ScienceDirect

ScienceDirect
Procedia Computer Science 00 (2020) 000–000
Procediaonline
Available Computer
at Science 00 (2020) 000–000
www.sciencedirect.com www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia

ScienceDirect
Procedia Computer Science 170 (2020) 905–910

The 5th International Workshop on Big Data and Networks Technologies (BDNT)
The 5th International Workshop
April 6-9,on2020,
Big Data
Warsaw,
and Networks
Poland Technologies (BDNT)
April 6-9, 2020, Warsaw, Poland
Predicting the Anthropometric Properties of Cranial Structures
Predicting the Anthropometric Properties
Using Big Data of Cranial Structures
Using Big Data
Amelec Viloriaa*, Yisel pinillos-Patiñob, Omar Bonerge Pinedac, Ligia Romerod
Amelec Viloriaa*, Yisel pinillos-Patiñob, Omar Bonerge Pinedac, Ligia Romerod
a, d
Universidad de la Costa, Barranquilla, Colombia
ba, d
Universidad Simon
Universidad de laBolivar, Barranquilla, Colombia
Costa, Barranquilla, Colombia
c
Universidad Tecnológica
b Centroamericana
Universidad (UNITEC), San
Simon Bolivar, Barranquilla, Pedro Sula, Honduras
Colombia
c
Universidad Tecnológica Centroamericana (UNITEC), San Pedro Sula, Honduras

Abstract
Abstract
The objective of this study was to generate predictive statistical models of the anthropometric dimensions of craniofacial
structures,
The fromofmedical
objective images
this study obtained
was by Computed
to generate Tomography
predictive statistical (CT).
modelsThe ofstudy consisted of two-dimensional
the anthropometric dimensions ofmeasurement
craniofacial
of the distances
structures, between
from medical the anthropometric
images points Glabella,
obtained by Computed Tomography Vertex,
(CT).Eurion,
The study Nasion and ofOpisthocranium
consisted two-dimensionaltomeasurement
achieve the
dimensions:
of skull between
the distances length (G-Op), head width (Eu-Eu)
the anthropometric and head Vertex,
points Glabella, height (V-N).
Eurion,The iQ-VIEW/
Nasion iQ-Lite softwaretowas
and Opisthocranium used the
achieve for
measurement.skull
dimensions: A total of 30
length adult skulls
(G-Op), between
head width the ages
(Eu-Eu) andofhead
50 and 70 (V-N).
height were measured, all inhabitants
The iQ-VIEW/ iQ-Lite of the cityof
software wasMedellin,
used for
Colombia. TheAmean
measurement. total and
of 30standard deviation
adult skulls values
between thewere
ages calculated.
of 50 and 70 A predictive modelall
were measured, was developedofusing
inhabitants multiple
the cityof linear
Medellin,
regression, The
Colombia. which predicts
mean the distance
and standard corresponding
deviation to head
values were height (V-N)
calculated. relative model
A predictive to G-Opwasand Eu-Eu regressors,
developed obtaining
using multiple lineara
square R value
regression, of 0.375.
which predictsPositive correlations
the distance were observed
corresponding between
to head height the threerelative
(V-N) craniofacial dimensions.
to G-Op and Eu-Eu regressors, obtaining a
square R value of 0.375. Positive correlations were observed between the three craniofacial dimensions.
© 2020 The Authors. Published by Elsevier B.V.
© 2020
This The
is an Authors.
open accessPublished by Elsevier
article under the CCB.V.
BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
© 2020 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the Conference Program Chairs.
Peer-review under responsibility of the Conference Programlicense
This is an open access article under the CC BY-NC-ND Chairs.(http://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the Conference Program Chairs.
Keywords: ANOVA, Anthropometry, Medical Imaging, Computed Tomography.
Keywords: ANOVA, Anthropometry, Medical Imaging, Computed Tomography.

* Corresponding author. Tel.: +57-3046238313.

* E-mail aviloria7@cuc.edu.co
address:author.
Corresponding Tel.: +57-3046238313.
E-mail address: aviloria7@cuc.edu.co
1877-0509 © 2020 The Authors. Published by Elsevier B.V.
This is an open
1877-0509 access
© 2020 Thearticle under
Authors. the CC BY-NC-ND
Published license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
by Elsevier B.V.
Peer-review under
This is an open responsibility
access of the
article under the Conference
CC BY-NC-NDProgram Chairs.
license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the Conference Program Chairs.

1877-0509 © 2020 The Authors. Published by Elsevier B.V.

