Breast Tumor Detection Using Efficient Machine Learning and Deep Learning Techniques

Machine Learning and Applications: An International Journal (MLAIJ) Vol.10, No.2/3, September 2023
DOI:10.5121/mlaij.2023.10302 17
BREAST TUMOR DETECTION USING EFFICIENT
MACHINE LEARNING AND DEEP LEARNING
TECHNIQUES
Ankita Patra1
, Santi Kumari Behera2
, Prabira Kumar Sethy3
, Nalini Kanta
Barpanda4
and Ipsa Mahapatra5
1,3,4
Department of Electronics, Sambalpur University, Burla Odisha
2,5
Department of CSE, VSSUT Burla, Odisha, India
ABSTRACT
Breast cancer tissues grow when cells in the breast expand and divide uncontrollably, resulting in a lump
of tissue commonly called and named tumor. Breast cancer is the second most prevalent cancer among
women, following skin cancer. While it is more commonly diagnosed in women aged 50 and above, it can
affect individuals of any age. Although it is rare, men can also develop breast cancer, accounting for less
than 1% of all cases, with approximately 2,600 cases reported annually in the United States. Early
detection of breast tumors is crucial in reducing the risk of developing breast cancer. A publicly available
dataset containing features of breast tumors was utilized to identify breast tumors using machine learning
and deep learning techniques. Various prediction models were constructed, including logistic regression
(LR), decision tree (DT), random forest (RF), support vector machine (SVM), Gradient Boosting (GB),
Extreme Gradient Boosting (XGB), Light GBM, and a recurrent neural network (RNN) model. These
models were trained to classify and predict breast tumor cases based on the provided features.
KEYWORDS
breast tumour, machine learning, deep learning, cancer, tumour detection
1. INTRODUCTION
Breast cancer is a serious health concern that affects a significant number of women worldwide.
It is characterized by the growth of abnormal cells in the breast tissue and is the most commonly
diagnosed cancer among women [1]. The prognosis of breast cancer refers to the likelihood or
risk of the disease occurring based on various clinical and non-clinical factors. Prognosis is an
essential aspect of breast cancer treatment and management. It helps patients and healthcare
providers understand the potential outcomes of the disease, which can help guide treatment
decisions and planning. Various factors can influence breast cancer prognosis, including the stage
and subtype of cancer, the patient's age, overall health status, and other medical conditions.
Breast cancer prognosis can be expressed in several ways, such as relapse-free survival (RFS)
and overall survival (OS). The 5-year RFS rate indicates the percentage of patients who have not
experienced a disease recurrence five years after treatment.
On the other hand, the 10-year OS rate reflects the percentage of patients still alive ten years after
diagnosis, regardless of disease recurrence or progression. Accurately predicting breast cancer
prognosis is crucial for patients, healthcare providers, researchers, and policymakers. It can help
guide treatment decisions, evaluate the effectiveness of various therapies, and identify patients
who may require more intensive monitoring or follow-up care. Several predictive models have
been developed and tested in different contexts [2]. Prabira Kumar et al. proposed a classification

