Morphometric and radiomics analysis toward the prediction of epilepsy associated with supratentorial low-grade glioma in children

Table 4 Performance of seizure prediction models using radiomics features plus tumor location features for pediatric patients with supratentorial LGG

Classifier	Precision	Recall	Specificity	Accuracy	F1-Score	TP	FN	FP	TN	AUC
Linear SVM	0.955	0.913	0.960	0.938	0.933	21	2	1	24	0.950
Linear Discriminant	0.955	0.913	0.960	0.938	0.933	21	2	1	24	0.929
Medium Gaussian SVM	0.909	0.870	0.920	0.896	0.889	20	3	2	23	0.908
Fine Gaussian SVM	0.870	0.870	0.880	0.875	0.870	20	3	3	22	0.920
Coarse Gaussian SVM	0.840	0.913	0.840	0.875	0.875	21	2	4	21	0.920
Weighted KNN	0.870	0.870	0.880	0.875	0.870	20	3	3	22	0.903
Medium KNN	0.895	0.739	0.920	0.833	0.810	17	6	2	23	0.906
Cosine KNN	0.895	0.739	0.920	0.833	0.810	17	6	2	23	0.910
Fine Tree	0.792	0.826	0.800	0.813	0.809	19	4	5	20	0.813
Medium Tree	0.792	0.826	0.800	0.813	0.809	19	4	5	20	0.813
Coarse Tree	0.792	0.826	0.800	0.813	0.809	19	4	5	20	0.765
Binary GLM Logistic Regression	0.792	0.826	0.800	0.813	0.809	19	4	5	20	0.847
Fine KNN	0.818	0.783	0.840	0.813	0.800	18	5	4	21	0.811
Cubic KNN	0.889	0.696	0.920	0.813	0.781	16	7	2	23	0.902
Kernel Naive Bayes	0.833	0.652	0.880	0.771	0.732	15	8	3	22	0.889
Quadratic Discriminant	0.824	0.609	0.880	0.750	0.700	14	9	3	22	0.819
Gaussian Naive Bayes	0.867	0.565	0.920	0.750	0.684	13	10	2	23	0.810
Cubic SVM	1.000	0.087	1.000	0.563	0.160	2	21	0	25	0.087
Quadratic SVM	1.000	0.044	1.000	0.542	0.083	1	22	0	25	0.043
Coarse KNN	NaN	0.000	1.000	0.521	NaN	0	23	0	25	0.000

AUC, area under receiver operating characteristic curve; FN: False Negative; FP: False Positive; GLM: Generalized linear models; KNN: k-nearest neighbors; SVM: Support Vector Machines; TN: True Negative TP: True Positive

ISSN: 1470-7330