A Multicriteria Model Applied in the Diagnosis of Alzheimer's Disease

1
A Multicriteria Model Applied in the Diagnosisof
Alzheimer's Disease

• Ana Karoline A. de Castro
• akcastro_at_gmail.com
• Plácido Rogério Pinheiro
• Mirian Caliope Dantas Pinheiro
• placido, caliope_at_unifor.br
• University of Fortaleza Ceara - Brazil

2
Introduction

• Many studies show that there has been a
  progressive increase of the elderly population in
  the last years.
• The main consequence of this fact, is the
  increase in life, expectancy and the increase of
  the prevalence of dementias.
• Alzheimers disease is the most frequent cause of
  dementia.

• Porto, C.S., Fichman, H.C., Caramelli, P.,
  Bahia, V. S., Nitrini, R. Brazilian Version of
  the Mattis Dementia Rating Scale Diagnosis of
  Mild Dementia in Alzheimer.s Disease, Arq
  Neuropsiquiatr 61(2-B) (2003) 339345.

3
Introduction

• Systems of classification with specific inclusion
  and exclusion criteria allow for the clinically
  probable definition and diagnosis of Alzheimers
  disease.
• The main focus of this work is to develop a
  multicriteria model for aiding in decision making
  for the diagnosis of Alzheimer's disease.

4
Diagnosis of Alzheimer's Disease

• Diagnosis of Alzheimer's disease is carried out
  in several steps. Initially, syndromic diagnosis
  is defined, which informs whether the patient
  presents the diagnostic criteria for dementia.
• After Dementia is confirmed, etiological
  diagnosis follows, which informs which disease is
  causing the dementia. In this case, we are
  looking for Alzheimer's disease.

• World Health Organization at http//www.who.int/c
  lassifications/icd/en/
• Hughes, C.P., Berg, L., Danzinger, W.L., Coben,
  L.A., Martin, R.L. A New Clinical Scale for the
  Staging of Dementia, British Journal of
  Psychiatry 140 (1982) 566572.
• Mckhann, G.D., Drachman, D., Folstein, M.,
  Katzman, R., Price, D., Stadlan, E.M. Clinical
  Diagnosis of Alzheimers Disease Report of the
  NINCDS-ADRDA Work Group under the Auspices of the
  Department of Health and Human Services Task
  Force on Alzheimers Disease, Neurology 34 (1984)
  939944.

5
Definition of Problem

• In studies developed by (Castro,2007a) and
  (Castro,2007b) the application of the
  multicriteria model for aiding in diagnosis of
  Alzheimer's disease was presented.
• Two MCDA tools were used to aid in solving the
  problem
• Hiview (www.catalyze.co.uk )
• M-MACBETH (www.m-macbeth.com)

• CASTRO, A. K. A. PINHEIRO, P. R. PINHEIRO, M.
  C. D. Applying a Decision Making Model in the
  Early Diagnosis of Alzheimers Disease. In Joint
  Rough Sets Symposium (JRS07), 2007. Proceedings
  of Joint Rough Sets Symposium. Berlin Springer
  Verlag, 2007a.
• CASTRO, A. K. A., PINHEIRO, P. R., PINHEIRO, M.
  C. D. A. Multicriteria Model Applied in the
  Early Diagnosis of Alzheimer's Disease A
  Bayesian Approach. In International Conference
  on Operational Research for Development
  (ICORDVI), Brazil, 2007b.

6
Definition of Problem

• In the present study, we sought to develop a
  multicriteria model that helps in the decision
  making related to the diagnosis of Alzheimer's
  disease.
• This model consists of several steps that will be
  described below.
• Initially were defined the problem value tree
  corresponding to the Fundamental point of view
  (FPV) that are used in evaluation of the
  diagnosis.

7
Definition of Problem - Describers

• An FPV is operational in the moment that has a
  set of levels of associated impacts (describers).
  These impacts are defined for Nj, that can be
  ordered in decreasing form according to the
  describers (Bana, 2003).
• For the evaluation of each FPV, the possible
  states were defined. Each FPV has a different
  quantity of states. These states were defined
  according to the exams or questions involved for
  each describer.
• Its important remember that the describers has a
  structure of complete pre-order, otherwise, a
  superior level is always preferable a least level.

