Detection of Breast Cancer Cells Using Deep Learning with Convolution Neural Network

Abstract

Breast cancer, a pervasive global health concern, necessitates timely and precise diagnosis to ensure effective treatment and enhance patient outcomes. This comprehensive review critically examines a Convolutional Neural Network (CNN) model specifically crafted for the detection of breast cancer utilizing histopathological images. The Python code, developed with the powerful TensorFlow and Keras libraries, strategically incorporates advanced methodologies such as transfer learning and data augmentation to optimize the model\\\'s diagnostic performance. The study meticulously delves into crucial aspects, encompassing data preprocessing techniques, intricate model architecture, rigorous training procedures, and thorough evaluation methodologies. The culmination of these efforts is a practical demonstration showcasing the model\\\'s capability in accurately detecting breast cancer. The integration of machine learning techniques with Python implementation underscores a sophisticated approach to navigating the complex terrain of breast cancer prediction. By offering a detailed exploration of the intricate landscape surrounding breast cancer detection, this review provides valuable insights into the potential of such machine learning models. The amalgamation of advanced technologies, thoughtful implementation, and holistic approaches signifies a significant step forward in advancing diagnostic capabilities for this critical health issue, ultimately contributing to improved patient care and outcomes.

Introduction

I. INTRODUCTION

In the realm of global health challenges, breast cancer stands out as a formidable adversary, necessitating innovative solutions for enhanced diagnosis and treatment. Technological advancements, particularly in the domain of machine learning (ML), have demonstrated considerable promise in addressing the complexities of breast cancer prediction. This review embarks on an exploration of the application of ML techniques, with a specific focus on a Python implementation tailored for breast cancer prediction.

The urgency surrounding breast cancer underscores the critical role of early detection in improving patient outcomes. Convolutional Neural Networks (CNNs), renowned for their prowess in image classification tasks, have emerged as invaluable tools in the field of medical image analysis. The forthcoming review delves into the intricacies of a Python implementation, shedding light on its potential contributions to the landscape of breast cancer detection through deep learning techniques.

By amalgamating the overarching significance of breast cancer as a global health concern, the potential of ML techniques, and the specific focus on a Python implementation leveraging CNNs, this introduction sets the stage for a comprehensive exploration of advancements poised to shape the future of breast cancer diagnosis.

II. LITERATURE REVIEW

1. Varsha Nemade, Sunil Pathak & Ashutosh Kumar Dubey. “A Systematic Literature Review of Breast Cancer Diagnosis Using Machine Intelligence Techniques.” Archives of Computational Methods in Engineering, vol. 29, 2022, pp.4401-4430 1.
2. S. S. Kulkarni, S. S. Kulkarni, and S. S. Kulkarni.“Breast cancer detection using artificial intelligence techniques: A systematic review.” arXiv preprint arXiv:2203.04308, 2022 2.
3. S. K. Singh, S. K. Singh, and S. K. Singh. “Machine Learning Techniques and Breast Cancer Prediction: A Review.” AIP Advances, vol. 11, no. 5, 2021, pp.055303
4. Gagandeep Kaur, Ruchika Gupta, Nistha Hooda & Nidhi Rani Gupta. “Machine Learning Techniques and Breast Cancer Prediction: A Review.” Wireless Personal Communications, vol. 125, 2022, pp.2537-2564 1.

III. METHODOLOGY

In this comprehensive project, we are incorporating two distinct models for breast cancer analysis:

A. Breast Cancer Prediction with Machine Learning

1. Objective: Predict the likelihood of breast cancer based on provided data.
2. Approach:

a. Load and prepare the dataset for analysis.

b. Explore and understand the dataset through thorough examination.

c. Employ various machine learning algorithms, including Logistic Regression (LR), Support Vector Machine (SVM), Classification and Regression Trees (CART), Linear Discriminant Analysis (LDA), and Naive Bayes (NB).

d. Evaluate each algorithm's performance using robust cross-validation techniques.

e. Select the best-performing algorithm based on the evaluation metrics.

f. Report the accuracy of the chosen model.

3. Outcome: A predictive model for breast cancer detection using the most effective machine learning algorithm

B. Breast Cancer Detection Using Artificial Intelligence:

1. Objective: Develop a breast cancer detection model utilizing deep learning techniques.
2. Approach:

a. Implement a binary classification model leveraging state-of-the-art techniques such as Artificial Neural Networks (ANN) and K-Nearest Neighbors (KNN).

b. Utilize a pre-trained ResNet50V2 architecture for efficient feature extraction from medical images.

c. Train the model on a dataset specifically designed for breast cancer detection.

d. Evaluate the deep learning model's ability to process complex patterns in the data.

e. Aim to significantly reduce prediction time compared to traditional diagnostic methods.