This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the Conference Program Chairs.
10.1016/j.procs.2020.03.112
906 Amelec Viloria et al. / Procedia Computer Science 170 (2020) 905–910
2 Author name / Procedia Computer Science 00 (2018) 000–000

1. Introduction

Injuries to the face, head and neck are relatively common. The incidence and severity of craniofacial lesions vary
greatly depending on the type of injury and the severity of the injury, for example, in some countries, road traffic
accidents are the most common cause of head fractures [1]. On the other hand, craniofacial defects can occur from
bone resorption, infection, tumors or traumatic loss associated with assaults, traffic accidents and sports.
Cranioplasty, or cranial reconstruction after these defects, uses bone implants to achieve repair. However, among the
main drawbacks are the complicated geometry that the defect may present [2].
Recent developments in craniofacial reconstructions have shown significant advances in both materials and
methods used, however, conventional craniofacial reconstruction methods still fail to mimic the complex 3D
anatomy and the biology of native tissues [3]. On the other hand, the average implant is not designed to
accommodate specific custom variations represents a major reconstructive challenge due to the difficulty of
predicting missing geometry. Biometric models can potentially solve this problem by statistically transforming the
average shape of the skull and applying the model to predict missing anatomy to generate custom implants [4].
There is clear evidence of variation between the cranial dimensions of different populations such as Koreans,
Caucasians, Indians, among others [5], [6]. Racial characteristics are claimed to be better defined in the skull [7]. As
a result, cranial dimensions are one of the most important properties for determining anthropometric parameters [8],
[9].
The purpose of this study is to generate predictive statistical models of the anthropometric dimensions of three-
dimensional craniofacial structures for bone implant design, based on a set of statistical data of adult skulls three-
dimensional of a sample of healthy adults in Colombia.

2. Methodology

2.1 Acquisition of TC images

Computerized Tomography (CT) images were provided by a public hospital in Colombia (for reasons of
confidentiality his name will not be displayed). The images must have met the following inclusion criteria: coming
from skulls of healthy adults without facial fractures or deformities. Exclusion criteria considered having facial
fractures, post-traumatic defects or deformities, congenital or developmental malformations or asymmetries, or the
presence of implants [10], [11].
The quality of the CT scan was at least HQ helical, 2.5 x 1.25 mm collimation, 7.5 mm / rotation, 120 kVp, 200
mA, 1 second and the use of an STD/bone algorithm. Out of a total of 100 skulls, 30 aged between 50 and 70 years
for the population of the state of Medellin in Colombia, met the inclusion criteria and were used for subsequent
anthropometric analysis.

2.2 Anthropometric analysis

The measurement technique used in this study was based on the two-dimensional measurement of the
anthropometric points located in the images of the three-dimensional computed tomography of the patient's face (see
Figure 1). The iQ-VIEW/ iQ-Lite software was used, which is software designed by radiologists for medical
imaging processing. Three craniofacial dimensions were measured corresponding to the distances between five
reference anthropometric anatomical points [12], [13]: Glabella, Vertex, Eurion, Nasion and Opisthocranium (see
Table 1). Table 2 shows the names of craniofacial anthropometric dimensions [8-9].

2.3 Statistical Analysis

The data was captured in Excel software® and then migrated for statistical analysis in The SPSSv17 software.
The search and processing of outliers was performed taking into account the acceptable intervals for each
dimension. The arithmetic mean, standard deviation, minimum, and maximum were calculated. For the development
of predictive models of anthropometric dimensions, multiple linear regression analysis was applied. A significance
 Amelec Viloria et al. / Procedia Computer Science 170 (2020) 905–910 907
Author name / Procedia Computer Science 00 (2018) 000–000 3

value of 0.05 was used for all analyses performed [14].

a) b) c)

Fig. 1. Measuring the distance between the anthropometric points (a) G-Op of an axial cut, (b) Eu-Eu of a coronal cut, (c) V-N of a sagittal cut of
a two-dimensional CT scan of the skull.

Tabla 1. Anthroropometric points

number Abbreviation Description

Eurion Eu The most lateral point of the parietal bone

Vertex V Higher area of the skull, is located near the center of the
sagittal suture.
Opisthocranium Op The back of the skull in the mid-sagittal plane

Glabella G The most prominent midline between the eyebrows

Nasion N Junction between the nasal and frontal bones, in the midline
of the frontonasal suture

Tabla 2. Antropometric measurements facial craneo.

