have the highest fractal
dimension in the three species investigated (Monteiro and Les-
sa, 2000).
If human cranial sutural morphogenesis is indeed similar to
a repeated iteration function in that a set of rules (genetically
determined) are repeatedly applied to allow the cells in the
sutures to react and interact with the intra- and extracranial
environmental perturbations (epigenetically determined), then
it should be possible to detect fractal objects in the end output.
Several years ago the possibility that human cranial sutures
possess self-similar qualities was informally tested in a pilot
study by tracing out the lambdoidal sutures on three dry human
skull specimens on transparent acetate paper that was super-
imposed on a square grid (Pashley et al., 2002). The logarithm
of Nr, the number of squares of length r occupied by the sutural
pattern, was plotted against the logarithm of 1/r. A straight line
was obtained. The slope of this straight line represented the
box dimension, which is an estimator of fractal dimension. The
box dimension is sometimes referred to as the Hausdorff-Be-
sicovitch dimension. It was determined that the box dimension
of the human lambdoidal suture was indeed greater than the
topological dimension (Pashley et al., 2002). Therefore, the
purpose of this study was to expand on that pilot investigation
and answer these three questions: (1) are human sagittal suture
patterns fractal; (2) what is their fractal dimension; and (3)
what can we infer about the underlying mechanism or the pro-
cess that produced these sutures in the first place?
SUBJECTS AND METHODS
One hundred three dry human skulls comprised of 53 males
and 50 females of African (n 5 56) or European (n 5 47)
ancestry, aged 14 to 60 years, were selected from the Terry
Collection at the Smithsonian Institution in Washington, DC.
Only sagittal sutures that were completely open were used;
sutures with partial fusion were not digitized. Utilizing a Pol-
hemus 3Draw Pro model, electromagnetic, three-dimensional
digitizing pen (Polhemus Incorporated, Colchester, VT), the
entire ectocranial aspect of the sagittal suture was traced from
lambda to bregma at a scale of 0.2 mm. The resulting data
were stored in a database as x, y, z coordinates and reconsti-
tuted in the three-dimensions by graphing the coordinates of
each data point using Mathematica (Wolfram Research, Inc.,
Champaign, IL). Each datum point was also projected to the
 Yu et al., FRACTAL ANALYSIS OF HUMAN CRANIAL SUTURES 411

TABLE 1 The Complete Listing of the Terry Collection TABLE 1 Continued

Identification Number, Race, Sex, and Age and the Sagittal
Suture Box-Counting Dimension (D) Specimen Race Sex Age D