18
model utilizing fine-KNN, which incorporates a merger of deep learning features and handcrafted
features extracted from H&E (Hematoxylin and Eosin) images. The aim is to automate
classifying these images for various purposes. With ongoing research and innovation, breast
cancer detection and treatment are continually improving, and survival rates are steadily
increasing. Early detection, personalized treatment, and a better understanding of the disease are
key to reducing the aftermath of breast cancer on women's health and well-being [3].
Turkki et al. argues that performing a prognostic evaluation for breast cancer without prior
knowledge of its histology is possible. They propose a novel approach to breast cancer
classification using five AI classification methods and a specialized CNN model. To assess their
performance, the study evaluates several classification techniques, including logistic regression,
random forests, support vector machine, voting classifier, decision tree, and a custom CNN
model [4-7]. Mohammad Monirujjaman et al. also managed to test different approaches for breast
cancer detection with a modified CNN network. According to a study, the new coronavirus
illness (COVID-19) pandemic may have hindered advances in cancer care. The biggest hurdles
were most likely delays and decreases in cancer screening, diagnosis, and treatment. Many
models have been developed to predict breast cancer prognosis.
Along with it, the range of models has grown significantly, as have their emergence methods,
predictive variables, results, demonstrations, and performance in various contexts [8-13]. Several
studies have explored the practicality of AI and deep learning in healthcare to enhance the quality
and assurance of treatments [14–17]. Bejnordi et al. employed a deep learning system to detect
breast cancer and collate the results with diagnoses made by pathologists. The findings revealed
that automated detection using a deep learning system surpassed human efforts in diagnosis [18-
23]. Patra et al. researched to investigate the performance of advanced deep learning methods in
detecting breast cancer by combining different features. Specifically, the study evaluates the
efficacy of fusing the features of AlexNet, vgg16, and vgg19 in a two-layer fusion approach
when applied to thermal images for breast cancer detection. Breast cancer identification at an
early stage also results in lower treatment costs. Building devices for the visual assessment of the
potential pleasing effects of Breast Conservative Treatment, with potential implications in
reducing the number of secondary measures or in the indicative outcomes of additional therapies
such as radiation therapy or systemic interventions. A method is proposed to augment the Kaggle
breast tumor dataset with synthetic data and apply it via different ML and deep computing
algorithms to compare its accuracy in detecting cancer [24].
2. MATERIALS AND METHODOLOGY
2.1. Dataset Description
The dataset is collected from Kaggle (breast tumor.csv) [25]. It has 568 samples, from which 357
samples indicated benign breast (Class:0) and 212 samples indicated Malignant breast tumors
(class:1) [23]. As we observe in Figure 1 the dataset comprised 32 features: diagnosis,
radius_mean, texture_mean, perimeter_mean, area_mean, smoothness_mean,
compactness_mean, concativity_mean, symmetry_mean, etc.

19
Figure 1. The First 5 Rows of the Dataset
2.2. Data Preprocessing
The breast tumor features dataset was utilized in this study. It has 568 rows and 32 breast tumor
characteristics. 357 data points in the target column (class:0) represent benign healthy breasts,
whereas 212 data points indicate malignant breast cancers. Before model improvement, data
preprocessing is required to remove unwanted noise and anomalies in a dataset that may cause
the model to deviate from the estimated preparation set. The model's evolution aims to remove
any barriers to its creation and evaluation. After gathering the data, it should be cleaned up and
prepared for model construction. Moving ahead, the dataset is examined for mistakes.
Regardless, there are correct attributes in this dataset. After checking, there was no indication of
missing data. Zero missing values according to the data presented in the Figure 2.
Figure 2. Zero Missing Values
2.2.1 Exploratory Data Analysis
Exploratory data analysis (EDA) has always been an excellent tool data scientists use to gain
insights and knowledge from large and complex data sets. Understanding data and its hidden
patterns is quite a task achieved through various statistical and visualization techniques, thus
exploring the given data. In Figure 3 examination is done for the target column containing unique
values.
The primary objective of EDA is to procure basic as well as an in-depth apprehension of the data
sets and their properties, such as their distribution, skewness, kurtosis, outliers, and missing
values. Data visualization is an essential aspect of EDA, as it enables data scientists to portray
obscure data through a visual format, making it clearer and easier to interpret studies. Graphs,
charts, and plots can provide valuable insights into the data, helping data scientists identify
patterns, correlations, and anomalies that may not be immediately apparent from raw data. EDA
techniques allow data scientists to identify data quality issues and perform data cleansing or
transformation if necessary. In Figure 4 Count Plot of the target column is done and in the graph
shown in Figure 5 is data distribution and outliers after plotting of Data.

20
Figure 3. The target column containing unique values
Figure 4. Count Plot of the target column
Figure 5. Data distribution and outliers
2.2.2. Feature Selection
We have used a correlation map to check the correlation among different features. After
analyzing the map properly, we got all the negatively and positively correlated features. And thus
selected the ones which are needed and positively related to the target column, diagnosis. Of
them, 23 features were further selected for modeling in different algorithms. Feature Correlation
Map of all 32 features is seen in Figure 6 and in Figure 7 feature correlation map of selected 23
features are observed.

