Breast Cancer Segmentation Methods: Current Status and Future Potentials

Table 2

Summary of reviewed works related to classical segmentation in mammogram image.

Subcategory	Related works	Year	Technique	Filter	Database	Evaluation metric
RM	[77]	1999	Adaptive and region growing	Gaussian	UMH	98.0% accuracy
RM	[102]	2001	Region growing	Kalman	DDSM	93.0% ROC with adaptive module and 86.0% ROC without the adaptive module
RM	[97]	2001	Partial loss of region	Sober	Japanese	97.0% true positive
RM	[99]	2004	Region growing		MIAS	90.0% TPR, and 1.3 FTR per image
RM	[103]	2004	Contour searching		MAGIC-5	ROC
RM	[87]	2005	Region growing	ANN	MIAS	92.5% accuracy
RM	[90]	2006	Morphological algorithm	Median	MIAS	95.0% detection rate
RM	[75]	2010	Harris corner	Median	MIAS	93.0% segmentation accuracy
RM	[91]	2010	Region growing		DDSM	78.0% sensitivity and 4.0% false positive
RM	[92]	2010	Watershed	Morphological	DDSM	Mean standard
RM	[74]	2011	Thresholding	Median	MIAS	99.0% segmentation accuracy
RM	[82]	2012	Region growing	Contrast	MIAS	94.59% sensitivity and 3.90 false positive
RM	[86]	2012	Morphological	Median	MIAS	95.0% detection rate
RM	[96]	2012	Region growing	Adaptive	DDSM	97.2% sensitivity and 1.83% false positive
RM	[80]	2012	Seed point selection	Mathematical morphology	NCSM	98.0% accuracy
RM	[81]	2013	Morphological gradient watershed	Adaptive median	MIAS and NMR	95.3% positive for MIAS and 94.0% for NMR
RM	[101]	2013	Improved watershed	Median	MIAS	92.0% accuracy
RM	[76]	2013	Otsu	Morphological	DEMS	95.06% accuracy
RM	[95]	2014	Marker-controlled watershed	Sober	MIAS	90.83% detection rate and 91.3% ROC
RM	[84]	2014	Wavelet and genetic algorithm	Wiener	MIAS and DDSM	mean and standard deviation
RM	[98]	2014	Watershed transformation		MSKE	90.47% sensitivity, 75.0% specificity, and 84.848% accuracy
RM	[79]	2015	Morphological operators	Alternating sequential filter	MIAS	99.2% sensitivity and 99.0% accuracy
RM	[83]	2017	Region growing	Sliding window	MIAS	91.3% accuracy
RM	[100]	2017	Region growing	Median	MIAS	94.0% accuracy
RM	[93]	2017	Watershed	Morphological	DDSM	80.5% similarity index, 75.7% overlap value
RM	[94]	2017	Bimodal-level set formulation		MIAS	96.72% precision and 97.22% recall
RM	[88]	2018	Hidden Markov and region growing		MIAS	91.92% accuracy and 8.07% error
RM	[89]	2018	Watershed combined with -NN	Sober	MIAS	83.33% accuracy
RM	[78]	2018	Region growing	Gaussian	DDSM	98.1% sensitivity, 97.8% specificity, and 90.0% accuracy
RM	[85]	2019	Watershed		MIAS	94.0% false detection and 18.0% positive detection
TM	[118]	2001	Otsu thresholding	Morphological	MIAS	1.7188 ME1, 0.0083 ME2, and 0.8702 MHD
TM	[120]	2001	Otsu	Median	MIAS	96.55% accuracy, 96.97% sensitivity, and 96.29% specificity
TM	[113]	2011			MIAS	97.0% accuracy, 97.03% specificity, and 97.0% sensitivity
TM	[117]	2012	Histogram thresholding	Morphological	DDSM	96.0% detection rate and 90.0% accuracy
TM	[119]	2012	Kittler’s optimal thresholding		BCCCF	92.0% to 95.0% Spearman and 6.9% average density
TM	[109]	2013	Otsu	Median
TM	[108]	2014	Rough set theory	Median	MIAS
TM	[107]	2014	Otsu thresholding	Morphological and median	DDSM
TM	[114]	2014	Threshold and evolutionary	Average	DDSM	95.2% accuracy
TM	[110]	2014	Otsu	Median	MIAS
TM	[115]	2015	Global threshold	Median	MIAS	92.86% accuracy and acceptable level of 4.97%
TM	[111]	2015	Global thresholding and merging	Wiener		82.0% accuracy and 18.0% error detection
TM	[105]	2016	Morphological threshold	Median	MIAS	94.54% accuracy and 5.45% false identification
TM	[106]	2016	Adaptive threshold			91.5% accuracy for SVM and 70.0% accuracy for -NN
TM	[121]	2016	Otsu	Morphological	WHC and DDSM	100.0% accuracy for WHC and 91.30% for DDSM
TM	[104]	2017	Otsu	Clahe	MIAS	96.0% accuracy
TM	[116]	2017	Histogram and edge detection	Gaussian	MIAS and EPIC	98.8% accuracy (MIAS) and 91.5% (EPIC)
TM	[112]	2018	Adaptive global and local threshold	Meteorological	MIAS	91.3% sensitivity and 0.71% false positive
EM	[128]	2004	Edge	2-D	MIAS	92.5% accuracy, 93.0% sensitivity, and 85.0% specificity
EM	[122]	2006	Edge		MAGIC-5 collaboration	86.20% ROC and 82.0% sensitivity
EM	[126]	2009	Histogram	Morphological	MIAS	97.0% accuracy
EM	[133]	2011	Active contour	Binary homogeneity	MIAS	99.6% CM, 98.7% CR, and 98.3% quality
EM	[131]	2011	Energy minimisation and contour		MIAS	90.0% accuracy and 92.27% precision
EM	[134]	2011	Edge	Median	KHCCJH	94.1% accuracy (CC), 81.4% MLO, and 90.0% accuracy
EM	[130]	2011	Sobel, Prewitt, Laplacian	Adobe Photoshop	NCSM	79.0% AUC for Sobel, 72.0% Prewitt, and 71.0% Laplacian
EM	[127]	2012	Edge	Median	MIAS	83.9% accuracy
EM	[132]	2014	Active contour			88.0% sensitivity
EM	[123]	2015	Dynamic graph cut		MIAS and DDSM	98.88% sensitivity, 98.89% specificity, and 93.0% for negative values
EM	[124]	2015	Canny edge detection	Median	MIAS, INbreast, and BCDR	98.8% Dice boundary of 97.8% MIAS, 98.9% for boundary 89.6% INbreast, and 99.2% for boundary, and 91.9% BCDR
EM	[135]	2017	Cascade	Gabor	UHGL	100.0% sensitivity and 3.4% false positives
EM	[129]	2017	Edge		NCSM	84.0% AUC