Breast Cancer Analysis and Predictions

Based on Ultrasonogram Using Machine

Learning Model
Md. Shakibul Islam (213-15-4422), Md. Rakib Hassan (213-15-4411), Mahmuda Akter Mumu
(213-15-4398)
Abstract— One of the most common cancers in women, breast cancer, tremendously affecting
women’s lives, extending beyond health risks to psychological, social, and economic repercussions.
This research paper represents the view on the uses of Machine Learning (ML) in combination with
ultrasonogram imaging for early detection and accurate diagnosis of cancer in breast. The study uses
a large data of ultrasonogram images of breast tissues, each carefully annotated by experienced
radiologists for building the ML model. A range of machine learning algorithms were tested where it
showed a significant result in enhancing diagnostic accuracy and consistency. The desire is to enable
precise classification of cancer in breast, evaluating risk factors for early and targeted interventions.
Thus, allowing the treatment to be faster and accurate with the help of ML model. And the main goal
of this paper is to lessen the burden of cancer in breast on women, their families, and the healthcare
system.
Index Terms— Breast Cancer, Ultrasound Imaging, Machine Learning (ML), Artificial Intelligence,
Medical Imaging, Early Detection, Accurate Diagnosis, Healthcare Resource Allocation,
Telemedicine, Resource-Limited Regions, Women’s Health.

Introduction
Worldwide, breast cancer in women is stated as the most common cancer in this modern day. This disease
not only have a negative impact on the individuals affected but also on those who are close related and it
also impose negative impact on the healthcare systems tasked with their care. To reduce the negative impact
of breast cancer on the individual suffering from it and also those who are treating her and most importantly
the family members, is by early detection and accurate diagnosis which can be done by modern medical
technologies. In field of modern medical technologies, ultrasound imaging in one of the most suitable tools
for imaging breast tissues. As a non-invasive, cost-efficient tool ultrasound imaging also know as
ultrasonogram has become of the most preferable tool for predicting breast cancer. Sound wave is used in
ultrasonogram to create images of the breast tissues. As it uses sound wave, it makes the work easier for
professionals to take images of dense breast tissues. Further it does not radiate any radiation, making it safer
option for repeated testing and for use in younger women. But now a days science has become so advanced
that we are able to train machines to do the job of detecting cancer by analyzing the imaging data. And this
technology is ML which is a subpart of Artificial Intelligence. In medical field, it has shown prominent
result analyzing large ultrasound image data and also in detecting and predicting the breast cancer. The
machine learning algorithms learn from large datasets of ultrasound images, which have been labeled by
skilled radiologists, to predict the chances of breast cancer in new, unseen images. In the world stage, the
application of ML in breast cancer prediction can reduce the shortage of skilled radiologist, particularly in
low-resource regions where resources are so low that everything needs to be utilize efficiently without
wasting any resources. Using it has opened the doors to more possibilities which has led to more precise
and accurate interpretation of ultrasound images, ultimately increasing the consistency and accuracy of the
breast cancer diagnosis. Furthermore, it can be implemented in telemedicine where remote diagnosis and
monitoring can be possible. Moreover, the uses of ML in medical imaging not limited to only breast cancer
detection but also lung cancer detection, brain tumor detection, diabetic retinopathy detection etc. Thus, the
combination of medical imaging and Machine Learning algorithm can open path to more effective and
accurate treatment which can be more accessible.
In conclusion, the combination of ultrasound imaging and Machine Learning algorithms produces a tool
for detecting and prognosis of breast cancer which have the potential to substantially enhance the health
outcomes worldwide. Even though these technologies are very efficient, they are just designed to aid the
healthcare professionals but not to replace them for the medical field. The ultimate diagnosis should
invariably be rendered by a competent healthcare professional, taking into consideration the patient's
comprehensive health status and medical history. As we progress, it's imperative to persist in research and
development in this domain, while concurrently contemplating the ethical ramifications of AI in healthcare.