21
Figure 6. Feature Correlation Map of all 32 features
Figure 7. Feature correlation map of selected 23 features
2.3. Proposed Methodology
The architecture of the prospective methodology is shown in Figure 5. The given dataset was
properly examined and available for model development. A preliminary processed dataset and
machine and deep learning algorithms are vital for model building. Among the techniques
employed are "Logistic Regression," "Decision Tree classification," "Random Forest
classification," and "SVM classifier." The accuracy metrics are the accuracy score, precision
score, recall score, and F1 score. They are used to assess five distinct machine learning models
and a deep learning model once built. As illustrated in the Figure 8, proposed architecture for
breast cancer prediction.
Figure 8. Proposed Architecture for Breast Cancer Prediction

22
2.3.1. Logistic Regression
Logistic regression is a less complex and more efficient solution for discrete and continuous
classification problems, which, unlike its name utilized much more efficiently for categorization.
This paradigm is easy to construct and works well with linearly independent classes. It is a
popular corporate categorizing strategy. The LR model is an approach to statistics used in binary
classification, although it is also used in multiclass classification. Scikit-learn provides an
extremely efficient logistic regression execution. Logistic regression uses the logistic function
below to describe a binary output variable.
Logistic Function=
2.3.2. Decision Tree
Decision tree learning utilizes a divide-and-conquer strategy to find optimal split positions within
a tree. After classifying information under unique labels, the recursive splitting procedure occurs
from top to bottom. Subgroups are formed by testing attributes, known as "recursive
partitioning." The process continues until further splits do not enhance predictions or subsets with
identical target variable numbers. When predicting a record's class label, the comparison starts at
the root, choosing the route based on attribute values.
2.3.3. Random Forest
Random Forest is an extension of bagging that adds randomness and creates a network of
decision trees. The random subspace strategy, or feature bagging, ensures little correlation among
trees by randomly selecting features. Unlike decision trees, random forests evaluate only a subset
of feature splits. It is a powerful and widely used method that handles categorical, continuous,
and binary data.
2.3.4. Support Vector Machine
Support Vector Machine (SVM) is popular due to its high accuracy and minimal resource
consumption. It solves regression and classification problems by identifying a hyperplane that
separates data points based on attributes. The hyperplane's size and placement depend on the
number of features. Support vectors close to the hyperplane impact its orientation and margin.
Utilizing these vectors increases the classifier's margin while removing them shifts the
hyperplane.
2.3.5. Recurrent Neural Network
Recurrent Neural Network (RNN) processes sequential or time series input, incorporating
previous information to influence the current input and output. Unlike conventional networks,
RNNs have memory and produce outputs based on previous sequence elements. Although they
cannot predict future events, they excel in determining the output order.
2.3.6. Ensemble Learning
Ensemble learning is a machine learning method that combines predictions from many models to
improve and enhance the accuracy and robustness of the final forecast. The core idea behind
ensemble learning is to reduce every possible mistake or bias in individual models by using the
aggregate intelligence of numerous models, resulting in a more exact prediction.

23
The primary sources of error in learning models are noise, variation, and bias. Machine learning
ensemble techniques aim to mitigate these error-inducing factors, ensuring the robustness and
longevity of machine learning (ML) computations. By harnessing the power of diverse models
and combining their predictions, ensemble approaches effectively reduce the impact of noise,
mitigate the effects of variation, and mitigate bias. This collective wisdom of multiple models
leads to improved accuracy, generalization, and overall performance in ML tasks.
2.3.7. Bagging and Boosting
Ensemble learning can be classified into two types: bagging and boosting. The primary difference
between these methods is how weak learners are trained. Bagging trains weak learners in parallel,
whereas boosting trains them sequentially. In boosting, a series of models are created, and the
weights of incorrectly classified data in the previous model are increased with each successive
iteration. This redistribution of weights aids the algorithm in identifying which parameters it
should focus on to enhance its performance.
AdaBoost is among the most well-known boosting algorithms, and it was one of the earliest
algorithms of its kind. Other examples of boosting algorithms include XGBoost, Gradient Boost,
and Brown Boost.
2.2.7.1. Bagging
Bagging, also known as Bootstrap Aggregation, is an ensemble learning strategy that aids ML
models' performance and accuracy. It mostly aids in overcoming the overfitting problem in
categorization. This procedure takes random subsets of a dataset and replaces them in a classifier
or regressor. Predictions for each subgroup are then determined using either a majority vote for
classification or averaging for regression, boosting prediction accuracy.
It should be noted that the random forest technique expands the bagging approach, employing
both bagging and attribute unpredictability to construct a forest without the correlation of
decision trees.
2.2.7.2. Boosting
Boosting is a widely used ensemble modeling strategy that aims to combine multiple weak
classifiers to form a single strong classifier. The primary goal of boosting is to boost and refine
the accuracy and generalizability of the model by iteratively learning from the mistakes made by
the previous models.
The boosting algorithm works by sequentially training a series of weak models on the same
training dataset, where each model attempts to correct the general mistakes made by the
previously existing model. During training, the algorithm assigns a higher weight to misclassified
samples, which are then used to train the next weak model. This way, the boosting algorithm
focuses on the most challenging examples, gradually reducing the overall error rate and
improving the model's performance.
2.2.7.2.1. Gradient Boosting
Gradient Boosting is a boosting algorithm that creates a strong learner from multiple weak
learners by training each new model to mitigate the loss function of the previous model using
gradient descent. The algorithm calculates the changing slope of the loss curve concerning the
current ensemble's predictions in each iteration and then trains a new weak model to minimize

