A Review on Automatic Mammographic Density and Parenchymal Segmentation

Table 4

Summary of representative studies using other less popular methods (e.g., collective multiple measurements) for mammographic tissue segmentation. denotes correlation coefficient; AUC denotes area under ROC curve. Note that () largely identical studies are excluded in the list and () in case of multiple results, only the best reported results are listed.

Study	Year	Number of density categories	Modalities	Number of views	Number of images	Segmentation evaluation	Risk/density estimation accuracy
Collective multiple measurements
Kallenberg et al. [67]	2011	Dense and fatty	Digitised SFM	MLO	1300	Visually assessed; Pearson (percent density, dense area) = 0.911, 0.895 (automatic-Cumulus)	N/A
Li et al. [68]	2012	Dense and fatty	Digitised SFM	MLO	765 cases and 747 controls	Visually assessed; = 0.884 (automatic-Cumulus); AUC = 0.589 (four densities: 5%, 25%–50%, 50%–75%, and 75%)	N/A
Other methods
Chen et al. [69]	2012	Dense and fatty	Digitised SFM	MLO	321 (MIAS)	Visually assessed	70% (BI-RADS)
Chen et al. [70]	2013	Dense and fatty	Digitised SFM	MLO and CC	321 (MIAS) and 831 (DDSM)	Visually assessed	76% MIAS, 81% DDSM (BI-RADS)