Pulmonary lesion subtypes recognition of COVID-19 from radiomics data with three-dimensional texture characterization in computed tomography images

Table 6 Performance of HAFS achieved by SVM, KNN, GaussianNB and QDA by using and not using HAFS

Method	\(\alpha \)	Label	Precision (%)	Recall (%)	Accuracy (%)	F-measure (%)
SVM		1	76.34	99.42	82.3	86.37
		2	99.48	62.07	92.31	76.45
		3	100.0	57.75	98.04	73.21
		4	96.84	58.2	91.75	72.71
HAFS-SVM	0.1	1	90.08	95.03	91.3	92.49
		2	91.64	87.03	95.8	89.28
		3	99.1	77.46	98.92	86.96
		4	84.98	80.14	93.58	82.49
KNN		1	88.04	86.32	85.66	87.17
		2	83.09	83.23	93.19	83.16
		3	78.05	86.49	98.17	82.05
		4	65.98	67.96	87.58	66.96
HAFS-KNN	0.5	1	89.01	87.01	86.6	88.0
		2	83.17	84.52	93.42	83.84
		3	80.77	85.14	98.31	82.89
		4	66.89	69.37	87.97	68.11
GaussianNB		1	88.57	59.6	72.88	71.25
		2	42.62	62.72	75.52	50.75
		3	14.64	86.62	76.01	25.05
		4	37.82	10.19	79.89	16.05
HAFS-GaussianNB	0.1	1	81.77	77.8	77.71	79.74
		2	46.21	51.38	78.19	48.66
		3	34.96	55.63	93.16	42.93
		4	46.67	41.11	80.02	43.71
QDA		1	95.14	36.67	63.72	52.94
		2	39.1	97.27	66.82	55.78
		3	100.0	99.3	99.97	99.65
		4	43.38	48.66	79.07	45.87
HAFS-QDA	0.3	1	86.07	82.19	82.69	84.09
		2	60.85	82.88	84.84	70.17
		3	58.93	92.96	96.68	72.13
		4	56.51	31.84	83.12	40.73

ISSN: 1475-925X