Table 5.

Segmentation results for the network fine-tuned on S1. Dice coefficient and percentage of misclassified pixels in terms of False Positives (FPs) for Experiment 1 (CNN trained on synthetic data) and Experiment 2 (CNN trained on phantom data), after fine-tuning on S1.

Dataset	Dice Exp.1	Dice Exp.2	FPs (%) Exp.1	FPs (%) Exp.2
T1	0.58	0.57	0.98	1.18
T2	0.78	0.76	2.35	2.85
T3	0.72	0.77	1.08	1.14
T4	0.71	0.73	1.82	2.55
T5	0.29	0.29	1.48	1.97
T6	0.13	0.18	1.33	2.10