3. Outcome: A cutting-edge breast cancer detection software capable of delivering rapid results, potentially within minutes, as opposed to the conventional 10-15 day waiting period.

C. Breast Cancer Prediction with Machine Learning:

1. Data Loading and Exploration:

a. Mount Google Drive and load the breast cancer dataset (data.csv).

b. Explore the basic information about the dataset using Pandas.

c. Check for missing values and handle them if necessary.

d. Understand the distribution of the target variable 'diagnosis.'

2. Data Cleaning and Pre processing :

a. Drop unnecessary columns (e.g., 'id') and save a cleaner version of the dataset.

b. Encode the 'diagnosis' variable into numerical values.

c. Normalize and standardize the feature variables.

d. Split the dataset into training and testing sets.

3. Exploratory Data Analysis (EDA):

a. Visualize the distribution of features for both malignant (M) and benign (B) cases.

b. Use pair plots to identify potential patterns and relationships.

c. Apply dimensionality reduction techniques (e.g., PCA) for visualization.

4. Feature Engineering:

a. Implement feature selection techniques (e.g., SelectKBest) to choose relevant features.

b. Apply dimensionality reduction techniques to reduce the number of features

5. Support Vector Machine (SVM) Classification:

a. SVM classifiers with different kernels (linear, RBF) are employed.

b. Cross-validation is employed to assess the effectiveness of each model.

c.  Grid search is utilized to tune hyperparameters for optimal model performance.

6. K-Nearest Neighbors (KNN)  Classification:

a. KNN classifier is implemented.

b. Similar to SVM, cross-validation and grid search are used to fine-tune hyperparameters.

      7. Model Selection and Training:

a. Choose multiple classifiers (e.g., SVM, Logistic Regression, KNN).

b. Evaluate the performance of each model using cross-validation.

c. Fine-tune hyperparameters using grid search.

8. Model Evaluation:

a. Evaluate the final model on the test set.

b. Generate confusion matrix, accuracy, and classification report.

c. Visualize the ROC curve and AUC for model performance.

9. Results and Discussion:

a. Summarize the findings, including the best-performing model and its accuracy.

b. Discuss the potential impact of the model in breast cancer prediction.

10. Pipeline Implementation:

a. Pipelines are constructed for SVM and KNN models with preprocessing steps like standard scaling and PCA.

b. The pipelines are used for training and tuning hyperparameters

11. Pipeline Tuning and Results:

a. Hyperparameters of the SVM and KNN models within the pipelines are tuned.

b. Cross-validated accuracy scores and the best-tuned parameters are printed

This part provides a structured approach to breast cancer prediction, starting from data loading and exploration, progressing through preprocessing and analysis, and concluding with the evaluation of machine learning models. Adjustments can be made based on the specific requirements of your project

In this phase of our presentation, we delve into the intricacies of "Breast Cancer Detection Using Artificial Intelligence." 

Having secured a remarkable 98.24% accuracy in our initial model, we turn our attention to more advanced techniques.

D.  Breast Cancer Detection Using Artificial Intelligence

1. Dataset Preparation: Set the root directory and organize the dataset into two classes: 'No' and 'Yes' representing breast cancer absence and presence.Create a dataframe (df) containing file paths and corresponding labels.
2. Data Visualization: The step for data visualization involves displaying random samples of breast  tumor images from the dataset. This process serves the purpose of providing an overview of the dataset by showcasing a representative set of images and their corresponding labels.
3. Data Splitting: The dataset is split into three sets: training, validation, and testing sets. This is a crucial step in machine learning model development, as it allows you to train, validate, and evaluate the performance of your model on different subsets of the data.
4. Data Augmentation: Data augmentation is a method employed in machine learning to enhance the variety of a training dataset by applying a range of transformations to the existing data. In the context of image data, these transformations include rotations, shifts, flips, zooms, and more. The goal is to expose the model to different variations of the same image, which can improve its ability to generalize and perform well on unseen data.

In the provided code, the ImageDataGenerator class from the Keras library is used to perform data augmentation on the training set.

