DATOS PRELIMINARES SOBRE UN MODELO DE ANÁLISIS DE IMAGEN PARA PREDECIR EL PORCENTAJE DE GRASA INFILTRADA EN EL LOMO DE BOVINOS DE RAZA WAGYU

Arantxa Echegaray Artajo¹, Nicolás Escartín Casas¹, Marta Guillén Palacín¹, Marta Oliván Nerín¹, Irene Fernández¹

(1)-Humeco

Introduction

The
prevalence of highly marbled cattle breeds, such as Wagyu, has increased around
the world due to consumer demand for higher-quality beef. The ability to grade
highly marbled carcasses in vivo accurately and consistently, particularly with
quantifying intramuscular fat content remains a challenging task for the Wagyu
beef industry, because of the poor quality ultrasound images caused by the high
subcutaneous and intramuscular fat content. We have previously developed software
for the prediction of the IMF content in beef with less than 6% IMF and now we are
trying to understand and develop a new model for animals with an intramuscular fat
content of 20% or more.

Methods

We evaluate
a group of 24 Wagyu animals, 15 raised in Argentina and 9 raised in Thailand.

The
ultrasonographic images were taken in both countries with an ultrasound scanner
(EXAPAD® model, IMV IMAGING, Angoulême, France) with a multifrequency linear
transducer, configured at 2.5 MHz and 160 mm deep. At least 2 raw files were
taken in a longitudinal plane, parallel to the midline of the animal and
between the 12th and 13th rib.

After a
week, the animals were slaughtered and a piece of this part of the loin was
collected per animal and analyzed by the Soxhlet method to investigate the real
percentage of fat in the sample.

Then we
analyze the histogram, texture and microstructure of the samples to obtain a
prediction of the value of fat of the sample by the analyses of the image using
mathematical methods.

Results
and discussion

As a preliminary
work, the most important steps to develop our prediction were:

The standardization
of the images. Before realising a mathematical analysis it was necessary to standardize
the images. For this, we defined a mathematical parameter (DKL from Drak Lines)
to investigate the level of ultrasound exposition of the longissimus dorsis superior
fascia and we adjusted all images of all animals to a determinate value. Then
we define another parameter named Zonification, related to the non-homogeneity
of grey levels in the image. When an image doesn’t meet the criteria of low
zonification it is discarded.

After that,
we studied the size of fat micro and macrostructures in the muscular tissue and
we determined directly ist percentage of occupation into the area of study. Differently
from the leaner cattle in Wagyu, we saw a progressive increase in the size of
these structures.

After detecting
the size and percentage of these structures, we obtained the median values for
all images in an animal and we compared them with the chemical fat values. The
animals have IMF values ranging from 5,3% to 21%. The preliminary regression
line has an r2 of 0.91 which is pretty encouraging.

  Conclusions

This study
is, to our knowledge, the first reported in vivo evaluation of the carcass
quality of these breeds. The preliminary regression line has an r2 of 0.91 which
is pretty encouraging.  We continue to take
samples to improve our model and validate it in a blind assay.