Related Works
A study was done on multimodal classification using mammogram of breast cancer and ultrasonogram
images by Atrey et al. [1]. They have implemented machine learning model to fuse features from both types
of images and achieved very prominent results regarding the prediction and breast cancer classification. It
leverages the strengths of both mammogram and ultrasonogram images and leads to more accurate
diagnosis. The paper “Breast cancer and surgery” [2] provides very detailed information on breast cancer,
including the use of machine learning models for analyzing ultrasound image data and predicting outcomes.
It is very valuable resource as it is really easy for understanding the implementation of advanced
technologies in breast cancer diagnosis and treatment. A survey by Debelee et al. [3] was conducted about
deep learning approach in breast cancer ultrasonogram images analysis. Their work represents a very clear
view on the potential of deep learning (DL) techniques that can be used for improving the result accuracy
rate and efficiency of model for prediction of cancer of breast tissues and diagnosis. A probability-based
optimal deep learning (DL) featuring fusion model for cancer of breast tissue classification from
ultrasonogram images have been developed by Jabeen et al. [4]. The model of theirs showed high accuracy,
stating the usefulness of deep learning (DL) in improving the diagnosis process of breast cancer. A
comparative analysis by Khan et al. [5] was conducted on machine learning techniques for breast cancer
prediction. Their study displays a distinct insight into the performance of different machine learning models
in predicting breast cancer is contributing to the continuous efforts to enhance early breast cancer prediction
and treatment. Ma et al. [6] predicted breast cancers’ molecular subtype and identified interpretable imaging
features utilizing ML algorithms. Their work contributes to personalized treatment strategies by providing
a more precise diagnosis. The study for regrowing of breast cancer segmentation presented a novel approach
for segmenting breast cancer regions in ultrasonogram images using machine learning. This work
contributes to the early prediction and efficient and right diagnosis of breast cancer. The study on
multimodal ultrasonogram [8] demonstrated the heights of machine learning in improving the diagnosis of
breast cancer by analyzing multimodal ultrasound images. Mao et al. [9] did a systematic scoping review
on breast tumor classification using ultrasound elastography with machine learning. And this work of theirs
gave us a clear overview of early detection and treatment for breast cancer. Using ultrasonogram images, a
pre-trained CNN was developed By Masud et al. [10], using ultrasound images, a pre-trained convolutional
neural network was developed for cancer of breast tissue for detection and diagnosis. Their research on
breast cancer detection showed the possibility of deep learning in increasing the accuracy and consistency
for cancer of breast tissues’ diagnosis. And an optimized framework for breast tissue cancer classification
using machine learning (ML) was developed with hard work of Michael et al. [11]. This work gives us a
very detailed approach to analyze and classify cancer and potentially contributing in improving the
diagnosis. It was demonstrated by Moustafa et al. [12] that color doppler ultrasonogram enhances machine
learning (ML) approach for diagnosis of cancer detection in breast tissue. It states that incorporating color
doppler ultrasound into machine learning (ML) model increases the performance of the diagnosis. A deep
learning (DL) approach for classifying the cancer in breast using ultrasonogram images was explored by
Paçal [13]. Using these techniques, could potentially widen the reach of the classification of the professional
who diagnoses breast cancer. Pourasad et al. [14] presented an architecture for analyzing, identifying and
diagnosis of cancer in breast using ultrasonogram images using Machine learning. This research can help
improving the localization of breast cancer, aiding in treatment planning for all the patients suffering from
breast cancer. Also based on ultrasonogram images and numerical dataset was used of identification of
Benign, Malignant and Normal state of cancer in breast was used by Ray et al. [15]. This approach of theirs
could potentially improve the differentiation and accurate detection rate whether it is benign, malignant or
normal sate of breast cancers. Rezazadeh et al. [16] developed a machine learning model that can explain
was built for diagnosis of cancer in breast tissue based on ultrasonogram image texture features. And
approach like this could definitely improve the interpretability of models in cancer detection and diagnosis
affected in breast tissue and also help the professionals to more accurate in treatment of breast cancer. An
exploration of image processing and machine learning (ML) for cancer detection and diagnosis in breast
tissues was done by Sadhukhan et al. [17]. This paper shows how the advanced image processing algorithms
are used for analyzing and detecting breast cancer and its diagnosis. This approach leverages the advanced
image processing algorithms. A algorithm based on bio-imaging for cancer of breast tissues detection was
developed by Safdar et al. [18]. It demonstrated that the potential of integrating bio-imaging techniques
with machine learning can enhance the accuracy of detection and prediction which remarkable. Using
unsupervised machine learning approach, Shia et al. [19] did classification of cancer state whether it is
benign, malignant or normal with ultrasonogram images. For risk prediction in ultrasonogram images,
Singh et al. [20] integrated radiologists feedback with computer supported diagnostic systems. By this
approach of theirs the accuracy of risk prediction could be improved through expert feedback. Further study
was done on deep learning and radiomics for predicting axillary lymph node metastasis by Sun et al. [21].
To analyze the breast cancers’ responds to chemotherapy a quantitative ultrasonogram multi-parametric
images at pre-trained deep learning (DL) was used by Taleghamar et al. [22]. And this approach of theirs
was appreciated by many and more intensive research is going on their work. A computer supported
diagnosis system for cancer in breast tissue, ultrasonogram images using deep learning (DL) developed by
Tanaka et a. [23]. Their work highlighted the role of deep learning (DL), enhancing diagnostic process. A
multi-view CNN with transfer learning for breast cancer identification in automated breast ultrasonogram
was used by Wang et al. [24]. This approach has the possibility to increase the output accuracy of cancer in
breast tissue in prediction. Also, machine learning (ML) model for breast tissue cancer diagnosis
ultrasonogram images of triple negative breast tissue cancer was used, this cancer is a subtype of breast
cancer that is particularly challenging to treat and was researched by Wu et al. [25]. By Xu et al. [26], a
machine learning model for medical breast ultrasound image segmentation was used. This work could
potentially increase the segmentation of breast ultrasound images which could help accurately diagnosis
the breast cancer. To predict HER2 state of cancer in breast tissue on ultrasonogram images, deep learning
(DL) model was used by Xu et al. [27]. Further Zhang et al. [28] evaluated the diagnosis effectivity of the
breast cancer ultrasonogram computer supported prediction model based on CNN for predicting and
detection breast cancer. And this work showed the potential of convolutional neural networks in expanding
the diagnosis process. The accuracy of molecular subtype diagnosis by ultrasonogram image using deep
learning (DL) model was evaluated by Zhang et al. [29]. The work demonstrated a very significant insights
into the performance of the model in diagnosis of breast tissue cancer and its molecular subtypes. But the
research does not show any about limitation and of their model. Finally, from the primary breast cancer US
images Zhou et al. [30] used model to predict lymph node metastasis. This work of theirs provided real
benefit of using deep learning in breast cancer diagnosis which appreciated all over.

