Está en la página 1de 4

A revised technique to calculate the

fractal dimension of cranial sutures

Josep M. ARQUÉS 1 Josep GIBERT2


`Universitat Autónoma de Barcelona.
Departament d'Informática.
Combinatória i Comunicació Digital.
08193 Bellaterra. Spain
'Institut de Paleontologia: Dr M. Crusafont.
Escola Industrial 23. 08201 Sabadell. Spain

ABSTRACT

Fractal analysis was applied in human palaeontology by Gibert and Palmqvist (1995) to estimate the value
of the fractal dimension obtained from the cranial sutures preserved in the fragment of occipital and parietals
bone (VM-0) found at the Venta Micena site. This paper also estimated the values of fractal dimension for
different specimens in order to establish a taxonomy. Although that initial study demonstrated that the technique
could be useful in human palaeontology, the large variability of sutures observed in the VM-0 sample requires
a mechanism that makes it possible to automatically obtain an objective plot of the suture to be analyzed.
Keywords: fractal dimension, "box counting" algorithm.

INTRODUCTION

In a generic sense, fractal dimension is a number which indicates the measure of regularity in a natural
object or geometric set. As Gibert and Palmqvist (1995) demonstrated, this value is useful to compare the
"content" of different cranial sutures.
In this paper we propose a new technique to improve the method described by Gibert and Palmqvist (1995),
computerise it, and attempt to reduce and limit the errors produced in the process of obtaining and processing
the image as well as the calculation of the fractal dimension. The results of our work can be seen in table 1.

ESTIMATION OF FRACTAL DIMENSIONS

To find the fractal dimension a "box counting" algorithm (Liebovitch and Toth, 1989) was used instead of
the "division method" (Gibert and Palmqvist 1995).
An alternative measure of the fractal dimension is the capacity

log NB(E)
dB = hm
E0 log (1/E)

This is called "box counting" because one counts the nunimal number of boxes, N B , that covers a curve
for boxes of size E.
To calculate the fractal dimension, Ver. 0.3 of the computer program FD3 developed in 1992 by John
Sarraille and Peter DiFalco (based on the "box counting" method) was used. This algorithm is parti-cularly useful
when the fractal dimension is low (as in our case). It is quite fast and it has a low complexity, 0(n log n), (where
n indicates the number of points used).

MATERIAL AND METHODS

Our experimenta were carried out using a digitalization obtained from the original models at a scale of 1:1
(the sutures were plotted using methods typically employed in palaeontology). Digitalization was performed using

83
ARQuÉs, J. M. & GIBERT, J.

a grey scale at 150 lpi and the PhotoStyler 2.0 software. This software was also used to convert the digitalization
to a binary image and 8-bit TIFF format. An aspect not taken into consideration in the measures ohtained
hitherto (Gibert and Palmqvist, 1995) was the choice of suture points. The following image (see figure 1.A)
shows a large section of a digitalized suture.
The method employed to calculate the fractal dimension requires only one point for each grid line.
Nevertheless, it can be seen that, for example, there are several points to choose from on Line 1 of the previous
grid pattern. The choice of points is not trivial since, when making the choice and regardless of the algorithm
used, an error is always produced in the measurements calculated to date, which was not taken into account
(Gibert and Palmqvist, 1995) and which could affect the precision of the fractal dimension calculation. Moreover,
since the way the points were chosen is not specified, it would be difficult to repeat the experiments and assess
the error (Gibert and Palmqvist, 1995). Our solution for solving this problem and standardizing criteria for
choosing the suture points is to use a thinning algorithm provided by the Visilog 4.1.3 Rev.6 software. This
algorithm was used to obtain a curve with a single pixel for each line (see figure 1.B). The resultant image is
saved in BMP format for further operations.
In arder to extract the cartesian coordinates from the points of the image resulting from the thinning
process, a curve plotting program was developed using the ViLi Language. This is a language specially adapted
to image processing which was designed by Xavier Sánchez, professor of the Computer Science Department at
the Universitat Autónoma of Barcelona. The input parameters provided to the ViLi program are: a file containing
the suture image in BMP format, the coordinates of the initial reference point used to find the curve and the
size (in millimetres) of the box containing the curve.
To calculate the fractal dimension the program FD3 was used, which is based on the number of occupied
cells in the entire box containing the curve and therefore, it is sensitive to a translation of the image along the
x-axil. For this reason, each image was placed inside the smallest box that would hold it (the upper left vertex
of the box represents the origin (0,0)). The authors of the program indicate that the error of the estimated value
is less than 5% of the real value of the dimension, if a uniform sample of 200 or more different 2-D points is
used (as is the case with our experiments).

