ESTIMATION OF BALANCE STATUS IN HEMIPARETICS

Completed

• Conditions: Hemiparesis

• Registration Number: NCT04423497
• Lead Sponsor: Pamukkale University

Brief Summary

Although Balance Evaluation Systems Test(BESTest) is an important balance assessment tool to differentiate balance deficits, it is time consuming and tiring for hemiparetic patients. Using artificial neural networks(ANNs) to estimate balance status can be a practical and useful tool for clinicians. The aim of this study was to compare manual BESTest results and ANNs predictive results and to determine the highest contributions of BESTest sections by using ANNs predictive results of BESTest sections. 66 hemiparetic individuals were included in the study. Balance status was evaluated using the BESTest. 70%(n=46), of the dataset was used for learning, 15%(n=10) for evaluation, and 15%(n=10) for testing purposes in order to model ANNs. Multiple linear regression model(MLR) was used to compare with ANNs.

Detailed Description

The demographics and clinical information of the participants' were recorded. Clinical information consists of some basic medical data for the patients. Hodkinson Mental Test was used to assess the cognitive status of the participants if they met inclusion criteria. Balance Evaluation Systems Test was used to assess balance status of the participants.

Feed-forward back-propagation ANNs was used in this study by employing Levenberg-Marquardt training algorithm. Tangent hyperbolic transfer functions were used in the hidden layer. Matlab (Version R2017b, Mathworks Inc, USA) was used in ANNs modeling. 70% (n=46), 15% (n=10) and 15% (n=10) of the data obtained from the participants were used for training, validation and test in the study, respectively. Multiple linear regression (MLR) models also were used to compare with ANNs.

Firstly, the ANNs were modeled for the first aim of the study. We used the data of the five traditional balance tests in the BESTest that did not use the real values (the timing or distance), but just the classified values (0-3 points in the BESTest) to train ANNs. Five balance tests were functional reach test (cm), one leg standing test for right and left side (sec), 6-metre timed walk test (sec) and timed up and go test (sec). Then, we compare the manual total BESTest scores with the predicted scores by the ANNs.

Secondly, we removed 6 sections of the BESTest one by one and modeled with the remaining 5 sections of the test to estimate the total BESTest score. After this modeling, we removed each item one by one in the first section and estimated the first section total score. We repeated the process for all the sections of the BESTest.

Statistical Analysis

Study Timeline

• Study Start: 2016-07-31
• Primary Completion: 2018-05-31
• Study Completion: 2018-05-31
• Status Verified: 2020-06
• Study First Submitted: 2020-06-02
• Study First Submitted QC: 2020-06-05
• Last Update Submitted: 2020-06-05
• Study First Posted: 2020-06-09
• Last Update Posted: 2020-06-09

Recruitment & Eligibility

• Status: COMPLETED
• Sex: All
• Target Recruitment: 66

Inclusion Criteria

• Being aged between 35-65 years,
• Able to walk independently or with a walking aid,
• Able to stand at least 1 minute independently,
• Having single hemiparesis,
• Getting at least 8 points from Hodkinson Mental Test.

Exclusion Criteria

• Having comorbidities affecting their balance,
• Having communication problems.
• Patients who cannot comprehend the directions given to them were excluded from the study.

Study & Design

• Study Type: OBSERVATIONAL
• Study Design: Not specified

Primary Outcome Measures

Name	Time	Method
Artificial Neural Networks Modeling	two years	determining the highest contributions of BESTest subsets in order to find ANNs predictive results of BESTest subsets.
Balance Evaluation Systems Test (BESTest)	two years	Biomechanical constraints, stability limits/verticality, anticipatory postural adjustments, postural responses, sensory orientation and stability in gait