24
this gradient. The new model's predictions are then combined with the group, and the process is
repeated until a stopping criterion is met. Loss functions used in gradient boosting can include
mean squared error or cross-entropy. Overall, gradient boosting is a powerful and effective
algorithm for creating strong learners from weak ones.
2.2.7.2.2. XG Boost
XGBoost is an optimized networked gradient boosting framework for efficient and flexible
machine learning model training. It is an ensemble learning approach that integrates the
predictions of numerous weak models to obtain a stronger prediction. Because of its capacity to
handle massive data sets and achieve state-of-the-art efficiency for numerous machine learning
operations such as regression and categorization, XGBoost has grown and expanded to become
one of the most prevalent as well as extensively utilized machine learning algorithms.
Among the numerous benefits of XGBoost, one peculiar asset is its ability to handle missing data
efficiently, eliminating the need for extensive preprocessing and enabling it to work with
incomplete real-life datasets. Additionally, XGBoost's parallel processing capabilities make it
feasible to train models on large datasets in a reasonable amount of time. Its flexibility is
noteworthy, as it allows for fine-tuning various model parameters to enhance overall
performance.
3. RESULTS AND ANALYSIS
The dataset, after being passed through different algorithms, both Machine Learning (Logistic
Regression, Decision Tree, Random Forest, SVM) and Deep Learning (CNN, RNN), the results
of their accuracy in determining whether it is a healthy breast or otherwise was noted and
analyzed. Following is the report obtained:
3.1. Logistic Regression
The Test Accuracy acquired is 97.6%, while the Train Accuracy remains 98.9%. Figure 9 shows
the Precision, Recall, f1-score, and Logistic regression support. In Figure 9, 0 denotes benign
tumor, and 1 denotes malignant tumor data. The f1-score of benign tumors is calculated to be
98%, whereas, for malignant tumors, it is 97%.
Figure 9. Logistic Regression Result

25
3.2. DECISION TREE
The Test Accuracy acquired is 97.4%, while the Train Accuracy remains 98.4%. Figure 10 shows
the Precision, Recall, f1-score, and support of the Decision Tree.
In Figure 10, 0 denotes benign tumor, and 1 denotes malignant tumor data. The f1-score was
calculated to be 96 % for benign tumors, whereas, for malignant tumors, it is 93%.
Figure 10. Decision Tree Result
3.3. XG Boost
The Test Accuracy acquired is 97%, while the Train Accuracy remains 98%. Figure 11 and figure
12 show the Precision, Recall, f1-score, and support of the XG Boost with its confusion matrix.
In Figure 11, 0 denotes benign tumor, and 1 denotes malignant tumor data. The f1-score was
calculated to be 98 % for benign tumors, whereas, for malignant tumors, it is 96%.
Figure 11. XG Boost Result

26
Figure 12. XG Boost Confusion Matrix
3.4. Random Forest
The Test Accuracy acquired is 95.9%, while the Train Accuracy remains 98%. Figure 13 shows
the Precision, Recall, f1-score, and support of the Random Forest. In Figure 13, 0 denotes benign
tumor, and 1 denotes malignant tumor data. The f1-score was calculated to be 97 % for benign
tumors, whereas, for malignant tumors, it is 94%.
Figure 13. Random Forest Result
3.5. Support Vector Machine
The Test Accuracy acquired is 98%, while the Train Accuracy remains 97.2%. Figure 14 shows
the Precision, Recall, f1-score, and support of the Support Vector Machine. In Figure 14, 0
denotes benign tumor, and 1 denotes malignant tumor data. The f1-score was calculated to be 98
% for benign tumors, whereas, for malignant tumors, it is 97%.

