ABSTRACT

 

Parkinson’s disease (PD) is a second most common neurodegenerative disease worldwide that mainly affect motor system. Movement Disorder Society-Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) is a golden standard in clinical assessment of people with PD. There have been many research on elaborating the association of total, or subtotal scores of the MDS-UPDRS and the severity of the disease. However, little attention is put on the inter-relationship amongst the measurements in MDS-UPDRS. Therefore, in this paper, we propose the mapping of network of measurements, particularly in MDS-UPDRS Part II measuring the activities of daily living.

Bayesian network is a graph probabilistic model (GPM) that works based on the Bayes theorem to get a conclusion. In this case, we use Bayesian network to understand the relationship of motor aspects of experience of daily living in Parkinson disease. We identify the relationship between activities, represented by nodes in the network. Bayesian network modelling is based on the probabilities of conditional aspects and the conformity of Directed Acyclic Graph (DAG). DAG consists of nodes, representing the variables and edges, representing the relationship between nodes. Nodes formed in the network is characterized by the conditional probability table (CPT). CPT showing the conditional probability of the events Y when an X events has been occurred and denoted by P(Y|X). In addition to CPT, the Bayesian network also produces a combined probability distribution (JPD), which represents the probability of occurrence of aggregate for all existing values present in any existing variables (on each aspect motor)

 

In this discussion will only show some conclusions from MDS-UPDRS PART II data set with different aspects and levels in each event. Represented in the table below :

 

 

 

Event

 

Aspect observed

 

Probability

 

BL

 

P(NP2TRMR|NP2SPCH,NP2HWRT)

• P(NP2TRMR=1|NP2SPCH=2,NP2HWRT=0) = 0,99
• P(NP2TRMR=1|NP2SPCH=0,NP2HWRT=2) = 0,97
• P(NP2TRMR=1|NP2SPCH=2,NP2HWRT=2) = 0,74

 

V04

 

P(NP2RISE|NP2TURN,NP2WALK)

• P(NP2RISE=2|NP2TURN=2,NP2WALK=0) = 0,95
• P(NP2RISE=2|NP2TURN=0,NP2WALK=2) = 0,33
• P(NP2RISE=2|NP2TURN=2,NP2WALK=2) = 0,33

 

V06

 

P(NP2SALV|NP2SPCH,NP2HOBB)

• P(NP2SALV=4|NP2SPCH=3,NP2HOBB=0) = 0,96
• P(NP2SALV=4|NP2SPCH=0,NP2HOBB=3) = 0,3
• P(NP2SALV=4|NP2SPCH=3,NP2HOBB=3) = 0,3

 

V08

 

P(NP2HWRT|NP2DRESS,NP2HOBB)

• P(NP2HWRT=4|NP2DRESS=0,NP2HOBB=1) = 0,0003
• P(NP2HWRT=4|NP2DRESS=0,NP2HOBB=2) = 0,0095
• P(NP2HWRT=4|NP2DRESS=0,NP2HOBB=3) = 0,0095
• P(NP2HWRT=4|NP2DRESS=0,NP2HOBB=4) = 0,2

 

V10

 

P(NP2HWRT|NP2SPCH,NP2TRMR)

• P(NP2HWRT=4|NP2SPCH=0,NP2TRMR=1) = 0,02
• P(NP2HWRT=4|NP2SPCH=0,NP2TRMR=2) = 0,083
• P(NP2HWRT=4|NP2SPCH=0,NP2TRMR=3) = 0,125
• P(NP2HWRT=4|NP2SPCH=4,NP2TRMR=4) = 0,96

 

V12

 

P(NP2HWRT|NP2SALV,NP2TRMR)

• P(NP2HWRT=4|NP2SALV=0,NP2TRMR=1) = 0,0001
• P(NP2HWRT=4|NP2SALV=0,NP2TRMR=2) = 0,0006
• P(NP2HWRT=4|NP2SALV=0,NP2TRMR=3) = 0,0026
• P(NP2HWRT=4|NP2SALV=0,NP2TRMR=3) = 0,2

 

From the table above, we gained some information. Among others, at baseline, speech with level 2 will most affect the tremor with level 1 where the resulting probability value is the largest. After 1 year of observation, in V04 turn with level 2 most affect the activity up from the chair in level 2 where the patient with Parkinson's need more than one try to get up or need occasional help. In the second year of observation, V06 speech ability with level 3 most affect saliva and drooling in level 4  where the patient with Parkinson's most or all of speech cannot be understood. In the third year of observation, V08 hobbies with level 4 most affect the handwriting in level 4 where the patient with Parkinson's most or all words cannot be read. In the fourth year of observation, V10 tremor with level 4 will affect the handwriting in level 4 where the patient with Parkinson's most or all words cannot be read. In the fifth year of observation, V12 tremor with level 4 most affect the handwriting in level 4 where the patient with Parkinson's most or all words cannot be read.

 

Some of the conclusions in the above table provide an overview for Parkinson's patients as early as possible to deal with if they sense symptoms according to the existing levels. Therefore, it is expected to prevent the worsening of motor problems of people with Parkinson's disease.

 

Keywords: Bayesian network, MDS-UPDRS Part II, Parkinson’s disease.

 

Acknowledgment: This research is funded by Indonesian university via PITTA 2018.