C44-Lancaster Farming, Saturday, July 11 1998
Dr. William B. Roush,
Associate Professor of
Poultry Science
Ascites (also known as Pul
monary Hypertension Syndrome)
is a metabolic disease in broilers
that increases in incidence when
broiler flocks grow rapidly and de
creases in incidence when flock
growth rates are restricted. Ascites
has become a major cause of eco
nomic loss
Joint studies between Dr. Bob
Wideman at the University of Ar
kansas and Dr. Bill Roush at Penn
Slate University are being con
ducted to examine methods of de
tection of birds that have a ten
dency to develop ascites.
The objective of these studies
is to give producers a tool for
identification of birds prone to as
cites so those birds can be re
moved from the genetic popula
tion A previous study using arti
This unit feat
fl Shenandoah P"1 WA I
V + a omsion of *HD ■>• Brooders VHli
A Division of
Feeding Systems
24 Hour
Service
7/11/98
(icial neural networks and several
physiological measurements has
been already reported (Lancaster
Farming, November 23, 1998,
42(3) - D2). The current study in
volves the monitoring of the daily
growth of broilers for evidence of
a tendency of the birds to develop
ascites
Growth is usually described as
a cumulative weight over time re
sulting in an S-shaped growth
curve. Growth can also be de
scribed in physics terminology as
velocity (growth rate) and accelera
tion (the rate of growth rate).
Previous work at Penn State
on nonlinear dynamics of growth
has shown that day-to-day growth
velocity and acceleration can be
divided into three distinct growth
phases The phases can be identi
fied as (1) 0-15 days, (2) 16-35
days and (3) 35 to 50 days
These growth phases exhibit
Thanks to the Richard Rhine Family
for choosing Farmer Boy Ag Systems to construct
Farmer Boy Ag Systems
Inc.
410 East Lincoln Avenue, Myerstown, PA 17067
717-866-7565 » 1-800-845-3374
That is, that an artificial neural
network would be able to differen
tiate between normal birds and
birds with ascites based on indi
vidual daily growth velocities and
accelerations. Artificial neural
networks are computer programs
that have been developed to mimic
the biological network of neurons
present in the biological brain.
Artificial neural networks have
been shown to be very successful
m prediction and classification
problems.
An experiment was conducted
involving 46 male broiler chicks
from a breeder pullet line. Growth
data form each bird was obtained
by manually weighing the birds
for each of 50 days on an electron
balance.
their 44’x500’ Broiler House
Ventilation
12’x12’
Generator
& Electric
Building
increased oscillating behavior.
Growth responses in the last two
phases can be very erratic or cha
otic (Lancaster Farming April 22.
1995).
A hypothesis was formed,
based on previous nonlinear re
search on heart rate, that normal
birds would exhibit oscillating
behavior while ascitic birds would
be more steady m growth behav
ior. Because of success in diagno
sis and prediction of complex data
using artificial neural networks, a
second hypothesis was proposed.
The birds were raised in a pen,
provided water, ad libitum, and
feed as mash for the first three
days and pellets thereafter. Birds
surviving to 50 days and prior
mortality were examined for the
presence or absence of ascites. Of
the 46 birds, 13 were identified as
having ascites and the remaining
33 were considered normal.
Average growth velocity and
acceleration and standard deviation
values were statistically evaluated
as response variables for each
growing phase Average values for
velocity and acceleration during
the third phase were different be
tween normal birds and those with
ascites.
The third phase standard devia
tions of velocity and acceleration
(reflecting oscillation for velocity
and acceleration), were greater for
normal birds as compared to birds
with ascites. The results suggest
that while strains of birds with
high growth rates are more prone
BUILDINGIHE FUTURE
,‘4^/
* u
14’x21’
Composting
Building
Hours:
Mon -Fn. 7 to 5:30
Sat. 7:30 to Noon
WATERING SYSTEMS
Poultry contract
provided by
ie« e
ft
Fredei
Icksburg, PA
to ascites, individual birds not get
ting ascites (within the high
growth strain) have higher growth
rates and more oscillation than
birds within the strain that are
prone to ascites.
An artificial neural network
(General Regression Neural Net
work) was trained to predict as
cites based on the day-to-day
growth velocity and acceleration
Data represented the first, first
two, and all three growth phases
The responses of birds in all three
data sets were successfully classi
fied (100%) as having or not hav
ing ascites.
These results are quite promis
ing. In the future, artificial neural
networks may have the potential
for computerized diagnostic
weighing of birds to determine
which birds have a tendency to de
velop ascites.
tlrovers