Artificial Intelligence-Based Breast Cancer Diagnosis Using Ultrasound Images and Grid-Based Deep Feature Generator

Figures & data

Figure 1 Graphical outline of the proposed grid-based deep framework.

Figure 2 The created best model for the used ultrasonic image dataset using our proposed grid-based deep transfer learning framework.

Table 1 The Generated Confusion Matrix

True Class	Predicted Class
	Benign	Malign	Normal
Benign	431	3	3
Malign	7	203	2
Normal	8	1	124
Recall (%)	98.63	96.67	93.23
Precision (%)	96.64	98.07	97.64
F1-score (%)	97.62	97.36	95.38

Table 2 Overall Results of the Proposed Grid-Based Deep Learning Model on the Used Ultrasonic Image Dataset

Performance Metric	Result (%)
Accuracy	97.18
Overall precision	97.45
Unweighted average recall	96.18
Overall F1-score	96.79
Geometric mean	96.15

Figure 3 Fold-wise accuracies of the grid-based deep learning model on the used dataset.

Table 3 Accuracies of the Used 16 Pre-Trained Networks Using Cubic SVM with 10-Fold Cross Validation

Network	Accuracy	Network	Accuracy
ResNet18	88.21	ShuffleNet	86.03
ResNet50	88.21	DensNet201	88.21
ResNet101	90.38	Inceptionv3	87.44
DarkNet19	88.46	InceptionResNetV2	88.33
MobileNetV2	89.10	GoogLeNet	86.92
DarkNet53	87.95	AlexNet	87.82
Xception	85.64	VGG16	85.64
EfficientNetb0	88.97	VGG19	85

Note: Bold values for models show the best three models chosen, which achieved best accuracies.

Figure 4 Number of features and misclassification rate (error rate) of the INCA for this work.

Table 4 Comparison with the State-of-the-Arts Using the Same BUS Dataset

Study	Year	Method	Classifier	Split Ratio	The Accuracy Result (%)
Moon et al^Citation40	2020	Convolutional neural networks	Ensemble classifiers	80:20	94.62 for 3 classes (Benign, Malignant, Normal)
Byra et al^Citation41	2020	Selective kernel U-Net convolutional neural network	Softmax	5-fold cross-validation	95.60 for 2 classes (Benign, Malignant)
Sadad et al^Citation42	2020	Hilbert transform	Ensemble classifiers	10-fold cross-validation	96.60 for 2 classes (Benign, Malignant)
Mishra et al^Citation43	2021	Grey-level co-occurrence matrix	AdaBoost classifier	70:30	97.40 for 2 classes (Benign, Malignant)
Muduli et al^Citation44	2021	Convolutional neural network	Softmax	5-fold cross-validation	89.73 for 2 classes (Benign, Malignant)
Joshi et al^Citation45	2021	Convolutional neural network	Softmax	5-fold cross-validation	95.25 for 3 classes (Benign, Malignant, Normal)
Zhang et al^Citation46	2021	Soft and hard attention multi-task learning model	Soft and hard attention multi-task learning model	7-fold cross-validation	94.12 for 2 classes (Benign, Malignant)
Our method		Grid-based deep feature generator	Deep neural network	10-fold cross-validation	97.18 for 3 classes (Benign, Malignant, Normal)

Moon WK, Lee Y-W, Ke -H-H, Lee SH, Huang C-S, Chang R-F. Computer‐aided diagnosis of breast ultrasound images using ensemble learning from convolutional neural networks. Comp Methods Prog Biomed. 2020;190:105361. doi:10.1016/j.cmpb.2020.105361

 PubMed Web of Science ®Google Scholar

Byra M, Jarosik P, Szubert A, et al. Breast mass segmentation in ultrasound with selective kernel U-Net convolutional neural network. Biomed Signal Proc Control. 2020;61:102027. doi:10.1016/j.bspc.2020.102027

 PubMed Web of Science ®Google Scholar

Sadad T, Hussain A, Munir A, et al. Identification of breast malignancy by marker-controlled watershed transformation and hybrid feature set for healthcare. Appl Sci. 2020;10(6):1900. doi:10.3390/app10061900

 Google Scholar

Mishra AK, Roy P, Bandyopadhyay S, Das SK. Breast ultrasound tumour classification: a Machine Learning—Radiomics based approach. Expert Syst. 2021;38:e12713.

 Web of Science ®Google Scholar

Muduli D, Dash R, Majhi B. Automated diagnosis of breast cancer using multi-modal datasets: a deep convolution neural network based approach. Biomed Signal Proc Control. 2021;71:102825.

 Web of Science ®Google Scholar

Joshi RC, Singh D, Tiwari V, Dutta MK. An efficient deep neural network based abnormality detection and multi-class breast tumor classification. Multimedia Tools Appl. 2021;1–21. doi:10.1007/s11042-021-11240-0

 PubMed Web of Science ®Google Scholar

Zhang G, Zhao K, Hong Y, Qiu X, Zhang K, Wei B. SHA-MTL: soft and hard attention multi-task learning for automated breast cancer ultrasound image segmentation and classification. Int J Comp Assisted Radiol Surg. 2021;1–7. doi:10.1007/s11548-021-02445-7

 Web of Science ®Google Scholar