1424 WH M 51 1.34442
Specimen Race* Sex* Age D
1481 WH M 54 1.36254
139R WH F 56 1.27293 1506 BL M 26 1.33822
335 WH M 18 1.43658 1507 BL F 23 1.35089
525 WH M 49 1.34646 1539 BL M 23 1.12404
529 BL F 37 1.32783 1544 BL F 23 1.28476
540 BL M 40 1.29605 1553 BL F 30 1.18061
541 BL F 28 1.27675 1563 WH F 29 1.20896
561 BL F 22 1.12936 1571 WH F 55 1.36239
562 BL F 17 1.37122 1572 WH F 45 1.27570
563 WH M 41 1.26303 1591 WH M 19 1.34547
565 BL M 32 1.22827 1594 WH F 40 1.24335
566 WH M 40 1.35203 1599 WH F 41 1.40929
568 BL F 27 1.06788 1600 BL F 26 1.17175
570 BL M 37 1.09270 1602 WH M 20 1.34160
574 BL M 35 1.38477 1607 WH M 32 1.21847
591 WH M 28 1.39704 1611 WH F 59 1.40362
592 BL M 25 1.22745 1617 WH F 35 1.14515
595 BL M 25 1.33257 1624 WH F 58 1.13098
614 WH M 50 1.29552 11R BL F 24 1.27427
676 BL M 46 1.27243 1289R BL F 28 1.38720
680 WH F 30 1.27634 1270R WH F 59 1.39326
719 BL M 25 1.25880 13R BL M 28 1.27155
760 BL M 19 1.25929 1482R WH F 35 1.40103
782 BL M 30 1.31759 161R WH F 50 1.38599
815 BL F 32 1.24656 187R WH M 29 1.47408
824 BL F 30 1.24445 225R WH F 55 1.21038
830 BL M 28 1.25485 256R WH F 50 1.38132
847 WH F 39 1.38411 311R WH M 27 1.20616
850 BL M 23 1.23709 351R WH F 30 1.42672
868 WH M 60 1.48340 405R WH F 34 1.25911
876 BL M 22 1.36455 41R WH F 41 1.34123
886 BL F 23 1.25261 612R WH F 60 1.38485
897 WH M 43 1.27859 64R WH F 57 1.26114
898 BL M 22 1.29651 69R WH F 52 1.21121
905 BL M 31 1.20557
* WH 5 white, BL 5 black, F 5 female; M 5 male.
915 BL M 27 1.19882
920 BL F 36 1.35970
926 BL F 23 1.29395
929 BL F 20 1.26876 xy-plane (two-dimensions) by graphing each the x and y co-
932 BL M 27 1.29185 ordinates of each datum point with Mathematica (Wolfram Re-
937 BL M 50 1.33633
949 BL F 24 1.40396 search). Each two-dimensional graphic was inverted (black to
970 BL M 21 1.22427 white, white to black) and saved as a bitmap file for analysis
983 WH F 30 1.28463 using a commercially available software program called Benoit
994 BL F 30 1.23571
1006 BL F 27 1.39021 (Trusoft, Inc., Tampa, FL). Benoit processed the inverted two-
1010 BL F 25 1.20597 dimensional bitmap files to compute the box-counting dimen-
1020 BL M 24 1.24048 sion, Db, also known more simply as the box dimension. This
1023 WH M 20 1.37788
1076 BL F 26 1.44519 was done by using progressively reduced sizes of the boxes in
1105 BL F 22 1.27625 a non-rotating square grid and counting the minimum number
1148 BL M 23 1.21682 of boxes containing part of the suture.
1165 BL M 26 1.30703
1186 WH F 44 1.28528 A nonrotating grid format was selected because of the an-
1187 BL M 21 1.21231 isotropic nature of the sutural pattern. The process is similar
1195 BL M 29 1.44737 to the hand calculations, except Benoit can generate approxi-
1225 WH M 59 1.23618
1232 WH M 57 1.32707 mately 10 times as many refinements (at least 50 data points)
1239R WH F 52 1.22564 per suture than the five refinements in hand calculations and
1271 WH M 58 1.03750 in a much shorter time. The slope of the graph plotting the
1274 WH M 56 1.23058
1286 WH M 50 1.33708 logarithm of the number of boxes needed to contain a given
1287 BL F 24 1.33757 suture pattern against the logarithm of the side length of the
1300 BL M 28 1.32507 boxes is referred to as the box-counting dimension, Db. These
1351 BL F 24 1.24387
1354 BL F 25 1.37791 box-counting dimensions were then statistically processed us-
1363 BL F 14 1.19994 ing Mathematica (Wolfram Research). Benoit (Trusoft) also
1383 WH M 55 1.29370 permitted the determination of Hurst exponents, H, which is
1413 BL F 27 1.37349
an indicator of autocorrelation of the data set being analyzed.
412 Cleft Palate–Craniofacial Journal, July 2003, Vol. 40 No. 4

FIGURE 2 The scattering of sagittal sutural dimension when plotted

against age. There is no concentration along the line of identity in agree-
ment with the analysis of variance result, indicating a lack of significant
correlation between age and Db.

H of any given set of data expressed in the xy-plane is cal-

culated by measuring the variation in the y values (dependent
variable) for a large number of different x ranges (independent
variable). This analysis is also known as the rescaled range
analysis. For example, a random process such as a one-dimen-
sional Brownian walk has a Hurst exponent of 0.5, and be-
cause the maximum possible fractal dimension for a planar
FIGURE 3 The human sagittal sutures. The top three are high Db su-
tracing is 2, the equation Db 1 H 5 2 allows for the easy tures, and the lower three are examples of low Db sutures. Posterior is
conversion between H and Db. Thus, a fractal dimension of toward left. They are visibly different in terms of how much serration the
1.5 means the process that generated the original tracing might, suture contains. Db is a quantitative measure of this qualitative feature.
but not necessarily, be a stochastic process. However, if the
fractal dimension is not 1.5, then the process cannot be com-
pletely stochastic (Glenny et al., 1991).