• Bana Costa, C.A., Corte, J.M.D., Vansnick,
  J.C. Macbeth, LSE-OR Working Paper, (2003).

8
Definition of Problem - Analysis of Impacts

• In this step, the analysis of impacts is carried
  out, according to each FPV (i) the lowest and
  highest values of the impacts and (ii) the
  relevant aspects of the distribution of the
  impacts in each one.
• In this work, for each describer, the same values
  were considered to get the value function for
  each FPV. Therefore, scores higher than 60,
  obtained through the judgments matrixes were
  considered risk describers during the evaluation
  of diagnosis.

9
Definition of Problem - Evaluation

• After the definition of the FPVs family and the
  construction of the describers, the next step is
  the construction of the cardinal value scales for
  each FPV.
• After evaluating the alternatives of all the FPVs
  individually, an evaluation of the FPVs in one
  matrix only was carried out.

10
Definition of Problem - Results

• In this step we show the final result of the
  model, the contribution of the criterion for the
  diagnosis of Alzheimer's disease.
• We can see the describer values for each
  criteria. With these values we can conclude that
  the questions that are part of these describers
  should be preferentially applied during the
  definition of diagnosis of Alzheimer's disease.

11
Definition of Problem - Evaluation Results

• In this step the value analyses of the
  sensitivity and the dominance of potential
  actions in the process of evaluation to discover
  which questions are most relevant for diagnosis
  of Alzheimer's disease by using data sets. HIVIEW
  software (Keysalis, 1995) was used to carry out
  these analyses.

• Keysalis. Hiview for Windows. London Krysalis,
  (1995).

12
Application of the Multicriteria Model

• For the application of the model two scenarios
• 1) The battery of standardized assessments
  developed by The Consortium to Establish a
  Registry for Alzheimer's disease (CERAD)
• 2) The questionnaire with 120 questions that was
  used in a study realized in the city of Sao Jose
  dos Campos, SP, Brazil were used.

13
Scenario 1

• The original mandate of the Consortium to
  Establish a Registry for Alzheimer's Disease
  (CERAD) in 1986 was to develop a battery of
  standardized assessments for the evaluation of
  cases with Alzheimer's disease who were enrolled
  in NIA-sponsored Alzheimer's Disease Centers
  (ADCs) or in other dementia research programs
  (Morris, 1989).
• CERAD was designed to create uniformity in
  enrollment criteria and methods of assessment in
  clinical studies of Alzheimer's Disease and to
  pool information collected from sites joining the
  Consortium.

• Morris JC, Heyman A, Mohs RC, et al. The
  Consortium to Establish a Registry for
  Alzheimer's Disease (CERAD) Part 1. Clinical and
  Neuropsychological Assessment of Alzheimer's
  Disease. Neurology,39(1989) 1159-1165.

14
Scenario 1 - Problem value tree
15
Scenario 1- Describers

• Each FPV has a different quantity of states. The
  FPV1 has 3 states. The FPV2 has 5 states. The
  FPV3 has 3 states. The FPV4 has 3 states. The
  FPV5 has 3 states. The FPV6 has 3 states. The
  FPV7 has 13 states.

16
Scenario 1- Evaluation

• Next, the evaluation of the FPVs was made through
  the construction of the judgment matrixes.

17
Scenario 1- Evaluation

• In the last step of the application of the model
  we made the global evaluation of the actions when
  they are confronted with the seven FPVs.

18
Scenario 1- Evaluation

• It was found that FPV1 to FPV7 have a total
  participation of 19,8, 13,9, 17,8, 10,9,
  1,0, 17,8 and 18,8.
• The action N03 proved to be potentially better
  with 85 points and the action N01 proved to be
  potentially least.

19
Scenario 2

• We used information obtained from the study
  realized in 2005 with 235 elderly people in the
  city of Sao Jose dos Campos, SP, Brazil.
• In this study a questionnaire with 120 questions
  that supplied demographic-social data, analyzed
  the subjective perception of the elderly, their
  mental and physical health (cognitive and
  emotional aspects), day-to-day independence, in
  addition to family and social support and the use
  of services (Belderrain, 2006).

• Belderrain, M. C. N., Silva, A. C. S., Vianna,
  P. V. C. Functional capacity of the elderly in
  city of Sao Jose dos Campos - Sao Paulo - Brazil.
  In 21st European Conference on Operational
  Research - EURO XXI, Iceland. EURO XXI
  Proceedings, (2006).

