1 Involuntary culling of sows Danish sow herds: ~ 15 % of sows sent - - PDF document

▶
1
SMART_READER_LITE
LIVE PREVIEW

1 Involuntary culling of sows Danish sow herds: ~ 15 % of sows sent - - PDF document

Sep 08, 2023 423 likes •677 views

Bayesian network 3 Case examples Advanced Herd Management October 2nd 2009 Tina Birk Jensen Outline 1. The Weak Sow Index model An example of a Bayesian network developed based on collected data 2. The PigLame model An

slide-1
SLIDE 1

1

Bayesian network 3 Case examples

Advanced Herd Management October 2nd 2009 Tina Birk Jensen

Outline

  • 1. The ”Weak Sow Index” model

An example of a Bayesian network developed based on collected data

  • 2. The ”PigLame” model

An example of an Object+Oriented Bayesian network developed based on expert opinions

The ”Weak Sow Index” model

An example of a Bayesian network developed based on collected data

slide-2
SLIDE 2

2

Involuntary culling of sows

Danish sow herds: ~ 15 % of sows sent to rendering plants A problem for the animal welfare and the economy Involuntary culling: Sows sent to slaughter due to a poor health status Sows being euthanized Sows experiencing sudden death

Involuntary culling of sows

The health status of individual sows is often characterized based on presence or absence of individual clinical signs There is a need to develop a way to combine information about several diseases

The ”Weak Sow Index” model

Purpose: Develop a model that characterizes the risk of involuntary culling of individual pregnant sows Weak Sow Index Weak Sow Index (WSI): Probability of a sow to be involuntarily culled

slide-3
SLIDE 3

3

Data used for the WSI

33 sow herds stratified by the feeding system Each herd visited twice: 49+76 pregnant sows randomly selected each time Clinical examinations of individual sows

The clinical protocol

Clinical signs States

Lameness

  • Pressure mark of knee
  • Pressure mark of digit
  • Pressure mark of hock
  • Claw lesion
  • Claw length

Leg position Reaction Shoulder ulcer ! Wounds at rear " Wounds at head " Wounds at shoulder " Vulva bite

  • Filthiness

#$%&$'() *&$'() Body condition score ++) Willingness to stand

  • Data used for the WSI

Farmers recorded all replacements (e.g. euthanasia and sudden death) and reasons for these actions A total of 2875 sows included in the study During a 3 month period: 119/2875 (4.1 %) sows involuntarily culled

slide-4
SLIDE 4

4

WSI: A Bayesian network model

Two steps in the modeling building process Step 1: Structural dependencies between nodes Step 2: Estimate the probabilities for the model

Leg position Filthiness Vulva bite Reaction Head wound Shoulder wound Rear wound Willing to stand Claw length Claw lesion

Step 1: Structural dependencies

PM of digit PM of knee PM of hock Lameness BSC Shoulder ulcer

Involuntary culling

Step 1: Structural dependencies

Using the 16 clinical variables to characterize the underlying correlation structure

We did that using statistical methods (Factor analysis)

slide-5
SLIDE 5

5

Leg position Filthiness Vulva bite Reaction Head wound Shoulder wound Rear wound Willing to stand Claw length Claw lesion

Step 1: Structural dependencies

PM of digit PM of knee PM of hock Lameness BSC Shoulder ulcer

Involuntary culling

Leg position Filthiness Vulva bite Reaction Head wound Shoulder wound Rear wound Willing to stand Claw length Claw lesion

Step 1: Structural dependencies

Lameness BSC Shoulder ulcer

Wounds Lameness Involuntary culling

Leg position Vulva bite Claw length Claw lesion BSC Shoulder ulcer PM of digit PM of knee PM of hock

Pressure mark

Step 2: Estimate the probabilities for the model

Link between the variables and the factors: Factor loadings from the clinical variables

slide-6
SLIDE 6

6

Step 2: Estimate the probabilities for the model

Link between factors and Involuntary culling: Parameter estimates from logistic regression model

Step 2: Estimate the probabilities for the model

Factor: Lameness (p=0.01)

Willing to stand

Weak Sow Index model: Prototype

Lameness

Lameness Involuntary culling

slide-7
SLIDE 7

7

Vulva bite Willing to stand

Weak Sow Index model: Prototype

Lameness

Lameness Involuntary culling

Vulva bite

Now, lets see how it looks like!

Comments!

