Table 4.

Evaluation results of the segmentation models trained using different numbers of training image patches (the sizes of image patches used to train FCNNs were 33^*33^*3 and 65^*65^*3 respectively, n = 5).

Dataset	No. of patches	Methods	Dice			PPV			Sensitivity
			Complete	Core	Enhancing	Complete	Core	Enhancing	Complete	Core	Enhancing
Challenge	1000*5*20	FCNNs	0.71	0.68	0.62	0.57	0.58	0.54	0.95	0.86	0.76
		FCNN + CRF	0.84	0.78	0.68	0.84	0.78	0.61	0.85	0.81	0.79
		FCNN + CRF + post-process	0.87	0.82	0.76	0.91	0.86	0.77	0.84	0.81	0.77
	3000*5*20	FCNNs	0.73	0.70	0.66	0.60	0.61	0.59	0.94	0.86	0.76
		FCNN + CRF	0.84	0.80	0.69	0.86	0.79	0.62	0.84	0.82	0.80
		FCNN + CRF + post-process	0.86	0.83	0.77	0.92	0.86	0.77	0.83	0.82	0.77
	5000*5*20	FCNNs	0.74	0.72	0.67	0.62	0.63	0.60	0.94	0.86	0.77
		FCNN + CRF	0.85	0.80	0.70	0.87	0.80	0.63	0.84	0.81	0.80
		FCNN + CRF + post-process	0.87	0.83	0.76	0.92	0.87	0.77	0.83	0.81	0.77
Learderboard	1000*5*20	FCNNs	0.66	0.58	0.51	0.53	0.53	0.45	0.96	0.74	0.67
		FCNN + CRF	0.83	0.65	0.56	0.82	0.71	0.51	0.86	0.69	0.70
		FCNN + CRF + post-process	0.86	0.72	0.61	0.88	0.76	0.63	0.85	0.77	0.68
	3000*5*20	FCNNs	0.69	0.60	0.52	0.57	0.56	0.47	0.96	0.74	0.67
		FCNN + CRF	0.83	0.65	0.56	0.84	0.70	0.49	0.85	0.69	0.71
		FCNN + CRF + post-process	0.85	0.72	0.61	0.89	0.75	0.59	0.84	0.77	0.68
	5000*5*20	FCNNs	0.70	0.61	0.54	0.58	0.57	0.49	0.96	0.74	0.67
		FCNN + CRF	0.83	0.66	0.57	0.85	0.71	0.50	0.85	0.69	0.71
		FCNN + CRF + post-process	0.86	0.73	0.62	0.89	0.76	0.64	0.84	0.78	0.68