Methodology
The dataset is the collection of ultrasound images of breast tissues which is later being used for analyzing
and predicting the breast cancer using machine learning algorithms. This dataset was collected form Kaggle
which is a large store house of datasets of numerous types. After getting this dataset, it was processed
according to the requirements. Then machine learning algorithms were applied and evaluated the results
using confusion matrix and finally demonstrated final decision.
A. Data Collection
The dataset is obtained from Kaggle which consists of ultrasound images of breast tissues. The dataset
contains 3 categories of images they are benign, malignant and normal which are the stated of breast
cancer. Here all these images are used for training, testing and validating after training the machine
learning models and also each categories contains masked images too which are used to show the
affected areas of breast tissues.

B. Data Preprocessing
Data preparation is a very critical part in preparing the dataset for analysis and modeling. In this
preparation a very careful measure must be taken so that best dataset could be prepared for our machine
learning models. It must ensure the cleanliness, organization and usefulness of the dataset for the
research stages. And all the necessary steps should be taken to get most organized and well fitted dataset
for research which will be later effective for training, testing and validation. Here in this research, we
have created an working_dir folder to work on and also train, testing and validation named folders
respectively were created for each containing benign, malignant and normal folder which contains the
images for each stages which will performed in this research and also another folder named
Overlayedimages were created and we have also resized the images to the requirements for the research.
And this way the preprocessing was done to have the processed dataset.

