Involvement of Machine Learning for Breast Cancer Image Classification: A Survey

Table 9

Convolutional Neural Network.

Reference	Descriptor	Image type	Number of images	Key findings
Wu et al. [78]	Global Features	Mammogram	40	Achieved Sensitivity 75.00% and Specificity 75.00%.
Sahiner et al. [79]	Global Features	Mammogram	168	The achieved ROC score is 0.87.
Lo et al. [80]	Density, size, Shape, Margin	Mammogram	144	The achieved ROC curve is 0.89.
Fonseca et al. [81]	Global Features	Mammogram	—	Breast density classification has been performed utilizing HT-L3 convolution.
				Average achieved obtained Kappa value is 0.58.
Arevalo et al. [82]	Global Features	Mammogram	736	The achieved ROC curve is 0.826.
Su et al. [83]	Global Features	Mammogram	92	Fast Scanning CNN (fCNN) method has been utilized to reduce the information loss.
				The average Precision, Recall, and 1 score are 91.00%, 82.00%, and 0.85, respectively.
Sharma and Preet [84]	GLCM, GLDM Geometrical	Mammogram	40	The best Accuracy achieved is 75.23% and 72.34%, respectively, for fatty and dense tissue classification.
Spanhol et al. [6]	Global Features	Histopathology	7909	The best Accuracy achieved 89 6.6%.
Rezaeilouyeh et al. [85]	Local and Global Features	Histopathology	—	Shearlet transform has been utilized for extracting local features.
				When they utilize RGB image along with magnitude of Shearlet transform together, the Achieved Sensitivity, Specificity, and Accuracy were 84.00 1.00%, 91.00 2.00%, and 84.00 4.00%; when they utilize RGB image along with both the phase and magnitude of Shearlet transform together, the achieved Sensitivity, Specificity, and Accuracy were 89.00 1.00%, 94.00 1.00%, and 88.00 5.00%.