Bayesian network allows the WSI to be presented even though some of the clinical variables are missing The WSI model is based solely on collected data Other variables may be important for involuntary culling Possible to include expert information in the WSI model The ”PigLame” model

An example of an Object+Oriented Bayesian network developed based on expert opinions

slide-8
SLIDE 8

8

Outline

  • 1. The case problem
  • 2. Modeling methods in general
  • 3. Qualitative structure of the model
  • 4. Elicitation of probabilities
  • 5. Use of the model (demonstration)

Leg disorders in finishers

Leg disorders: Any lesion or dysfunction of the leg or claw that might give rise to lameness Lameness: Deterioration in the gait and/or posture

The case problem

Leg disorders in finishers

An economical problem for farmers due to:

  • Increased work load
  • Cost of treatments
  • Reduced productivity
  • Risk of condemnations
  • A negative impact on animal welfare

The case problem

slide-9
SLIDE 9

9

Causes of leg disorders

  • 1. Infectious

Mycoplasma hyosynoviae, Erysipelothrix rhusiopathiae, Haemophilus parasuis, Streptococcus suis

  • 2. Physical

Fracture, lesion to the claw wall, lesion to the claw sole

  • 3. Inherited

Osteochondrosis manifesta, osteochondrosis dissecans

The case problem

<2% 5080% 70% <1% 30%

Control strategies

Control strategies against leg disorders will depend on the cause category

  • Infectious leg disorders
  • Physical leg disorders
  • Inherited leg disorders

The case problem

Antibiotics Reconstructing the pen Boar semen Weight gain

Useful information

Herd level

  • Herd size (number of pigs delivered)
  • Stocking density (high/low)
  • Floor type in pens (slatted/concrete)
  • Supply of straw in pens (deep/sparse/no)
  • Purchase policy (own piglets/1/>1)
  • Production type (sectioned/continuous)

The case problem

slide-10
SLIDE 10

10

Useful information

Pig level

  • Observe pigs from outside the pen
  • Clinical investigation
  • Bacteriological investigation
  • Pathological investigation

The case problem

Cheap Cheap Expensive Expensive

To make a herd diagnosis of leg disorders

Challenges

  • What information to use
  • How much information to use
  • How to collect the information

The case problem

The ”PigLame” model

Purpose of the model To estimate probability distributions of different manageable causes of leg disorders in finisher herds

The case problem

slide-11
SLIDE 11

11

Strategy 2 Pig Info Leg disorder Strategy 1 Herd Info Strategy 3

The case problem

Qualitative structure of the model

Characteristics

Based on information from the literature All nodes are discrete Each causecategory defined as a risk index I on an arbitrary scale from 0 to 9

Qualitative structure

Qualitative structure of the model

Characteristics

Object oriented structure Ease the specification of the Bayesian network Hierarchical structure Two classes: Herd class and pig class

Qualitative structure

slide-12
SLIDE 12

12

Objectoriented Bayesian network

The Pig class

Individual pig information

Gender Lean meat percentage Diagnostic test results Leg disorder

Qualitative structure

Objectoriented Bayesian network

The Herd class

Herd size Production Purchase Pen density Floor type Straw Cause category

Pig object Pig object Pig object

Pen Den Floor Straw Physical Infectious Inherited Herd Size Produc+ tion Gain Breed Feed strat Pur+ chase

Herd class

slide-13
SLIDE 13

13

Pen Den Floor Straw Physical Infectious Inherited Herd Size Produc+ tion Gain Breed Feed strat Pur+ chase

Herd class

Obs Lame Pig Lame Frac ture Claw Wall Claw Sole Myco Strep Erysi Hae mo OCM OCD Gen der LMP

Physical Infectious Inherited

C1 P1 C2 P2 C3 P3 C4 P4 B1 C5 P5 B2 C6 P6 B3 C7 P7 B4 C8 P8 C9 P9

Pig class

Physical Infectios Inherited Floor Pen Den Gain

……

Frac ture Claw Wall Claw Sole Myco Strep Erysi Haemo OCM OCD

Herd class Pig class

Qualitative structure

slide-14
SLIDE 14

14

Physical Infectios Inherited Floor Pen Den Gain

……

Frac ture Claw Wall Claw Sole Myco Strep Erysi Haemo OCM OCD

Pig class

Qualitative structure

Herd class

Physical Infectios Inherited Floor Pen Den Gain

……

Frac ture Claw Wall Claw Sole Myco Strep Erysi Haemo OCM OCD

Pig class Herd class

Qualitative structure

Elicitation of probabilities

The probabilities in the model are based on

  • 1. Results from published literature

Conversion of odds ratios to conditional probabilities <40 conditional probabilities

  • 2. Expert opinions (9 experts)

>150 conditional probabilities Not randomly distributed Average of individual elicitations

Elicitation of probabilities

slide-15
SLIDE 15

15

Physical Infectios Inherited Floor Pen Den Gain

……

Frac ture Claw Wall Claw Sole Myco Strep Erysi Haemo OCM OCD

Pig class

Elicitation of probabilities

Herd class

Physical Infectios Inherited Floor Pen Den Gain

……

Frac ture Claw Wall Claw Sole Myco Strep Erysi Haemo OCM OCD

Pig class

Elicitation of probabilities

Herd class Example: P(Fracture|fully slatted floors)

Elicitation of probabilities

slide-16
SLIDE 16

16

Consider 100 pigs examined individually at a herd visit. The herd has fully slatted floors in the pens. How often do you, during the examination expect to find a pig with a fracture?