No Anthropometric dimension Anthropometric Reference Points

1 Head width I-I

2 Skull length G-Op
3 Head height V-N

3. Results

3.1 Descriptive statistics

According to a sample of 27 healthy adult craniofacial structures of the Medellin city, an average skull length (G-
Op) value of 175.40 mm-5.25479 was obtained, an average head width value (Eu-Eu) of 145.41 mm x 3.29852 and
an average of 145.41 mm x 3.29852 and an average of 145.41 mm x 3.29852 and an average of 80.94 mm x 2.37458
corresponding to the head height (V-N). Table 3 shows the values for the arithmetic mean, standard deviation,
minimum and maximum [15].
908 Amelec Viloria et al. / Procedia Computer Science 170 (2020) 905–910
4 Author name / Procedia Computer Science 00 (2018) 000–000

3.2 Predictive models

Data analysis for predictive model generation was performed using multiple linear regression analysis in
SPSSv17 software by obtaining the prediction equation (1). This equation predicts the result of head height (V-N)
based on the variables wide head (Eu-Eu) and skull length (G-Op). The model was obtained with a squared R value
of 0.3789 and a positive correlation of 0.677. Table 4 shows a summary of the properties of the generated predictive
model.

-17.476 + 0.476*(G-Op) + 0.401*(Eu-Eu) = V-N (1)

The result of the variance analysis study (ANOVA) is shown in Table 5, which shows that factors compared
against an alpha value of 0.05 are significant [16].

Tabla 3. Descriptive statistics of structures facial craneo

Dimensions N Minimum Maximum Media Typic Deviation

G-Op 28 175.40 172.69 180.8433 5.25479
Eu-Eu 2 145.41 153.58 157.4733 3.29852
V-N 28 80.94 92.47 87.4122 2.37458

Tabla 4. Summary of the predictive model

Model R R2 R2 Fixed Typical estimate error

1 0.752 0.452 0.451 2.854215

Table 5. Anova corresponding to the predictive model

Model Sum of squares Gl Media F Sig.

Quadratic
Regression 111.254 4 55.78 7.789 .005
Residual 200.152 24 6.3125

The following is the residual histogram (see Figure 2), which shows that the residuals are normally distributed
with a zero mean, indicating that the data do not include outliers.

Fig. 2. Residuals histogram.

 Author name / Procedia Computer Science 00 (2018) 000–000 5
Amelec Viloria et al. / Procedia Computer Science 170 (2020) 905–910 909

Figure 3 shows a graph of the residuals against the predicted values of the V-N head height, which shows that the
residues are randomly distributed and have a constant variance.

Fig. 3. Gráfico de residuos contra los valores predichos V-N

4. Discussion

This study provides a series of statistical data related to the main craniofacial dimensions for a sample of healthy
adults in Colombia and the correlation between them (75.2%).
The predictive model can be used to estimate head height (V-N) from the wide head craniofacial dimensions (Eu-
Eu) and skull length (G-Op), which can serve as a reference to reconstruct the area of damage in this area by using
the statistical approximation of the actual dimensions of the skull. However, multiple linear regression analysis
showed that there is a difference between the actual values versus the predicted values; so it is important to note that
the reliability of the model only accounts for a variance of 45.1%, which is low, but significant [17], [18], [19].
The residual histogram (Figure 2) verifies that the residuals are normally distributed indicating that outliers are
not included in the data. The residual plot against predicted values (Figure 3) indicates that there is no pattern of
residual behavior relative to the V-N response variable. Therefore, in the proposed multiple linear regression model,
the G-Op and Eu-Eu regressors are linearly related to the V-N response variable, however, it is possible to improve
the model by adding a larger number of significant variables.
In terms of anthropometry it is not reliable to estimate a "longitudinal" dimension based on another "transverse"
dimension, so it is necessary to improve the model by adding new anthropometric points, as well as obtaining a
larger sample of individuals for the estimation of craniofacial parameters and by incorporating manual
anthropometric measurements.

5. Conclusions

A positive correlation was observed between craniofacial dimensions. A multiple linear regression model was
generated for the prediction of head height (V-N) from skull length (G-Op) and head width (Eu-Eu) dimensions.
Cranial dimensions provide a significant means of estimating the dimensions of an individual's missing geometry at
trauma, which will make it easier for the surgeon to design custom implants by using data from the present research
as a statistical reference.
The predictive model derived from this study can be used for the design of cranial bone implants, however,
within the limitations of the model results in the fact that it was developed using a small sample of individuals. As
an improvement and continuation of this research, it is recommended to add a greater number of measurements,
which could be performed manually by adding at least 10 craniofacial anthropometric points.
910 Amelec Viloria et al. / Procedia Computer Science 170 (2020) 905–910
6 Author name / Procedia Computer Science 00 (2018) 000–000

