Diagnosis System of Microscopic Hyperspectral Image of Hepatobiliary Tumors Based on Convolutional Neural Network

Table 3

Comparison results of S4CNN with different methods in the Pavia University data set.

Method	Type
	SVM	3DCNN	BASE-3DCNN	SSRN	S4CNN
Asphalt	94.21	92.31	97.78	98.41	98.68
Meadows	98.43	98.01	98.47	99.56	99.94
Gravel	75.31	84.86	94.73	92.48	97.66
Trees	92.11	97.02	92.25	95.34	98.04
Painted metal sheets	99.84	97.93	99.64	99.43	99.92
Bare soil	90.01	98.94	100.00	99.26	99.84
Shadows	93.67	96.44	96.14	97.82	97.88
OA (%)	94.39	95.76	96.97	98.57	99.11
AA (%)	91.64	93.09	95.93	98.03	98.32
Kappa × 100	92.53	94.39	96.21	98.12	98.82