RESULTS AND DISCUSSION

As we have shown in the previous section, our method allows to obtain an objective plot of a cranial suture.
We do not use a digital tablet for taking cartesian coordinates, so the whole process is absolutely automatic and
very quick.
Also, the box counting algorithm allows to calculate the fractal dimension of any curve, including those
which have bifurcations and loops, although we have not considered it in this paper in order to keep the same
assumptions used in the method proposed by Gibert and Palmqvist (1995).

CONCLUSIONS

We conclude that it is not possible to establish any reliable discri-mination with this sample of values. Our
working team propases an Mercase in the number of specimens studied as well as a complete automation of the
process used to obtain the fractal dimension from a photograph or X-ray, stressing the importante of obtaining
reliable bounds for the error. It can be observed that the values calculated by Gibert and Palmqvist (1995) do
not include any error bound, when this is essential, given the small variability between hominids (1.047-1.173)
and cquids (1.123-1.374) (Table 1).

ACKNOWLEDGMENTS

Finally, we would like to express our appreciation for the help and support given by Dr. Julia Minguillón
and Toni Codina (CCD, Computer Science Department at the UAB) and Joan Masoliver (Image Processing
Services at the UAB). This work has been partially funded by the Spanish Government Commission DGICYT
through its project No. TEL 97-0663.

84
A • • • A :72:,.!ENS:•2N CT- SL:TURES

REFERENCES

AGUSTÍ J. and MOYA-SOLA S. Sobre la :de=dad del &len-len:o craneal atribuido a Horno sp. en Venta
Micena (Orce. Granada. 535-33.
GIBERT T. and PALMQVIST P. :1 .992,. Dimensión de las suturas del cráneo de Orce. Revista Española
de Paleontología, 2: 154-167; .
GIBERT J. and PALMQVIST P. :'1995. Fractal analysis of the Orce skull sutures. Journal of Human Evolution,
28: 561-575.
GIBERT J., CAMPILLO D., ARQUÉS GARCIA ouvAREs E.. BORJA C. and LOWESTEIN J.
(1998). Hominid status of the Orce cranial fragment reasserted. riurnal or Human Evolution 34: 203-217.
SÁNCHEZ X. and CVG (1994). Image Processing in TiLi. Technicá Report 1.4, Universitat Autónoma de
Barcelona.
LIEBOVITCH L.S. and TOTH T. (1989). A fast algorithm to determine fractal dimensions by box count ng.
Physícs Letters A, 141: 386-390.
MANDELBROT B. (1987). Los objetos fractales. Metatemas 13, Tusquets Editores.
MANDELBROT B. (1997). La geometría fractal de la naturaleza Metatemas 49, Tusquets Editores.
MOYA-SOLA S. and KÓHHLER M. (1997). The Orce skull: Anatomy of a mistake. Journal of Human Evolution
33: 91-97.
HARTWIG W.C. (1991). Fractal analysis of saggital suture morphology. j Morphol. 210: 289-298.
LONG C.A. (1985). Intricate sutures as fractal curves. J. Morphol. 185: 285-295.

85
ARQUÉS, J. M. & GIBERT, J.

TABLE 1:
Fractal dimension of sagittal sutures of young equids (by external face), infant current humans
(internal and external ) and the values of the different traces drawn for VM-0 (internal face:
M-96 (Moyá & Knhler 1997), A-87 (Agustí & Moyá 1987) and external face: G-95
(Gibert & Palmqvist 1995) and G-96 (Gibert et al. 1998).

Equids (ext. face) Infant Humans (EX. faceY Infant Humans (llsaf.) VM-0 (int.f.)

N° D P N° D P N° D P N° D P
EQAC-1 1.281 2952 IPC-6 1.138 1488 IPSHI 1.097 1D26 M-96 1.162 1169

LZH7476 L269 2300 C-47 1.196 1499 1 PSH2 1.115 1084 -A-87 1.128 1175

12E5444 1.319 2114 C-6 1.201 1271 ELS1006 1.127 1229 VM-0 (ext.f.)

6 1.966 1219 ELS 1005 1.167 950 N° D P

EL-81007 1.098 1267 0-95 1.110 1127


IPS (4) 1.051 1044 0-96 1.106 1270

IPC 64 1.124 1059

IPC 32 1.092 1065

M26 1.078 1008

INIZRZMIN'"'5111M•111111=
MEILESEffle'ZI"EMIZZIMIZE
•IIIIIIMMIEWHIERIME
II••II■IERECECEM•••
MIII•11•11~1111
MMUIZEIZINII, -mai=
MEIIIIMIBIZEUMMEMIZZ
■■■■■■ftweima■■ ME
•BIIIMINEEEZERZZIMIZEZ■■
• ■■•EEK2ZE■■■I•
• ■llIUMNRE■■I•I■■E
• I■■IREEEE■■■■■■■■■■

A B

86

También podría gustarte