Example: P(Fracture|fully slatted floors)

Elicitation of probabilities

Always (almost) 15 25 50 75 85 100 Usually Often As often as not Sometimes Once in a while (Almost) never

Consider 100 pigs examined individually at a herd visit. The herd has fully slatted floors in the pens. How often do you, during the examination expect to find a pig with a fracture?

Example: P(Fracture|fully slatted floors)

Elicitation of probabilities

Elicitation of probabilities R2 R1 Rn I D1 D2 Dn

Pig class

slide-17
SLIDE 17

17

Elicitation of probabilities

Causecategories: Defined as Risk Index I on an arbitrary scale 0: Low risk 9: High risk

Elicitation of probabilities R2 R1 Rn I D1 D2 Dn

Pig class

Elicitation of probabilities R2 R1 Rn I D1 D2 Dn

Pig class

slide-18
SLIDE 18

18

Elicitation of probabilities

Risk index based on a linear equation I the resulting risk index I the intercept ρk the systematic effect of risk factor k ε random residuals Assumptions: No interactions between the risk factors The effects are additive

R2 R1 Rn I D1 D2 Dn

Pig class

Elicitation of probabilities Elicitation of probabilities

Leg disorder nodes: Modeled using a logistic regression

Logit(P(Dk)) Logistic transformation of the conditional probability of a pig to have the leg disorder α Intercept indicating the base prevalence

  • f the leg disorder

β Slope indicating the sensitivity to changes in the risk level of the herd I Risk Index

I D P Logit

k k k

β α + = )) ( (

slide-19
SLIDE 19

19

Elicitation of probabilities

Parameter estimates for the Risk Index and leg disorder nodes found by:

Using the probabilities elicited by experts or literature Fitting a logistic linear model Optimizing the fit

Elicitation of probabilities

Obs Lame Pig Lame Frac ture Claw Wall Claw Sole Myco Strep Erysi Hae mo OCM OCD Gen der LMP

Physical Infectious Inherited

C1 P1 C2 P2 C3 P3 C4 P4 B1 C5 P5 B2 C6 P6 B3 C7 P7 B4 C8 P8 C9 P9

Pig class

Obs Lame Pig Lame Frac ture Claw Wall Claw Sole Myco Strep Erysi Hae mo OCM OCD

Each expert answered questions regarding the probability of lameness given each cause

Noisy or

slide-20
SLIDE 20

20

Noisy or

For example: P(Lamenessyes|Fracture, ClawWall, ClawSole, Myco, Strep, Erysi, Haemo,OCM, OCD) = 1 – (qFractureqClawWallqClawSoleqMycoqStrepqErysiqHaemoqOCMqOCD)

Use of the model

Decide on the level of information needed in order to identify the most likely causecategory Is it necessary to investigate individual pigs in a herd? Which diagnostic test(s) should be performed? How should pigs for diagnostic examination be selected? How many pigs should be selected?

Use of the model

Two fictitious herds

Same prevalence of lameness: 20% pigs are lame due to Mycoplasma hyosynoviae

  • Low risk herd:

Deliver 2000 finishers annually Sectioned production Produce own piglets Low pen densities Solid floors No supply of straw

  • High risk herd:

Deliver 6000 finishers annually Continuous production Purchase from several herds High pen densities Partially slatted floors Sparse supply of straw

slide-21
SLIDE 21

21

Use of the model

Different scenarios investigated:

  • 1. Herd evidence
  • 2. Herd evidence and observing 50 randomly selected pigs for

lameness

  • 3. Herd evidence and performing diagnostic ex. of lame pigs
  • 4. Herd evidence and performing diagnostic ex. of all pigs
Risk index Risk index Risk index Risk index
slide-22
SLIDE 22

22

Risk index Risk index Risk index Risk index

Use of the model

Low risk herd Necessary to perform diagnostic examination of pigs High risk herd Information regarding the herd characteristics is sufficient More economic benefit in performing diagnostic examination of individual pigs in the low risk herd

slide-23
SLIDE 23

23

Comments!

”PigLame” model is an OOBN model

Ease the specification of the model Suitable method of combining information from two different levels

Probabilities mainly from experts

Prone to subjectivity

Third mandatory report

Use of a Bayesian network to model litter sizes in sheep

Assume the litter size at weaning ,n of parity is determined by an underlying continuous variable n: We assume that (Y1, Y2,…..,Y10) has a 10 dimensional normal distribution N(E,∑)

Third mandatory report

Use of a Bayesian network to model litter sizes in sheep

Exercise:

Based on knowledge in regard to genetics and body condition score: How can we construct more simple Baysian networks?