27
Figure 14. SVM Result
3.6. Gradient Boosting
The Test Accuracy acquired is 95.6%, while the Train Accuracy remains 100%. Figure 15 and
figure 16 show the Precision, Recall, f1-score, and support with the confusion matrix of the
Gradient Boosting.
Figure 15. Gradient Boosting Result

28
Figure 16. Confusion Matrix of Gradient Boosting
3.7. Light GBM
The Test Accuracy acquired is 98.2 %, while the Train Accuracy remains 100 %. Figure 17,18,19
shows the result of the accuracy of Light GBM.
Figure 17. Accuracy Result of Light GBM

29
Figure 18. Result for Light GBM
Figure 19. Confusion Matrix of Light GBM
3.8. RNN
The Test Accuracy acquired is 97%, while the Train Accuracy remains 98.7%. Figure 20 shows
the result of the accuracy of RNN. It has been depicted in the graph of Model Loss in Figure 21.
Figure 20. Result of RNN

30
Figure 21. Model Loss
3.9. Performance Analysis of Models
We evaluated the performance of different machine learning and deep learning algorithms using
metrics such as F1 score, precision, recall, and accuracy. Our analysis determined that the Light
GBM algorithm achieved the highest accuracy of 98.2%, followed closely by the Support Vector
Machine and Logistic Regression, which achieved an accuracy of 98%. The RNN deep learning
algorithm yielded the third-highest accuracy, i.e., 97%. Decision Tree, on the other hand, had the
lowest accuracy at 95%.
Additionally, the F1 scores for Logistic Regression, Support Vector Machine, and Xtreme
Gradient Boosting were all the same and the highest among the evaluated algorithms. The
Logistic Regression had an accuracy of 98% compared to [4] with 96%. The SVM Classifier
used here also gives higher and more efficient results than other previously used models, with an
accuracy of 98%; [4] compared to it has submitted 97% accuracy with the same approach.
Table 1. Performance Analysis of Models for Breast Tumor Prediction
Models F1 score precision recall accuracy
0 1 0 1 0 1
LR 0.98 0.97 0.97 0.98 0.99 0.95 0.98
DT 0.96 0.93 0.98 0.90 0.94 0.97 0.95
RF 0.97 0.94 0.97 0.94 0.96 0.95 0.96
SVM 0.98 0.97 0.97 0.98 0.99 0.95 0.98
GB 0.97 0.95 0.96 0.97 0.98 0.94 0.96
XGB 0.98 0.96 0.97 0.97 0.98 0.95 0.97
RNN 0.95 0.96 0.97 0.98 0.96 0.98 0.97
LGBM 0.99 0.98 0.98 0.98 0.99 0.97 0.982

31
This model has presented a model with more algorithms, such as RNN and Gradient Boosting.
Thus, the average accuracy rate of the proposed model is higher, with Light GBM having the
highest accuracy rate of 98.2%.
4. CONCLUSION
Breast cancer presents a unique environment for medical diagnosis by taking the patient's
condition and treatment response into account. Machine learning has offered much for the
detection of breast cancer. Despite technological advancements, there are still challenges in
achieving reliable breast cancer screening and monitoring. It is essential to incorporate biological,
social, and demographic data to improve prediction models. In this study, we worked towards
evaluating the effectiveness of several machine learning and deep learning approaches for
classification.
We investigated the performance of four commonly used classifiers: logistic regression, random
forest, decision tree, and support vector machine with sequential minimum optimization.
Additionally, we explored ensemble methods, such as the voting classifier, and deep learning
techniques, specifically the Recurrent Neural Network (RNN).
To assess the performance of these seven classifiers, we utilized various peculiar performance
measures, including accuracy, precision, recall, and F1 score. These metrics provide insights into
the models' predictive capabilities, capturing aspects such as overall correctness, the precision of
positive predictions, the ability to identify true positives, and the balance between precision and
recall.
By evaluating these classifiers and their performance measures, we aimed to contribute to
developing improved breast cancer prediction models, considering both machine learning and
deep learning techniques.
In this study, Light GBM achieved 98.2% accuracy. On the other hand, LR, DT, RF, SVM, GB,
RNN, and XGB got accuracy 98%, 95%,96%,98%, 96%, 97%, 97%, and 97%, respectively. We
can use the fine-tuning of the parameters to boost the accuracy and score of the model.
REFERENCES
[1] Chagpar, Anees B., and Mario Coccia. "Factors associated with breast cancer mortality-per-incident
case in low-to-middle income countries (LMICs)." (2019): 1566-1566.
[2] Sharma, Ganesh N., et al. "Various types and management of breast cancer: an overview." Journal
of advanced pharmaceutical technology & Research 1.2 (2010): 109.
[3] Sethy, Prabira Kumar, and Santi Kumari Behera. "Automatic classification with concatenation of
deep and handcrafted features of histological images for breast carcinoma diagnosis." Multimedia
Tools and Applications 81.7 (2022): 9631-9643.
[4] Turkki, Riku, et al. "Breast cancer outcome prediction with tumor tissue images and machine
learning." Breast cancer research and Treatment 177 (2019): 41-52.
[5] Bychkov D, Linder N, Turkki R et al (2018) Deep learning based tissue analysis predicts outcome in
colorectal cancer. Sci Rep 8:3395. https://doi.org/10.1038/s41598-018-21758-3.
[6] Mobadersany P, Yousef S, Amgad M et al (2018) Predicting cancer outcomes from histology and
genomics using convolutional networks. Proc Natl Acad Sci USA 115:E2970–E2979. https://doi.
org/10.1073/pnas.1717139115.
[7] Sun C, Shrivastava A, Singh S, Gupta A (2017) Revisiting unreasonable efectiveness of data in deep
learning era. In: 2017 IEEE International conference on computer vision (ICCV), Venice, Italy, 22–
29 October 2017, pp 843–852.

