Table 9.

Analysis of BC diagnosis based on DL and NN approaches in Hybrid using information from reviewed articles.

References	Approach	Targets for ML	Targets for BC	Advantages of the ML approach	Advantages of the BC approach	Disadvantages	Evaluation criteria
							AUC	ACC	SPE	SEN	Recall	F-Score
[108]	DBN	•Algorithm optimization	•Optimization of BC classification	•Assists in optimizing decisions •Assists in the removal of noise	•Decrease noise •Better performance •Automated analysis of images	•A lack of use of clinical biomarker databases	✘	✔	✔	✔	✘	✔
[109]	CNN	•Present a deep learning-based approach to image analysis	•Improvement of BC classification	•Assisting the optimizer in making predictions	•Assisting the experts in reducing genomic tests	•A lack of multiclass classification •Failure to use effective clinical parameters	✘	✔	✔	✔	✘	✘
[110]	CNN	•Present a novel approach to detection and classification based on deep learning	•Improve the early detection and classification	•Enhance the effectiveness of diagnosis and classification	•Achieving a more accurate diagnosis and classification	•Insufficient use of effective clinical parameters	✘	✔	✘	✔	✔	✔
[111]	ANN	•Proposed a CAD system based on meta-heuristic algorithm for tuning the parameters of the neural network	•Identifying the type of BC	•Improve the detection and classification processes •Reduce the cost of computing	•Increasing the potential and accuracy of diagnosis and classification	•Inability to implement the proposed approach on large and real clinical datasets	✘	✔	✔	✔	✘	✘
[112]	CNN	•Developing a deep learning model for training and regulating models	•Improved accuracy in the classification of BC	•Improved accuracy and reduced time of image recognition and classification	•Increased efficiency in diagnostics •Reduction in the cost of diagnosis	•Lack of selection of effective clinical parameters in the process of BC diagnosis •Lack of comparison between the proposed approach and existing powerful techniques	✘	✔	✘	✘	✘	✘
[113]	ANN	•Presenting a comparative process between two neural network-based algorithm	•Finding factors associated with de novo metastases in invasive BC	•Introduction of ANN as a reliable model in the process of identifying new metastases in invasive BC compared to the traditional model	•Improved detection of new metastases of BC	•Unreliability of results due to the comparison between only two techniques	✔	✔	✔	✔	✘	✘
[114]	CNN	•Propose an approach for combining deep learning, transfer learning, and generative adversarial network to improve classification performance	•Effectively classifying BC	•Optimizing image classification performance	•Reducing noises caused by GAN images •Enhancing classification accuracy based on a limited training dataset.	•Ineffective selection of clinical parameters during the diagnostic process	✘	✔	✘	✘	✘	✘
[115]	CNN	•Propose an automated method for the binary classification of BC tumors as either malignant or benign	•Enhancing the classification mechanism in the diagnosis process	•Improving classification performance and accuracy based on new models	•Assisting in the accurate identification of dangerous malignant tumors from harmless benign tumors •Providing assistance to doctors in low-resource environments	•Inability to create effective classification models using updated CNN models •Inability to implement the proposed approach when confronted with large and real datasets	✘	✔	✔	✔	✘	✔
[116]	DNN	•Developing a new method based on DNN that can be used to detect BC	•Achieve high-quality and efficient image classification	•Provide very accurate and effective results compared to similar methods	•Accurate diagnosis of BC using the proposed approach •Early detection of BC using the proposed approach	•Inadequate use of powerful prediction and comparison techniques	✘	✔	✔	✘	✔	✘
[117]	CNN	•Presenting a novel deep learning model to achieve enhanced classification.	•Automating the classification of BC	•Improve the speed of diagnosis •Enhance the accuracy of diagnosis	•Ability to accurately detect and classify images using automatic classification approach	•Inability to use the proposed approach on other datasets •Lack of application of the proposed approach to actual datasets	✘	✔	✔	✔	✘	✔
[118]	CNN	•Presenting a hybrid approach for accurate classification of breast images	•Enhancing the accuracy of image classification	•Improve classification performance in interactive cross task extreme learning machine.	•Providing an efficient tool for BC classification in clinical settings •Accurate classification performance of remarkable quality	•A lack of consideration for images that cause noise to reduce the accuracy of classification •Failure to test the proposed approach in different environments	✘	✔	✔	✔	✘	✘
[119]	CNN	•Propose a modified architecture for extracting images' features efficiently	•Incorporating effective features into BC classification	•Enhancing image classification accuracy through automatic classification models	•Assist to classify breast lesions as benign and malignant •Increasing diagnostic reliability	•Failure to use CADx systems to extract lesion characteristics and robust classification of BC effectively •Failure to evaluate the proposed approach with specific algorithms for detecting lesions more accurately	✔	✘	✔	✔	✘	✘
[120]	DNN	•Propose a novel hybrid convolutional and recurrent deep neural network for image classification.	•Achieve an effective classification system for BC	•Using an automated method for analyzing images accurately	•Alleviate the problem of relatively low classification accuracy of benign images •The possibility of more authentic and finer recognition in classifying classes	•No direct use of low-resolution images as an input •Failure to apply the proposed approach to all available images in datasets	✘	✔	✘	✔	✘	✘
[121]	DNN	•Propose a deep learning-based approach to breast mass classification in sonography	•Finding a mechanism to improve the accuracy of image classification	•Improve the overall performance of the transfer learning techniques using DNN •Improved classification accuracy by using pre-trained DNN	•Providing assistance to radiologists in the classification of breast masses •Efficacy of the proposed approach as a clinical tool for diagnosis	•Failure to determine the effectiveness of the presented approach in the other datasets •Failure to compare the proposed approach with other algorithms for accurate diagnosis	✔	✔	✔	✔	✘	✘
[122]	DNN	•Developing an automated BC detection model based on images classification	•Implementing a strategy aimed at improving image classification	•Resolving the shortcomings caused by image classification in existing methods	•Detects BC masses with a high degree of accuracy •Provided a performance comparable to human sonographers in image classification.	•Failure to use the proposed approach in large data sets and real-world scenarios •Lack of use of the technique of combining different image methods to improve accuracy	✘	✔	✔	✔	✘	✔