Ribeiro F. A. L., Monteiro V. F., Longo E., Ferreira M. M. C., "Quantitative Discrimination of AFM Images of Human Hair from Different Ethnic Groups: African, Caucasian and Oriental". Águas de Lindóia, SP, Brazil, 10-15/09/2006: 10th International Conference on Chemometrics in Analytical Chemistry (CAC-2006, CAC-X), Book of Abstracts (2006) P085. Poster 085.

10th International Conference on Chemometrics in Analytical Chemistry P085

Quantitative Discrimination of AFM Images of Human Hair from
Different Ethnic Groups: African, Caucasian and Oriental

Fabiana Alves de Lima Ribeiro¹*, Valéria Fernandes Monteiro², Elson Longo², Márcia
M. C. Ferreira¹   fabianaalr@yahoo.com.br

1.   Laboratório de Quimiometria Teórica e Aplicada,   Instituto  de Química,  Universidade Estadual de
Campinas (UNICAMP), Brasil.  2.  Laboratório Interdisciplinar de Eletroquímica e Cerâmica, Instituto de
Química,  Universidade Federal de São Carlos (UFSCAR), Brasil.

Keywords: Image Analysis, Human Hair, Discriminant NPLS, Atomic Force Microscopy
_____________________________________________________________________________________
 

    Purpose.  The human hair is a complex tissue consisting  of  several morphological components. The
physical integrity  of  the fiber and the major physical and chemical characteristics are determined  mainly
by the three most important constituents  acting together:  cuticles,  cortex  and  intercellular components.
The ethnic aspects of human hair  are  resulting from  the  differences  in  the constitution of theses three
components,   and  they  are  reflected  in  the characteristical  aspect of the fiber  surface¹.   Quantitative
methods to  identify and classify such characteriscs could  be  very useful  to  cosmetic science,  forensic
investigations and medical diagnosis.
    Method.  This work presents a quantitative method  to  discriminate images of human hair fibers  from
three ethnic groups:  african,  caucasian  and  oriental, based on the utilization of a multi-way partial least
squares  (NPLS²)  technique.   The samples were colected from the root end area,   where  the fibers are
yonger and most preserved  from effects of combing, weathering and cosmetic action. Samples of human
hair were obtained from De Meo Brothers, New York, USA. Atomic Force Microscopy (AFM) images were
obtained using  a  Digital Instruments NanoScope IIa instrument,   under atmospheric conditions  at  25°C
and a loading force of 3.6 nN.

Results. The data set consists of a 3D array X (38 x
256 x 256)  where  the  38  pixels images  (256 x 256)
are superimposed,  and  a bidimensional array Y_ij (38
x 3),  where y_i1 = caucasian  fiber,  y_i2 = african  fiber,
y_i3 = oriental fiber.  Discreete categorical values of  0
and 1  were  attributed  to each column,  in which 0 =
samples that  don´t  belong to that category  and  1 =
means the opposite. The X array was submitted to  a
logarithmic   transformation³   and   it   was   modeled
using NPLS.  The best results were obtained without
any   further   preprocessing  and  4  latent  variables
were  necessary  to   describe  88.21%  on  the  total
variance  of  the  X block  and  66.29%  on  the  total
variance   of   the  Y  block.   Figure 1   presents  the
distribution of samples in the LV1 x LV2 x LV3 loading plot,  where (circles) african,  (squares)  caucasian
and  (diamonds)  oriental.   The hair fibers surface  are  quite irregular and  present  very  heterogeneous
aspects  as  intrinsic characteristic.   The samples are well clustered  in specific areas although  a  few of
them occupy  the intermediary regions.
    Conclusions.   The Discriminant NPLS model allowed  to  discriminate the 3 classes of  samples  with
success,  and it can be used succesfully for classification and discrimination of greyscale images.
    Acknowledgments.   F.A.L.R.  and  M.M.C.F.  greatfully acknowlegde financial support from FAPESP
and CAPES.   V.F.M.  and  E.L.  greatfully acknowlegde financial support  from  FAPESP  and the CEPID
Program and CNPQ.
__________________________________________________________________________________________________________________________

References

¹ Robbins, C. R.,  Chemical and Physical Behavior of Human Hair  (3rd edn),  Springer: New York,  USA,
1994, 1-92.
² Bro, R. J. Chemometrics 1996, 10, 47-62.
³ Huang, J., Wium, H., Qvist, K. B., Esbensen, K. H. Chemolab 2003, 66, 141-158.