Micro-anatomical quantitative optical imaging: toward automated assessment of breast tissues

Breast Cancer Research

Table 2 Summary of top performing parameters for distinguishing between benign and malignant sites measured using the nuclear and ductal segmentation algorithms

Group	Performance metric	Standard deviation of inter-nuclear distance	Area fraction	Range of inter-nuclear distance
A. Classification by nuclear parameter – all sites	AUC	0.87	0.86	0.87
	Sensitivity	78	76	76
	Specificity	82	79	85
B. Classification by nuclear parameter – non-duct sites	AUC	0.93	0.92	0.91
	Sensitivity	85	80	81
	Specificity	88	87	88
C. Classification by nuclear parameter – duct sites	AUC	0.68	0.72	0.74
	Sensitivity	46	65	62
	Specificity	100	70	96
Group	Performance metric	Number of lumens	Minor dimension of outer ellipse	Area of outer ellipse
D. Classification by duct parameter – duct sites	AUC	0.92	0.83	0.82
	Sensitivity	88	73	81
	Specificity	88	79	75

A: Nuclear parameters measured at all sites using the nuclear segmentation algorithm. B: Nuclear parameters measured at all sites except those with breast ducts (normal ducts, hyperplastic ducts, and ductal carcinoma in situ) using the nuclear segmentation algorithm. C: Nuclear parameters measured at sites with breast ducts using the nuclear segmentation algorithm. D: Ductal parameters measured at sites with breast ducts using the ductal segmentation algorithm. AUC area under the curve

ISSN: 1465-542X