Aggregation Operators Enhance the Classification of ACL-Ruptured Knees Using Arthrometric Data

Abstract

Many people suffer from the anterior cruciate ligament (ACL) injury, which can lead to knee instability associated with damage to other knee structures
Purpose: In this study we present a classification method based on aggregation operators, using Adaptive Network-based Fuzzy Inference System (ANFIS) and Multilayer Perceptron (MLP) neural network to differentiate between arthrometric data of normal and ACL-ruptured knees.
Methods: The data involves 132 samples consisting of 59 patients with injured knee and73 normal subjects. ANFIS hybrid training algorithm is implemented using Fuzzy C-Means (FCM) and subtractive data clustering. The Levenberg–Marquardt (LM) training algorithm is used for MLP neural network. The results of ANFIS and MLP are then combined using aggregation operators.
Results: The best accuracy (96%) is obtained by applying Choquet integral to the outputs of ANFIS classifier with the antecedent parameters selected using FCM algorithm.
Conclusion: The experimental results show that aggregation operators enhance the outcomes of ANFIS and MLP classifiers in discriminating between ACL raptured knees and normal subjects.