Fig-1.1: Resized Images

Fig-1.2: Overlayed Image
 Fig-1.3: Actual, Mask and Overlayed Images

C. Model Description
Some models had been selected for our research and that is because of these algorithms are efficiency
and reliability. And these models are AlexNet, DenseNet121, EfficientNet, GoogLeNet, MobileNet and
VGG-16.
• AlexNet: A classical convolutional neural network (CNN) method that consists of max pooling
and dense layers is primarily use for images recognition and classification of images and has
been instrumental in the evolution of technologies such as facial recognition, object detection
and automated image tagging and also in many other image processing tasks.
• DenseNet121: DenseNet121 uses dense connections within layers through Dense Blocks. And
each of the layer is interconnected to every other layer and bottleneck layers are used to reduce
the number of parameters without even reducing the number of features learned by the network.
It is widely used for image classification tasks, including transfer learning where pre-trained
models on datasets like ImageNet are adapted for other specific tasks.
• EfficientNet: EfficientNet uniformly measures all dimensions of depth, width and resolution
using a compound coefficient. It uses a set of fixed scaling coefficients. It is highly relevant for
 analyzing large amounts of visual data quickly, making it useful for tasks such as customer
behavior analysis, market trend identification and product or service improvement.
• GoogLeNet: GoogLeNet is Inception method based. It is developed by researchers at Google.
And it is built on 22 layer deep.
• MobileNet: It is spacially designed for mobile and embedded vision applications that uses
depthwise separable convolutions instead of ordinary convolutions, reducing the amount of
computation and improving the computational efficiency of the network. This architecture is
lightweight and efficient that makes it suitable for real-world applications which includes
object detection, fine-grained classifications, face attributes and localization.
• VGG-16: The simplicity of the VGG-16 is the main attraction. It is primarily used for image
recognition tasks which includes object detection, facial recognition and image identification
and also has attained high level of performance in the ImageNet Large Scale Visual Recognition
Challenge for image classification. It is really efficient as image processing model and it very
fast and reliable for large datasets.

D. Training
For training, testing and validation a certain amount data was kept to build a model that could
potentially predict breast cancer based on ultrasonogram images. A part of dataset comprising around
72.25% of the dataset was chosen to train the models. This training dataset was used to build and fit
machine learning models such AlexNet, DenseNet121, EfficientNet, GoogLeNet, MobileNet and
VGG-16. These models were trained with large amounts of images from the dataset to provide the
knowledge to accurately make predictions and classifications. This approach made way to further
model evaluation and model validation. And this is very import task in assessing the models’
functionality and ability to predict the cancer in breast tissue based on ultrasonogram images.

E. Testing
For testing the models 15% of the image dataset was chosen to test accuracy of the models. This
segmentation ensured that a significant portion of the dataset was kept out of model training to test the
models’ generalization abilities effectively. The confusion matrix, a popular evaluation approach was
used to determine how well the models predicted the benign, malignant and normal stages of the breast
cancer. And the confusion matrix provided a detailed breakdown of the model’s outputs which allows
us to analyze the result. In the context of investigation, the confusion matrix was used to construct many
performance measures including precision, recall, f1-score and ROC curve.

Table I - Confusion Matrix Measures

Measures Definition Formula
Accuracy establishes how well the
Accuracy Accuracy = (TP+TN) / (Total no of samples)
program predicts instances
Precision assesses the accuracy and
Precision Precision = TP / (TP+ FP)
correctness of the classification factors
Recall is utilized to assess the vividness
Recall Recall = TP / (TP+FN)
or fairness of the classifier
The calculated mean of recall and
F1-Score F1-Score = 2*(P*R) / (P+R)
precision is known as F1-Score
Test utility is compared using ROC
ROC Curve
(Receiver Operating Curve) curves.
 Result Analysis and Discussion
A total six number of classifiers and boosting strategies were used to evaluate the dataset consisting of
ultrasound images of breast tissues.