20
Scenario 2 - Problem value tree
21
Scenario 2- Describers

• The FPV1 has 4 states. The FPV2 has 6 states. The
  FPV3 has 7 states. The FPV4 has 8 states. The
  FPV5 has 6 states. The FPV6 has 4 states.

22
Scenario 2- Evaluation

• Next, the evaluation of the FPVs was made through
  the construction of the judgment matrixes.

23
Scenario 2- Evaluation

• In the last step of the application of the model
  we made the global evaluation of the actions when
  they are confronted with the six FPVs.

24
Scenario 2- Evaluation

• It was found that FPV1 to FPV6 have a total
  participation of 14,3, 22,9, 14,3, 14,3,
  20,0 and 14,3.
• Thus, action N04 proved to be potentially better
  with 99 points and the action N01 proved to be
  potentially least.

25
Comparation of the Results

• In scenario 1 we have the FPV7 (Neuro-pathological
  Diagnoses) as the criterion with the largest
  number of describers which obtained the minimum
  values required by the model. This result shows
  the importance of the definition of the diagnosis
  of Alzheimer's.
• This is merely to underline the importance of
  this criterion for the solution of this problem.
  In scenario 2 we have FPV4 (Community Affairs) as
  the criterion with the largest number of
  describers which obtained the minimum values
  required by the model.

26
Comparation of the Results

• Analyzing the characteristics of these two
  criteria, we see that despite being in different
  scenarios, they have similarities, as both the
  FPV7 in scenario 1 and FPV4 in scenario 2 are
  related primarily, the possible mental health
  problems that may occur with the patient and that
  may cause confusion in the definition of the
  diagnosis of Alzheimer's disease. The first
  result was obtained from the judgment of matrixes
  through the M-MACBETH software.

27
Conclusions

• The MACBETH multicriteria method was used to aid
  in decision making. The criteria were defined
  according to the CERAD areas of assessment. In
  the questionnaire applied in the Brazilian study
  we defined the criteria based on to
  demographic-social data, analyzed the subjective
  perception of the elderly, their mental and
  physical health.
• The questions that make up the battery of
  assessments and in the questionnaire were defined
  as the describers of the problem. With this
  information, the judgment matrixes were
  constructed using M-MACBETH software.
• After evaluating the matrixes, a ranking was
  obtained showing all the questions, from most
  important to least important with respect to the
  diagnosis of Alzheimer's. At the end of this
  study, a comparison was made of the results
  obtained in the application of the model in both
  data sets.

28
Future Work

• An extension of the model is in the submission
  process (Castro, 2008) using Bayesian Networks.
  We defined the mapping of the questions of CERAD
  and in the questionnaire applied in the Brazilian
  study and through these mappings we can observe
  the questions of the biggest decisive impact for
  the diagnosis.
• In addition, with this Bayesian Network, we
  structured a multicriteria model that shows the
  questions of the biggest decisive impact for the
  diagnosis. Were showing the percentage of elderly
  people that presented the symptoms of Alzheimers
  disease.

• Castro, A. K. A., Pinheiro, P. R., Pinheiro, M.
  C. D. Applying Bayesian Networks and a
  Multicriteria Model in the Diagnosis of
  Alzheimer's Disease. BMC Medical Informatics and
  Decision Making, (2008). ISSN 1472-6947. Paper
  has been submitted.

29
Future Work

• As a future project, this model can be extended
  with the inclusion of new criteria or new models
  which can be developed using other data sets
  (Belderrain, 2006).

• Belderrain, M. C. N., Silva, A. C. S., Vianna,
  P. V. C. Functional capacity of the elderly in
  city of Sao Jose dos Campos - Sao Paulo - Brazil.
  In 21st European Conference on Operational
  Research - EURO XXI, Iceland. EURO XXI
  Proceedings, (2006).

30
Acknowledgment

• The authors thank the Consortium to Establish a
  Registry for Alzheimer's Disease (CERAD), Amanda
  Cecilia Simoes da Silva and Mischel Carmem Neyra
  Belderrain (Instituto Tecnologico de Aeronautica)
  for the divulgation of the data utilized in this
  case study.
• We are thankful to FUNCAP for the support we have
  received for this project.

31
Thank you for coming !