Table 5.

Studies with ML-driven post-treatment planning techniques of KOA.

Author	Year	Data	Feature engineering	Learning Algorithm	Validation	Results
Chen, H·P. [123]	2016	Biomechanical data	Tilt angle calculation and initial posture classification algorithm	Multi-layer SVM	10-fold cross validation	90.6% on layer-1 SVM & 92.7% on layer-2 SVM
Huang, P·C. [124]	2017	Biomechanical data	Sequential forward feature selection (SFS)	Multi-class SVM	10-fold cross validation	Accuracy for rehabilitation exercises recognition is 100% and for motion identification is 97.7%.
Levinger, P. [121]	2009	Biomechanical data	SVM	SVM	LOOCV	Accuracy of 100% for the training set and 88.89% for the test set
Wittevrongel, B [122].	2015	Biomechanical data	k-equal frequency binning	Decision tree & Rule sets	LOOCV	Best accuracy 92.9% & 76.5% respectively