Table 1 Summary of literature review.

Humayun M. et al²¹.

InceptionResNetV2 deep learning model for breast cancer risk prediction

Achieved 91% accuracy, integrated risk markers for improved assessment

Limited to specific datasets, may not generalize well

Obayya M. et al²².

AOADL-HBCC technique using SqueezeNet and Adamax optimizer for breast cancer classification

Maximum accuracy of 96.77%, enhanced decision-making

Complex model integration, high computational cost

Al-Jabbar M. et al²³.

Combined CNN models and handcrafted features for breast cancer identification

High accuracy, robust feature extraction

Complexity in model integration, may not scale well

Kode H. & Barkana B. D²⁴.

Assessed efficacy of VGG16, CNN, and knowledge-based systems for feature extraction

Up to 98% accuracy with knowledge-based characteristics

Reliance on extensive preprocessing, high computational resources required

Abdulaal A. H. et al²⁵.

Pre-trained deep neural networks with hierarchical self-learning for sub-image classification

Achieved 99.1% accuracy, improved label accuracy

High complexity, potential issues with noisy labels

Mondol R. K. et al²⁶.

Hist2RNA method using H&E-stained WSIs for gene expression prediction

High accuracy in gene prediction, significant prognostic value

Limited to specific data types, may not generalize across different datasets

Amin M. S. & Ahn H²⁷.

FabNet model for multi-scale histopathological image analysis

High classification accuracy with fewer parameters

Complexity in hierarchical model structure

Guleria H. V. et al²⁸.

CNN model with VAE and DVAE for image reconstruction and prediction

Improved prediction accuracy

Lower accuracy compared to specialized CNN models

Peta J. & Koppu S²⁹.

Deep learning and federated learning-based automated illness diagnosis system

High accuracy and specificity, enhanced security through encryption

High computational cost, complexity in implementation

Peta J. & Koppu S³⁰.

Explainable deep learning method with Adaptive Unsharp Mask Filtering (AUMF) and ESAE-Net for breast cancer

Enhanced precision and interpretability

Increased computational overhead due to multiple explainable algorithms

Sharmin S. et al³¹.

Ensemble-based machine learning with ResNet50 V2 for breast cancer detection

High precision and accuracy, effective pattern recognition

High computational overhead, potential issues with model interpretability

Patel V. et al³².

GARL-Net: Graph-based adaptive regularized learning system using DenseNet121 for breast cancer classification

Achieved precision of 99.00%, high recall and F1 scores

Potential misclassification issues, high complexity in loss function

Jabeen K. et al³³.

Haze-reduced local-global enhancement technique combined with EfficientNet-b0 for breast cancer classification

High accuracy on multiple datasets, robust against visual contrast variations

High complexity, may require extensive computational resources

Anwar F. et al³⁴.

CAD system combining ResNet, DCNN features with WPD and HOG, followed by PCA for feature reduction

High accuracy in distinguishing benign and malignant BC

Dependence on extensive feature extraction and preprocessing steps

Attallah O. et al³⁵.

Histo-CADx: CAD system with two-fold feature fusion and MCS for breast cancer classification

High accuracy, reduced computation cost through auto-encoder fusion

Complexity in feature fusion process, may require extensive validation