Threshold estimation based on local minima for nucleus and cytoplasm segmentation

BMC Medical Imaging

Table 3 The performance of the proposed method for WBC segmentation compared to deep neural network (LeukocyteMask(Aug-ET)), U-Net (implemented in [22] ) and WBC-Net based on the data set two presented using average in each measure of similarity

	JI	DSC	Specificity	Sensitivity	Precision
Proposed method	0.942908	0.970176	0.975164	0.992475	0.950133
U-Net [36]	0.941467	0.969628	0.989037	0.962368	0.977735
Deep Neural Network [22]	–	0.981960	–	–	0.995440
U-Net [22]	–	0.961630	–	–	0.933190
WBC-Net [23]	–	0.989200	–	–	0.989700

ISSN: 1471-2342