Table II - Accuracy Table

Precision Recall F1-Score
Name Accuracy
Benign Malignant Normal Benign Malignant Normal Benign Malignant Normal
AlexNet 0. 97 0. 94 1.0 0. 97 0. 94 1.0 0. 97 0. 94 1.0 96.58%
DenseNet121 0.98 1.0 1.0 1.0 0.97 1.0 0.99 0.98 1.0 99.15%
EfficientNet 0.98 1.0 0.95 0.98 0.97 1.0 0.98 0.98 0.98 98.30%
GoogLeNet 0.97 1.0 1.0 1.0 0.94 1.0 0.98 0.97 1.0 98.30%
MobileNet 0.97 0.97 1.0 0.98 0.94 1.0 0.98 0.95 1.0 97.44%
VGG-16 0.97 1.0 1.0 1.0 0.94 1.0 099 0.97 1.0 98.29%

To predict the cancer in breast tissue based on ultrasonogram images, numerous machine learning
techniques have been applied. From above table DenseNet121 was most accurate among the other models,
scoring with the highest of 99.15% accuracy. The second spot was taken by EfficientNet and GoogLeNet
with the accuracy of 98.30%. The third place was taken by VGG-16. This model scored 98. 29% which is
0.01% less than the EfficientNet and GoogLeNet. Then fourth came MobileNet which scored 97.44%. And
finally, the last position was taken by AlexNet which scored 96.58%. Considering all the techniques
DenseNet121 perform the best when it comes to accuracy and predicting the accurate result. And poorest
among all was AlexNet. These findings were taken together to enhance the understanding of predicting the
breast cancer and provide useful information for future reseach and usage.

Fig-02: Accuracy Comparison.

Features with the greater significance scores are thought to have more of an impact when classifying things.
A. Evaluation
Using the combination of Numpy, Seaborn and Matplotlib a complete confusion matrix has been
constructed to showcase the output result of the model DenseNet121. With an exceptional accuracy of
99.15% surpassing other machine learning models. When compared with other models, this algorithm
showed greater precision, recall and F1 score. The evaluation is supported by the confusion matrix and
highlights DenseNet121’s ability to provide very promising outcomes.

Fig-03: Confusion matrix and Classification Report of DenseNet121

This confusion matrix shows DenseNet121 model only predict one wrong result after observing the Fig-
03. It is evident that DenseNet121 model showed most accuracy for our research. And it is really efficient
and have greater significance when classifying things using this algorithm.

Conclusion and Future Work

Using different kinds of machine learning model for to examine our dataset for predicting breast cancer.
The finding shows how machine learning model can improve the knowledge of breast cancer assessment.
The investigation of the model showed notable accomplishment with the DenseNet121 algorithm which
showed highest accuracy of 99.15%. Although this research has offered important new information on
predicting breast cancer base on ultrasound images, it is very important to acknowledge its shortcomings
and any potential gaps in the research. In this model there is always exists computational complexity,
difficulty in modifying output layers, overfitting, require large amount data and most importantly long
training time. With these shortcomings can affect results in real world execution and also new unique issues
might show up based on the uses. By addressing these issues, we can improve our model and increase
prediction ability of the model. The future research should be concentrating on these issues on improving
model for cancer in breast tissue prediction. There are existing opportunities to improve breast cancer
prediction capability of the model with real world clinical applications giving access to new data and more
complex ones which will lead to a better model that can predict cancer in breast tissue based on
ultrasonogram images with most accuracy and most right results. Thus, resulting in a more defined and
well-organized model can be built out it with the help of professional in medical field. And the more
machine learning model are used in medical field in real world environment, the more betterment of the
model is possible which result in more accurate result in future usage.

Acknowledgement
The Kaggle has been thanked by the authors for delivering the dataset of ultrasound images of breast tissues
of breast consisting of three stages which are benign, malignant and normal which were utilized in this
research.