References

[1] P. Ward Booth, B. Eppley y R. Schmelzeisen, Maxillofacial Trauma and Esthetic Facial Reconstruction, 2nd Edition, St. Louis, Missouri:
Saunders, 2016.
[2] INEGI, “Estadistica a Proporsito del Día Mundial de la Diabetes,” Día Mund. la Diabetes., p. 18, 2013.
[3] T. Santhanam and M. S. Padmavathi, “Application of K-Means and genetic algorithms for dimension reduction by integrating SVM for
diabetes diagnosis,” Procedia Comput. Sci., vol. 47, no. C, pp. 76–83, 2014.
[4] Viloria A., Lis-Gutiérrez JP., Gaitán-Angulo M., Godoy A.R.M., Moreno G.C., Kamatkar S.J. (2018) Methodology for the Design of a
Student Pattern Recognition Tool to Facilitate the Teaching - Learning Process Through Knowledge Data Discovery (Big Data). In: Tan Y.,
Shi Y., Tang Q. (eds) Data Mining and Big Data. DMBD 2018. Lecture Notes in Computer Science, vol 10943. Springer, Cham
[5] C. Díaz; R. García; G. Santacruz; G. Aguilar, “Craneoplastía con implante de titanio individualizado mediante tecnología CAD/CAM”,
Implantol Act, vol 12, no. 24, pp. 4-7, 2016.
[6] D. O. Visscher, E. Farré-Guasch, M. N. Helder, S. Gibbs, T. Forouzanfar, P. P. van Zuijlen y J. Wolff, “Advances in Bioprinting
Technologies for Craniofacial Reconstruction,” Trends Biotechnol., vol. 34, no. 9, pp. 700–710, 2016. DOI:
https://doi.org/10.1016/j.tibtech.2016.04.001
[7] T. Teshima, V. Patel, J. Mainprize, G. Edwards y O. Antonyshyn, “Three-Dimensional Statistical Average Skull: Application of Biometric
Morphing in Generating Missing Anatomy,” J. Craniofac. Surg., vol. 26, no. 5, pp. 1634– 1638, 2015. DOI: 10.1097 /
SCS.0000000000001869.
[8] K. V. S. R. P. Varma, A. A. Rao, T. Sita Maha Lakshmi, and P. V. Nageswara Rao, “A computational intelligence approach for a better
diagnosis of diabetic patients,” Comput. Electr. Eng., vol. 40, no. 5, pp. 1758–1765, 2014.
[9] K. Krishan, “Anthropometry in Forensic Medicine and Forensic Science-Forensic Anthropometry”. Int. J. Foren. Sci., vol. 2, no. 1, pp. 1-8,
2007.
[10] R. Ward, P. Jamison, “Measurement precision and reliability in craniofacial anthropometry: implications and suggestions for clinical
applications”, J Craniofac Genet Dev Biol., vol. 11, no. 3, pp. 156-64, 1991.
[11] Mellado A., Suárez, N., Altimir, C., Martínez, C., Pérez J. C., Krause, M., & Horvath, A. (2017) Disentangling the change-alliance
relationship: Observational assessment of the therapeutic alliance during change and stuck episodes. Psychotherapy Research, 27(5), 595-607.
doi: 10.1080/10503307.2016.1147657
[12] Ogles, B. M. (2013). Measuring change in psychotherapy research. En M. J. Lambert (Ed.), Bergin and Garfields’s Handbook of
Psychotherapy and Behavior Change (pp.134– 166). New Jersey: Wiley.
[13] El Pasante, «Ventajas y desventajas de las bases de datos,» 17 Junio 2015. [En línea]. Available: https://educacion.elpensante.com/ventajas-
y-desventajas-de-las-bases-de- datos/. [Último acceso: 12 Noviembre 2018].
[14] Probability Formula, «Hypergeometric Distribution,» [En línea]. Available: http://www.probabilityformula.org/hypergeometric-
distribution.html. [Último acceso: 16 Noviembre 2018].
[15] Skrondal, A., & Rabe-Hesketh, S. (2004). Generalized latent variable modeling. Boca Raton: Chapman & Hall/CRC
[16] Y. Hwang, K. H. Lee, B. Choi, K. S. Lee, H. Y. Lee, W. S. Sir, “Study on the Korean adult cranical capacity”, J. Korean Sci., vol. 10, pp.
239-42, 1995.
[17] S. Arif, J. Holliday y P. Willett, «Comparison of chemical similarity measures using different numbers of query structures,» Journal of
Information Science, vol. 39, nº 1, pp. 1-8, 2013.
[18] Bucci, N., Luna, M., Viloria, A., García, J. H., Parody, A., Varela, N., & López, L. A. B. [2018, June). Factor analysis of the psychosocial
risk assessment instrument. In International Conference on Data Mining and Big Data [pp. 149-158). Springer, Cham
[19] Viloria, A., Bucci, N., Luna, M., Lis-Gutiérrez, J. P., Parody, A., Bent, D. E. S., & López, L. A. B. (2018, June). Determination of
dimensionality of the psychosocial risk assessment of internal, individual, double presence and external factors in work environments. In
International Conference on Data Mining and Big Data (pp. 304-313). Springer, Cham..