RESULTS

Of the 103 skulls analyzed, the mean box dimension was

1.29289 with SD of 0.078457, median was 1.28528, and var-
iance was 0.007694. The results of the 103 skulls together with
their sex, age, and race are listed in Table 1. Figure 2 shows
the scattering of the sutural fractal dimension when plotted
against age, indicating a lack of correlation. Figure 3 shows
three examples of high Db sutures and three examples of low
Db sutures. There is a visible, qualitative difference in the com-
plexity of the patterns. Selected sutures with high Db were
further analyzed with the above-mentioned rescaled range
analysis. The transverse (mediolateral) variation, Xn, was plot-
ted against n, where n 5 number of pixels along the suture
and X 5 the amplitude of oscillation in the transverse plane.
This transformation allows for the decomposition of a suture
into more fundamental oscillations that together make up the
suture. An example of this rectified suture tracing is shown in
Figure 4. This was then used for two additional analyses: pow- FIGURE 4 A rectified sagittal suture. The graph was generated by plot-
er spectral density and pseudophase space trajectory. Figure 5 ting Xn against n, where X represents the amplitude of the mediolateral
shows an example of power spectral density and Figure 6 oscillation and n represents the sequential number of the pixel.
 Yu et al., FRACTAL ANALYSIS OF HUMAN CRANIAL SUTURES
FIGURE 5 Power spectral analysis plotting power spectral density
against individual frequencies. With a relatively dominant peak at fre-
quency of 0.01 (arrow) the corresponding period is in the range of 102, a
value used in constructing Figure 6.
shows the pseudophase space trajectory using the appropriate
lag, Dn, which was calculated based on the most dominant
frequency from the power spectral analysis. The resulting
Poincaré map clearly demonstrates the presence of an attractor,
confirming the nonrandom nature of the sutural pattern sug-
gested by Db (Fig. 6).
The results of analysis of variance using Db as the dependent
variable and age, sex, and race as independent variables indi-
cate that none of the independent variables had a significant
effect on Db (p . .05). Frequency distribution analysis of the
entire cohort is shown in Figure 7, and, although none of the
independent variables had an effect on Db, a bimodal distri-
bution is evident. However, the cause of this bimodal distri-
bution is not immediately obvious.
DISCUSSION
Sutural homeostasis is of critical importance for the normal
and rapid expansion of the cranium to occur from the initial
appearance of the sutures in utero to the cessation of the
brain’s expansion. The fact that cranial growth is always in
pace with the brain, no more and no less, is a form of opti-
mization. If the rate of brain growth is expressed as dBr/dt and
the rate of cranial growth is dC/dt, the above observation can
be summarized by dBr/dt 2 dC/dt 5 0. If dBr/dt exceeds dC/
dt for a period of time, intracranial pressure will increase be-
413
FIGURE 6 Pseudophase space portrait of a high Db suture. There is a
clear nonrandom distribution of the trajectory supporting the conclusion
based on the rescaled range analysis that the mechanism of this suture
pattern generation is not stochastic.
cause of the high bulk compressive modulus of the brain and
the high tensile modulus of the cranium. This becomes partic-
ularly important during the late third trimester when dBr/dt
reaches nearly 1000 mg/hour (Potter and Craig, 1975). On the
other hand, if dC/dt exceeds dBr/dt for a given period of time,
there will be unoccupied intracranial space. The dC/dt can be
further expressed by the rate of growth of the two components
that make up the cranium: sutures (S) and bone (B). Thus, dC/
dt 5 dS/dt 1 dB/dt. Because the only suture that normally
fuses during the period when brain growth is still taking place
is the metopic suture, the sum total of all the other sutures can
be considered homeostatic. That is, the sutures remain patent
with a constant physical dimension between the two opposing
cranial bone plates, thus dS/dt 5 0. The biological mechanisms
responsible for this tissue homeostasis are being actively mod-
FIGURE 7 The frequency distribution Db. There is a bimodal appearance
to this data set which may reflect the noncontinuous nature of sutural
