Table 1.

Comparison of intra-cohort performance of CNNs with different baselines in terms of Dice score for detecting active BAT

	Basel	Granada	Zurich	MSKCC
HU thresholding -180, -10	0.427	0.421	0.248†	0.115
CNN	0.745†	0.521†	0.189	0.130

† indicates a statistically significant difference between the results of CNN and the HU thresholding method with p < < 1% and bold indicates the best results.