References
1. Atrey, K., Singh, B. K., & Bodhey, N. K. (2023). Multimodal classification of breast cancer using
feature level fusion of mammogram and ultrasound images in machine learning paradigm. Multimedia
Tools and Applications, 83(7), 21347–21368.
2. Breast cancer and surgery. (n.d.). Google Books.
3. Debelee, T. G., Schwenker, F., Ibenthal, A., & Yohannes, D. (2019). Survey of deep learning in breast
cancer image analysis. Evolving Systems, 11(1), 143–163.
4. Jabeen, K., Khan, M. A., Alhaisoni, M., Tariq, U., Zhang, Y., Hamza, A., Mickus, A., & Damaševičius,
R. (2022). Breast Cancer Classification from Ultrasound Images Using Probability-Based Optimal
Deep Learning Feature Fusion. Sensors, 22(3), 807.
5. Khan, M. M., Islam, S., Sarkar, S., Ayaz, F. I., Ananda, M. K., Tazin, T., Albraikan, A. A., & Almalki,
F. A. (2022). Machine learning based comparative analysis for breast cancer prediction. Journal of
Healthcare Engineering, 2022, 1–15.
6. Ma, M., Liu, R., Chen, W., Xu, W., Xu, Z., Wang, S., Wu, J., Pan, D., Zheng, B., Qin, G., & Chen, W.
(2021). Predicting the molecular subtype of breast cancer and identifying interpretable imaging features
using machine learning algorithms. European Radiology, 32(3), 1652–1662.
7. Machine learning and Region Growing for Breast Cancer Segmentation. (2019, April 1). IEEE
Conference Publication | IEEE Xplore.
8. Machine learning to improve breast cancer diagnosis by multimodal ultrasound. (2018, October 1).
IEEE Conference Publication | IEEE Xplore.
9. Mao, Y., Lim, H., Ni, M., Yan, W., Wong, D. W., & Cheung, J. C. (2022). Breast Tumour Classification
Using Ultrasound Elastography with Machine Learning: A Systematic Scoping Review. Cancers, 14(2),
367.
10. Masud, M., Hossain, M. S., Alhumyani, H., Alshamrani, S. S., Cheikhrouhou, O., Ibrahim, S.,
Muhammad, G., Rashed, A. E. E., & Gupta, B. B. (2021). Pre-Trained Convolutional Neural Networks
for breast cancer detection using ultrasound images. ACM Transactions on Internet Technology, 21(4),
1–17.
11. Michael, E., Ma, H., Li, H., & Qi, S. (2022). An optimized framework for breast cancer classification
using machine learning. BioMed Research International, 2022, 1–18.
12. Moustafa, A. F., Cary, T. W., Sultan, L. R., Schultz, S. M., Conant, E. F., Venkatesh, S. S., & Sehgal, C.
M. (2020). Color doppler ultrasound improves machine learning diagnosis of breast cancer.
Diagnostics, 10(9), 631.
13. Paçal, İ. (2022). Deep learning approaches for classification of breast cancer in ultrasound (US) images.
Iğdır Üniversitesi Fen Bilimleri Enstitüsü Dergisi (Online), 12(4), 1917–1927.
14. Pourasad, Y., Zarouri, E., Parizi, M. S., & Mohammed, A. S. (2021). Presentation of novel architecture
for diagnosis and identifying breast cancer location based on ultrasound images using Machine
learning. Diagnostics, 11(10), 1870.
15. Ray, R., Abdullah, A. A., Mallick, D. K., & Dash, S. R. (2019). Classification of Benign and Malignant
Breast Cancer using Supervised Machine Learning Algorithms Based on Image and Numeric Datasets.
Journal of Physics. Conference Series, 1372(1), 012062.
16. Rezazadeh, A., Jafarian, Y., & Valeshabad, A. K. (2022). Explainable ensemble machine learning for
breast cancer diagnosis based on ultrasound image texture features. Forecasting, 4(1), 262–274.