32
[8] Khan, Mohammad Monirujjaman, et al. "Breast Tumor Detection Using Robust and Efficient
Machine Learning and Convolutional Neural Network Approaches." Computational Intelligence
and Neuroscience 2022 (2022).
[9] A. H. Osman and H. M. A. Aljahdali, “An effective of ensemble boosting learning method for breast
cancer virtual screening using neural network model,” IEEE Access, vol. 8, pp. 39165–39174, 2020.
[10] K. M. R. Alam, N. Siddique, and H. Adeli, “A dynamic ensemble learning algorithm for neural
networks,” Neural Computing & Applications, vol. 32, no. 12, pp. 8675–8690, 2019.
[11] M. Abdar and V. Makarenkov, “CWV-BANN-SVM ensemble learning classifier for an accurate
diagnosis of breast cancer,” Measurement, vol. 146, pp. 557–570, 2019.
[12] S. Lee, M Amgad, M. Masoud, R. Subramanian, D. Gutman, and L. Cooper, “An ensemble-based
active learning for breast cancer classification,” in Proceedings of the 2019 IEEE International
Conference on Bioinformatics and Biomedicine, pp. 2549–2553, BIBM), San Diego, CA, USA,
November 2019.
[13] J.-B. Lamy, B. Sekar, G Guezennec, J Bouaud, and B Seroussi, ´ “Explainable artificial intelligence
for breast cancer: a visual case-based reasoning approach,” Artificial Intelligence in Medicine, vol.
94, pp. 42–53, 2019.
[14] Guo, Yang, Xuequn Shang, and Zhanhuai Li. "Identification of cancer subtypes by integrating
multiple types of transcriptomics data with deep learning in breast cancer." Neurocomputing 324
(2019): 20-30
[15] Golden, Jeffrey Alan. "Deep learning algorithms for detection of lymph node metastases from breast
cancer: helping artificial intelligence be seen." Jama 318.22 (2017): 2184-2186.
[16] Li, Lingqiao, et al. "Multi-task deep learning for fine-grained classification and grading in breast
cancer histopathological images." Multimedia Tools and Applications 79 (2020): 14509-14528.
[17] Zhu, Zhe, et al. "Deep learning for identifying radiogenic associations in breast cancer." Computers
in Biology and Medicine 109 (2019): 85-90.
[18] Bejnordi, Babak Ehteshami, et al. "Diagnostic assessment of deep learning algorithms for detection
of lymph node metastases in women with breast cancer." Jama 318.22 (2017): 2199-2210.
[19] Litjens G, Sánchez CI, Timofeeva N, et al. Deep learning as a tool for increased accuracy and
efficiency of histopathological diagnosis. Sci Rep. 2016;6:26286.
[20] Pendas S, Dauway E, Cox CE, et al. Sentinel node biopsy and cytokeratin staining for the accurate
staging of 478 breast cancer patients. Am Surg. 1999;65(6):500-505.
[21] Bi, Wenya Linda, et al. "Artificial intelligence in cancer imaging: clinical challenges and
applications." CA: a cancer journal for Clinicians 69.2 (2019): 127-157.
[22] Lamy, Jean-Baptiste, et al. "Explainable artificial intelligence for breast cancer: A visual case-based
reasoning approach." Artificial Intelligence in Medicine 94 (2019): 42-53.
[23] Shen, Li, et al. "Deep learning to improve breast cancer detection on screening
mammography." Scientific Reports 9.1 (2019): 12495.
[24] Patra, Ankita, et al. "Two-layer deep feature fusion for detection of breast cancer using
thermography images." Onkologia i Radiotherapy 16.8 (2022): 6-8.
[25] https://www.kaggle.com/datasets/uciml/breast-cancer-wisconsin-data

