The aim of this research study is to develop hybrid Computational Intelligence algorithms that searches for the minimum possible training set that achieves the maximum performance in terms of classification accuracy and rehabilitation prognosis in patients undergoing Reverse Total Shoulder Arthroplasty (RTSA).The aim of this research study is to develop hybrid Computational Intelligence algorithms that searches for the minimum possible training set that achieves the maximum performance in terms of classification accuracy and rehabilitation prognosis in patients undergoing Reverse Total Shoulder Arthroplasty (RTSA).
The study was retrospective and patient consent was not applied for this research. The study was approved by the Research Ethics and Ethics Committee of the Democritus University of Thrace (58134/541/12-7-2023).
The data were collected at the Arthroscopy & Shoulder Surgery Center, which operates within the 3rd Orthopedic Clinic of the Hygeia Group, Athens, Greece. Data consisted of the following features: gender, age, type of surgery, range of all motions, numeric pain rating scale, number of physical therapy sessions and session duration, that were reported in the week before surgery, in the first trimester as well as six months after surgery. A number of 120 patients from 2015 to present who underwent RTSA were included. Computer experiments were conducted in two phases, I and II. In phase I, there were applied two Machine Learning algorithms, namely the K-Nearest-Neighbors classification algorithm (K-NN) and the K-MEANS clustering algorithm, as well as, a Genetic Algorithm Clustering method (GAClust). In phase II, three hybrid Computational Intelligence algorithms were applied, namely GAKNN, GAKMEANS and GA2Clust, where Genetic Algorithm search was utilized in order to optimize the training procedure and maximize classification and prognostic performance of the simple algorithms in Phase I (K-NN, KMEANS, GAClust, respectively).
Simple algorithms in Phase I achieved 100% classification and prognostic performance using 90% of available data for training. The Genetic Algorithm driven, hybrid algorithms in Phase II achieved 100% performance by using only the 35.83% of available data for training.
The proposed hybrid Computational Algorithms outperformed simple Machine Learning Algorithms for RTSA classification and rehabilitation prognosis, in terms of minimization of the size of essential data needed for training, as well as the reduction of the computational cost.
The proposed hybrid algorithms may be used as a Computational Intelligence tool for evaluating patients postoperatively during rehabilitation by completing scores in each direction of motion as well as the measuring time and early detects non-positive rehabilitation progress.
Genetic Algorithms
Ηybrid Machine Learning Algorithms