17. Sadhukhan, S., Upadhyay, N., & Chakraborty, P. (2019). Breast cancer diagnosis using image
processing and machine learning. In Advances in intelligent systems and computing (pp. 113–127).
18. Safdar, S., Rizwan, M., Gadekallu, T. R., Javed, A. R., Rahmani, M. K. I., Jawad, K., & Khan, S. B.
(2022). Bio-Imaging-Based machine learning algorithm for breast cancer detection. Diagnostics, 12(5),
1134.
19. Shia, W., Lin, L., & Chen, D. (2021). Classification of malignant tumours in breast ultrasound using
unsupervised machine learning approaches. Scientific Reports, 11(1).
20. Singh, B. K., Verma, K., Panigrahi, L., & Thoke, A. S. (2017). Integrating radiologist feedback with
computer aided diagnostic systems for breast cancer risk prediction in ultrasonic images: An
experimental investigation in machine learning paradigm. Expert Systems With Applications, 90, 209–
223.
21. Sun, Q., Lin, X., Zhao, Y., Li, L., Yan, K., Liang, D., Sun, D., & Li, Z. C. (2020). Deep Learning vs.
Radiomics for Predicting Axillary Lymph Node Metastasis of Breast Cancer Using Ultrasound Images:
Don’t Forget the Peritumoral Region. Frontiers in Oncology, 10.
22. Taleghamar, H., Jalalifar, S. A., Czarnota, G. J., & Sadeghi‐Naini, A. (2022). Deep learning of
quantitative ultrasound multi-parametric images at pre-treatment to predict breast cancer response to
chemotherapy. Scientific Reports, 12(1).
23. Tanaka, H., Chiu, S. W., Watanabe, T., Kaoku, S., & Yamaguchi, T. (2019). Computer-aided diagnosis
system for breast ultrasound images using deep learning. Physics in Medicine & Biology/Physics in
Medicine and Biology, 64(23), 235013.
24. Wang, Y., Choi, E. J., Choi, Y., Zhang, H., Jin, G. Y., & Ko, S. (2020). Breast Cancer Classification in
Automated Breast Ultrasound Using Multiview Convolutional Neural Network with Transfer Learning.
Ultrasound in Medicine & Biology, 46(5), 1119–1132.
25. Wu, T., Sultan, L. R., Tian, J., Cary, T. W., & Sehgal, C. M. (2018). Machine learning for diagnostic
ultrasound of triple-negative breast cancer. Breast Cancer Research and Treatment, 173(2), 365–373.
26. Xu, Y., Wang, Y., Yuan, J., Cheng, Q., Wang, X., & Carson, P. L. (2019). Medical breast ultrasound
image segmentation by machine learning. Ultrasonics, 91, 1–9.
27. Xu, Z., Yang, Q., Li, M., Gu, J., Du, C., Yang, C., & Li, B. (2022). Predicting HER2 status in breast
cancer on ultrasound images using deep learning method. Frontiers in Oncology, 12.
28. Zhang, H., Han, L., Chen, K., Peng, Y., & Lin, J. (2020). Diagnostic efficiency of the Breast Ultrasound
Computer-Aided Prediction Model based on convolutional neural network in breast cancer. Journal of
Digital Imaging, 33(5), 1218–1223.
29. Zhang, X., Li, H., Wang, C., Cheng, W., Zhu, Y., Li, D., Jing, H., Li, S., Hou, J., Li, J., Li, Y., Zhao, Y.,
Mo, H., & Pang, D. (2021). Evaluating the accuracy of breast cancer and molecular subtype diagnosis
by ultrasound image Deep learning model. Frontiers in Oncology, 11.
30. Zhou, L., Wu, X. Q., Huang, S., Wu, G., Ye, H., Wei, Q., Bao, L., Deng, Y., Li, X., Cui, X., & Dietrich,
C. F. (2020). Lymph Node Metastasis Prediction from Primary Breast Cancer US Images Using Deep
Learning. Radiology, 294(1), 19–28.

“The End"