33
AUTHORS
Mrs. Ankita Patra (PhD Scholar) currently doing her PhD research work at in the
Department of Electronics at Sambalpur University. She is a Sambalpur University
Research Fellowship (SURF) Scholar. She has Two years of teaching, research &
administrative experience. She has received her M. Tech degree from Veer Surendra Sai
University of Technology. Her research area is image processing, machine learning, and
deep learning. She has published 3 research papers in different reputed journals and
conferences. She is currently doing her research in the field of Cancer Detection using image processing
techniques and evaluation of various in-use detection methods.
She can be contacted at Email: ankitapatra033@gmail.com
Prabira Kumar Sethy (Senior Member IEEE) currently working as an Assistant Professor
in the Department of Electronics at Sambalpur University since 2013. He has nine years of
teaching, research & administrative experience and four years of Industry experience.
Previously he worked as Engineer in Doordarshan, Prashar Bharati, from 2009 to 2013. He
has received his Ph.D. and M. Tech degree from Sambalpur University and IIT (ISM)
Dhanbad, respectively. His research area is image processing, machine learning, and deep
learning. He has published 80 research papers in different reputed journals and conferences. In addition, he
has two patents. He is the Associate Editor of Onkologia I Raditherapy. He is an editorial board member of
the International Journal of Electrical and Computer Engineering. He is also Editorial Board Member in
Ingénierie des Systèmes d’Information. IIETA. He received the “InSc Young Achiever Award” for the
research paper “Detection of coronavirus (COVID-19) based on Deep Features and Support Vector
Machine, organized by the Institute of Scholars, Ministry of MSME, Government of India in the year 2020.
He is a Senior Member of IEEE. He is a frequent reviewer of many journals and session chair of
international conferences. He can be contacted at Email: prabirsethy.05@gmail.com
Dr. Nalini Kanta Barpanda received Ph.D. in Engineering from the Sambalpur University
in the year 2012. He is now working as Associate Professor and Head of the Department of
Electronics, Sambalpur University, Odisha. He has published over 65 number of research
articles in various areas of Performance Analysis of Communication Interconnection
Networks, Quality of Service of Optical Interconnection Networks, Image Processing
Applications in Agriculture and IoT applications in Sensor Calibration System for Improving
Information Quality. She can be contacted at email: nkbarpanda@suniv.ac.in
Santi Kumari Behera (Senior Member IEEE) currently working as an Assistant Professor
in the Department of Computer Science and Engineering at Veer Surendra Sai University
of Technology since 2015. Previously She has working as Assistant Professor in MNNIT
Prayagraj for 4 years. She has 10 years of teaching, research & administrative experience.
She has received his Ph.D. and M. Tech degree from VSSUT Burla and NIT Rourkela,
respectively. Her research area is image processing, machine learning, and deep learning.
She has published 70 research papers in different reputed journals and conferences. In
addition, he has one patent and one copyright. She is guiding 4 Ph.D. scholar and guided more than 20 P.G.
Scholar. She can be contacted at email: b.santibehera@gmail.com
Ipsa Mahapatra has received her Bachelor’s in technology degree in the field of computer
science and engineering in the year 2023. She has executed her knowledge in Machine
learning and Deep learning through different internships and projects. She has worked
closely as a project intern at National Institute of Technology, Meghalaya on Image
forgery detection. She is keen to learn, understand and execute her knowledge in the best
way possible with her meticulous and high problem solving potential. She can be contacted
at email: ipsamahapatra06@gmail.com

Breast Tumor Detection Using Efficient Machine Learning and Deep Learning Techniques

More Related Content

Similar to Breast Tumor Detection Using Efficient Machine Learning and Deep Learning Techniques (20)

Recently uploaded (20)

Breast Tumor Detection Using Efficient Machine Learning and Deep Learning Techniques