morphology because of thresholding that produces this form of epigenetic
polymorphism.
414 Cleft Palate–Craniofacial Journal, July 2003, Vol. 40 No. 4
eled and experimentally verified. Several antagonistic cou-
plings within the sutures have recently been confirmed: Twist
and fibroblast growth factor receptor 2 (Johnson et al., 2000),
Tbx2, and connexin 43 (Chen et al., 2001), noggin and bone
morphogenic proteins (Warren et al., 2002). These signaling
peptides provide the means for a closed-loop control system
at the transcriptional level, which has the mature patent suture
as an end product.
By critically examining the characteristic of the sutures, im-
portant information can be extracted. One such parameter is
how serrated the suture is, quantitatively measured by Db. Fig-
ure 2 plots all 103 Db measured against the ages of the skulls.
The lack of correlation between the ages of the skulls and their
sagittal suture fractal dimensions is graphically depicted. It is
conceivable that in some special conditions, such as hydro-
cephalus, the fractal dimension will increase as time progresses
because of the persistent high local strain. Such studies are
currently being designed and conducted. In this study of the
human sagittal sutures, the anterior-posterior axis (from breg-
ma to lambda) is assigned the x-axis, and the medial-lateral
variations are expressed in the corresponding y-axis (Fig. 3).
Variations in the sagittal plane (i.e., the z-axis data) were not
included to simplify the calculations. For all tracings generated
by truly stochastic processes, the fractal dimension would be
1.5 with their Hurst exponents equal to 0.5, indicating the lack
of both persistence and antipersistence within the data set. If,
on the other hand, the data points are not independent, the
deviation from H 5 0.5 tells whether a more positive y value
at x 5 a is more likely to have a more negative y value (an-
tipersistence) or a more positive y value (persistence) at x 5
a 1 1. This type of analysis is known as autocorrelation. Based
on the 103 skulls studied, the human sagittal sutures demon-
strated definite antipersistence with H 5 0.70711. In other
words, sagittal suture morphology demonstrated memory. It is
entirely possible that not only does H differ between different
sutures but also within each individual suture. In the case of
sagittal sutures, the more anterior portion (closer to bregma)
indeed appears to have less medial/lateral variation amplitude.
Future studies will have to address this quantitatively.
Whatever the mechanisms that produced these sagittal su-
ture patterns, each one being unique, the end output must obey
a simple rule. This simple rule is that log Nr 5 1.29 log 1/r,
where Nr is the number of square boxes with sides of length
r, which are needed to contain the sutural pattern and r equals
to the length of the sides of the square boxes. An understand-
ing of this is the focus of our current investigative effort. Even
though this study failed to show sexual dimorphism, others
have reported, at least in some human craniofacial sutures, the
fractal dimension can be used for reliable morphognostic sex
determination (Schiwy-Bochat, 2001).
CONCLUSION
One hundred three documented dry human skull specimens
were analyzed for fractal dimension calculations by the box-
counting method. The Db was 1.29289 for the group with SD
of 0.078457. The 95% confidence interval was 1.27634 to
1.30944. Thus, with a high degree of confidence, it appears
that human sagittal sutures are fractal and can be described in
terms of repeated iteration functions. The underlying mecha-
nism that produced these oscillations is not random. Because
the Hurst exponent is greater than 0.5, sagittal suture patterns
have long-range correlation: the systems demonstrate memory.
More importantly, this type of analysis allows for the quanti-
tative characterization of sutural morphology and permits the
testing of future hypotheses. One such hypothesis is that clo-
sure of the coronal suture increases tensile strain across the
sagittal suture. This can be tested by comparing the sagittal
suture fractal dimension between normal specimens and those
showing coronal craniosynostosis (if the local-regional bio-
mechanical strain data can be correlated with the fractal di-
mension of the suture). Efforts are under way to confirm or
