Table 1. Summary of previous studies.

Authors (Year)	Dataset/Availability	Classification algorithm	Classification type
Qian et al. (2024)	INbreast/Moreira et al. (2012)	Multi-feature fusion neural network (MFNet)	Benign group (BI-RADS 1, 2, and 3), malign group (BI-RADS 4, 5, and 6), BI-RADS 2 and 3 are discarded
Achak & Hedyehzadeh (2023)	CDD-CESM/Khaled et al. (2022)	Resnet-50, Resnet-18, and Densenet-201; the K-Fold10 technique	Benign and malignant non-mass enhancement (NME) lesions
He et al. (2023)	CBIS-DDSM/Lee et al. (2017)	Semantic Pyramid Network with a Transformer Self-attention (SPN-TS)	Detection/Benign and malign
Lou et al. (2022)	INbreast/Moreira et al. (2012)	Deep CNN	Benign group (BI-RADS 1 and 2), malign group (BI-RADS 4, 5, and 6), BI-RADS 2 and 3 are discarded
Walton, Kim & Mullen (2022)	CBIS-DDSM/Lee et al. (2017)	CNN	Co-locating lesions, classification according to the size of the mass type lesions
Malebary & Hashmi (2021)	DDSM + MIAS/Heath et al. (1998), Suckling et al. (2015)	ResNet, LongShortTermMemory of RecurrentNeuralNetwork(RNN-LSTM) + ResNet-ResNetnetwork + ResNet-VGG	Normal (BI-RADS 1), benign (BI-RADS 2 and 3), malign (BI-RADS 4, 5, 6)
Aly et al. (2021)	INbreat/Moreira et al. (2012)	YOLO-V3; ResNet and InceptionV3	Mass classification, benign (BI-RADS 2 and 3), malign (BI-RADS 4, 5, and 6)
Cai et al. (2020)	Nanfang Hospital (NFH), Guangzhou/Cai et al. (2019)	Deep learning method using neutrosophic boosting	Grading micro-calcification clustering: Three group classification, BI-RADS 3, 4, and 5
Al-antari, Han & Kim (2020)	DDSM + INbreast/Heath et al. (1998), Moreira et al. (2012)	YOLO detector, modified InceptionResNet-V2 classifier	Clasificaton of breast lesions (Benign and malign)
Lu, Loh & Huang (2019)	Local dataset	Fully cannected CNN	Binary (Benign and malign)
Gandomkar et al. (2019)	Local dataset	Convolutional Neural Networks (CNN), Visual Analogue Scales (VAS)Inception-V3 pre-trained on the ImageNet	The low-risk group (BI-RADS 1 and 2), the high-risk group (BI-RADS 3 and 4)
Kim et al. (2018)	Yonsel University health system/In-house dataset	ResNet-5 0	Sorting out cases (BI-RADS 1 cases and others)
Jung et al. (2018)	INbreast + GURO/Moreira et al. (2012), In-house dataset	RetinaNet	A mass detection based model, malignancy Binary (Benign and malignant)
Dhungel, Carneiro & Bradley (2017)	INbreat/Moreira et al. (2012)	Deep learning, Bayesian optimization, Transfer learning	No mass (BI-RADS 1), benign mass (BI-RADS 2 and 3), malignant mass (BI-RADS 4, 5, and 6)
Geras et al. (2017)	Health Insurance Portability and Accountability (HIPAA)	Multi-ViewDeepConvolutionalNeuralNetwork	Incomplete (BI-RADS 0), negative (BI-RADS 1), benign mass (BI-RADS 2)
Arevalo et al. (2016)	BCDR/(Lopez et al., 2012)	Various CNN algorithms: CNN3, CNN2, HGD, HOG, DeCAF, and Hcfeats	Mass malignancy Binary (Benign and malign)
Lévy & Jain (2016)	DDSM/(Heath et al., 1998)	ShallowCNN (the baseline model), AlexNet, and GoogLeNet	Binary (Benign and malign) BI-RADS 0, 1, 2, 3 as benign and BI-RADS 4,5