Current updates in machine learning in the prediction of therapeutic outcome of hepatocellular carcinoma: what should we know?

Insights into Imaging

Table 4 Characteristics of ML-based predictive model after RFA

Author	Study type	No. of patients	Modality	Model	Outcomes	AUC	Conclusion
Liang [69], 2014	Retrospective Single center	83	US guided	SVM	RR	0.69	The SA + RF SVM method had the best accuracy for predicting high-risk recurrent patients
Wu [70], 2017	Retrospective Single center	431	CT guided	MLP	1,2-year DFS	D: 0.94 (1-year) 0.88 (2-year) V: 0.77 (1-year) 0.72 (2-year)	The MLP-based model with 15 clinical HCC relevant features achieved satisfactory prediction performance for 1-year DFS

ML machine learning, RFA radiofrequency ablation, AUC area under the curve, US ultrasound, SVM support vector machine, RR recurrence rate, SA simulated annealing algorithm, RF random forest, CT computed tomography, MLP multilayer perceptron, DFS disease free survival, D development cohort, V validation cohort