refute this conjecture of strain-induced Db alteration as well as
the potential mechanisms to increase the information dimen-
sion, such as oscillation of intracellular calcium ion concen-
tration, that is necessary during the mechanotransduction pro-
cess (Yu et al., 2001a).
Sutures are records of dynamic interactions of past events.
Detailed fractal analysis permits the numerical characterization
of the end product of this complex and robustly adaptive sys-
tem. The next task is to determine how such macroscopic frac-
tal curves are generated from the coordinated antagonistic in-
teractions between osteoblastic and osteoclastic mechanisms
based on the genetic and environmental inputs.
Acknowledgments. This study was supported, in part, by a grant from the
American Society of Maxillofacial Surgeons and the Ronald E. McNair Foun-
dation. The authors also wish to thank the Smithsonian Institution for the use
of the Terry Collection.
REFERENCES
Bassingthwaite JB, Leibovitch LS, West BJ. The surprising statistical properties
of fractals. In: Bassingthwaite JB, Leibovitch LS, West BJ, eds. Fractal
Physiology. London: Oxford University Press; 1994:33–44.
Chen J, Zhong Q, Wang J, Cameron RS, Borke JL, Isales CM, Bollag RJ.
Microarray analysis of Tbx2-directed gene expression: a possible role in
osteogenesis. Mol Cell Endocrinol. 2001;177:43–54.
Enlow DH. Introductory concepts of the growth process. In: DH Enlow, ed.
Handbook of Facial Growth. Philadelphia: Saunders Co.; 1975:18–47.
Glenny RW, Robertson HT, Yamashiro S, Bassingthwaighte JB. Application of
fractal analysis to physiology. J Appl Physiol. 1991;70:2351–2367.
Hanken J, Hall BK. Mechanisms of skull diversity and evolution. In: J Hanken,
BK Hall, ed. The Skull. Vol. 3: Functional and Evolutionary Mechanisms.
Chicago: University of Chicago Press; 1993:1–26.
Johnson D, Iseki S, Wilkie AOM, Morris-Kay GM. Expression patterns of
Twist and Fgfr1, -2 and -3 in the developing mouse coronal suture suggest
a key role for Twist in suture initiation and biogenesis. Mech Dev. 2000;91:
341–345.
Kauffman SA. Self-organization and adaptation in complex systems. In: SA
Kauffman, ed. The Origins of Order. New York: Oxford University Press;
1993:175–181.
Long CA. Intricate sutures as fractal curves. J Morphol. 1985;185:285–295.
Long CA, Long JE. Fractal dimension of cranial sutures and waveforms. Acta
Anat. 1992;145:201–206.
Mandelbrot BB. The Fractal Geometry of Nature. New York: WH Freeman
and Co.; 1983.
 Yu et al., FRACTAL ANALYSIS OF HUMAN CRANIAL SUTURES
Monteiro LR, Lessa LG. Comparative analysis of cranial suture complexity in
the genus Caiman (Crododylia alligatoridae). Rev Bras Biol. 2000;60:689–
694.
Pashley DH, Borke JL, Yu JC. Biomechanics and craniofacial morphogenesis.
In: Lin K, Ogle RC, Jane JA, eds. Craniofacial Surgery: Science and Sur-
gical Technique. Philadelphia: Saunders Co.; 2002:84–100.
Potter EL, Craig JM. Rate of antenatal growth. In: Potter EL, Craig JM, eds.
Pathology of the Fetus and the Infant. 3rd ed. Chicago: Year Book Medical
Publishers, Inc.; 1975:15–24.
Russell AP, Thomason JJ. Mechanical analysis of the mammalian head skeleton.
In: Hanken J, Hall BK, eds. The Skull: Vol. 3: Functional and Evolutionary
Mechanisms. Chicago: University of Chicago Press; 1993:345–383.
Schiwy-Bochat KH. The roughness of the supranasal region—a morphological
sex trait. For Sci Int. 2001;117:7–13.
415
Warren SM, Brunet LJ, Economides AN, Harland RM, Fong KD, Longaker
MT. Apert and Crouzon fibroblast growth factor receptor gain-of-function
mutations suppress noggin: a mechanism for FGF-mediated cranial suture
fusion. In: Proceedings of the 47th Plastic Surgery Research Council. Han-
over, NH: Plastic Surgery Research Council; 2002:93.
Yu JC, Lucas JH, Borke JL. Tension-induced FGF-2 release in neonatal rat
coronal sutures: a self-organizing mechanism in cranial development. J
Craniofac Surg. 1999;8:29–32.
Yu JC, Braselton JP, Abell ML, Borke JL. The identification of calcium oscil-
lators in immature rat cranial sutures. In: Lauritzen CGK, ed. Craniofacial
Surgery. Bologna: Monduzzi Editore, 2001a;9:139–141.
Yu JC, Lucas JH, Fryberg K, Borke JL. Extrinsic tension results in FGF-2
release, membrane permeability change, and intracellular Ca11 increase in
immature cranial sutures. J Craniofac Surg. 2001b;12:391–398.