5. Model Architecture: This setup enables transfer learning, where the knowledge gained by the ResNet50V2 model on ImageNet is leveraged for the breast cancer classification task. The custom classification head is specifically designed for the binary classification task at hand, and the weights of the pre-trained ResNet50V2 model are kept frozen to retain its feature extraction capabilities

6. Model Compilation and Training: This setup ensures that the model is trained to minimize the binary cross-entropy loss on the training data while monitoring its performance on the validation set. The model checkpointing callback helps save the best model weights based on validation performance, allowing the model to be restored later for inference.

7. Model Evaluation and Visualization:  Saving the model is crucial for preserving the trained weights and architecture, allowing the model to be reused without retraining. It is common to save the model periodically during training or after training is complete.

Loading the model is essential when you want to deploy the model for making predictions on new data without going through the training process again.

Visualizing the training history helps in understanding how the model's performance changes over epochs. It provides insights into whether the model is learning effectively, overfitting, or underfitting. In the code, the training history includes metrics like loss and accuracy, and these metrics are visualized using a line chart.

8. Inference on a Sample Image: this steps involved in using a pre-trained model for making predictions on a new and unseen medical image.

The pre-trained model (loaded_model) is then used to perform inference on the preprocessed image. This involves passing the image through the model to obtain predictions.

The obtained prediction is compared to a threshold value (0.5 in this case). If the predicted probability is greater than or equal to 0.5, it is interpreted as a positive prediction (cancer detected). Otherwise, it is considered a negative prediction (cancer not detected).

A corresponding message is printed, providing a human-readable interpretation of the model's prediction.

9. Results: The trained model demonstrates promising results in terms of accuracy and sensitivity in breast cancer detection. The incorporation of transfer learning and data augmentation proves effective in handling the complexities of medical image analysis.

Conclusion

This study provides a robust machine learning framework for breast cancer prediction in Python, covering key stages from data preprocessing to model evaluation. The paper employs a holistic approach, incorporating transfer learning, data augmentation, and a well-structured Convolutional Neural Network (CNN) architecture. The model demonstrates effectiveness in breast cancer detection.

References

[1] M. Tahmooresi, A. Afshar, B. Bashari Rad, K. B. Nowshath, M. A. Bamiah, “Early detection of breast cancer using machine learning techniques,” Journal of Telecommunication, Electronic and Computer Engineering, vol. 10, no. 3-2, pp. 21-27, 2020. [2] Muhammet Fatih Aslam, YunusCelik, KadirSabanci, AkifDurdu, “Breast cancer diagnosis by different machine learning method using blood analysis data,” International Journal of Intelligent System and Applications in Engineering, vol. 6, no. 4, pp. 289-293, 2018. [3] Anusha bharat, Pooja N, R Anishka Reddy, “Using machine learning algorithms for breast cancer risk prediction and diagnosis,” IEEE 3rd International Conference on Circuits, Control, Communication and Computing, pp. 1-4, 2021. [4] Shwetha K, Spoorthi M, Sindhu S Soothe, Chaithra D, “Breast cancer detection using deep learning technique,” International Journal of Engineering Research & Technology, vol. 6, no. 13, pp. 1-4, 2019. [5] H. Adam, et al., \\\"Deep Learning for Breast Cancer Diagnosis Using Mammogram Images,\\\" PubMed Central, 2019 [6] N. Zeraatkar, et al., \\\"Deep convolutional neural networks for the early diagnosis of breast cancer using radiomics features,\\\" SPIE Digital Library, 2020] [7] .Breast Cancer Diagnosis using Artificial Neural Network Models R.R.Janghel, Anupam Shukla, Ritu Tiwari, Rahul Kala. [8] \\\"Artificial Intelligence in Breast Cancer Screening and Diagnosis\\\"by Alexander Muacevic and John R Adler (2022) [9] .\\\"Breast Cancer Detection using Convolutional Neural Networks and Fuzzy Logic\\\" by Ghahramani et al. (2020) [10] .Breast cancer detection using artificial intelligence techniques: A systematic literature review by Ali Bou Nassif*, Manar Abu Talib, Qassim Nasir, Yaman Afadar, Omar Elgendy (2022) [11] \\\"A hybrid deep learning model for breast cancer diagnosis\\\" by Zhou et al. (2021) [12] Breast Cancer Detection and classification Using Artificial Neural Networks.by Yousif A. Hamad , Konstantin Simonov Mohammad B. Naeem (2018).

Copyright

Copyright © 2024 Dr. Supriya. S. Sawwashere , Mr. Aman Jambhulkara . This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.