Breast Cancer Detection and classification using ML and Histopathological image Processing

Abstract

Introduction

I. Introduction

Breast cancer is the second most common cancer in women worldwide, and early detection is crucial for effective treatment and improved survival. This project introduces an intelligent diagnostic system that combines image pre-processing, deep learning, and ensemble learning to detect breast cancer more accurately and efficiently.

Project Objectives:

• Detect breast cancer using real-world histopathological datasets.

• Automate the diagnostic process through image analysis and machine learning (ML).

• Evaluate model performance using accuracy, sensitivity, and specificity.

• Visualize tumor regions for better clinical interpretation.

• Support early-stage diagnosis and clinical decision-making.

II. Related Work

A. Traditional Methods

• Includes mammography, biopsies, and ultrasound.

• These are often invasive, time-consuming, and depend heavily on human interpretation.

• Can produce false positives and diagnostic delays.

B. ML-Based Approaches

• Algorithms like SVM, Random Forest, Decision Trees, and k-NN have been used for tumor classification.

• Effective but often require manual feature engineering, and may perform poorly in noisy or unstructured real-world scenarios.

C. Deep Learning Advances

• CNNs have revolutionized medical image analysis by automating feature extraction.

• Models like VGG16, ResNet, and InceptionNet show high accuracy in tumor detection.

• Challenges include interpretability and high computational cost.

D. Ensemble Learning

• Combines outputs from multiple models to improve robustness and accuracy.

• Methods like bagging, boosting, and stacking help manage class imbalance and noisy data.

E. Key Challenges in Existing Research

• Class imbalance in datasets,

• Lack of model interpretability for clinicians,

• Deployment issues in low-resource settings.

F. Contribution of This Work

• A hybrid ensemble learning system combining SVM, Decision Tree, Random Forest, and ANN.

• Uses a voting mechanism for better diagnostic precision.

• Provides visual aids to help radiologists identify potentially malignant regions.

• Designed for clinical and remote use with computational efficiency.

III. Dataset Overview

A. Source

• Dataset obtained from Kaggle, containing histopathological breast tissue images.

• Labeled into benign and malignant categories.

B. Composition

• Malignant images: 7,890

• Benign images: 5,925

• Each image shows microscopic breast cell nuclei for tumor classification.

C. Preprocessing & Augmentation

• Data Augmentation: Flipping, rotation, brightness adjustments, etc.

• Normalization and resizing (64×64 pixels) for CNN compatibility.

• Class balancing techniques to ensure even model learning.

D. Dataset Challenges

• Minor class imbalance addressed with oversampling.

• Variability in image quality due to magnification and staining.

• Need to reduce false positives in benign cases while maintaining sensitivity.

IV. Methodology

A. System Pipeline

1. Dataset Loading: Import real-world medical imaging data.

2. Image Pre-processing: Enhance image quality, reduce noise, extract features.

3. Model Training: Use multiple ML classifiers (SVM, ANN, Decision Tree, Random Forest).

4. Ensemble Classification: Voting-based system to combine classifier outputs.

5. Prediction & Visualization: Display results and highlight suspected cancerous regions.

B. Deep Learning with CNN

• CNNs are trained to distinguish benign vs malignant tumors using image features.

• Automatic feature extraction removes the need for manual preprocessing.

• Aids clinicians by accelerating diagnosis and improving reliability.

Conclusion

1) The Breast Cancer Detection System developed in this research utilizes advanced image pre-processing techniques and deep learning models to enable accurate and efficient diagnosis of breast cancer from histopathological images. By combining Convolutional Neural Networks (CNNs) with an ensemble of machine learning classifiers, the system effectively classifies tumors as benign or malignant, providing immediate and interpretable diagnostic results to support early intervention. 2) The system demonstrates strong performance across key evaluation metrics such as accuracy, precision, recall, and F1-score, confirming its reliability in real-world clinical settings. The integration of data augmentation and class balancing techniques further enhances its robustness, allowing the model to generalize well across varying image qualities and cellular structures. 3) Despite its high performance, the system faces challenges including class imbalance, limited dataset diversity, and the interpretability of deep learning outputs. Future improvements could include integration with explainable AI tools, multi-modal imaging fusion, and deployment on portable edge devices for use in resource-constrained healthcare environments. 4) This work contributes to the growing field of AI-assisted medical diagnostics by offering a scalable and adaptive solution for breast cancer detection. It holds significant potential to aid radiologists and pathologists in making faster, more informed decisions, ultimately improving patient outcomes through timely diagnosis and treatment.

References

[1] Devika Menon, M. K., & Rodrigues, J. (2023). Efficient Ultra Wideband Radar Based Non-Invasive Early Breast Cancer Detection. IEEE Access. https://doi.org/10.1109/ACCESS.2023.3303333 [2] Kathale, P., & Thorat, S. (2020). Breast Cancer Detection and Classification. In 2020 International Conference on Emerging Trends in Information Technology and Engineering (ic-ETITE). IEEE. https://doi.org/10.1109/ic-ETITE47296.2020.9077814 [3] Priyanka, & Sanjeev, K. (2021). A Review Paper on Breast Cancer Detection Using Deep Learning. IOP Conf. Series: Materials Science and Engineering, 1022(1), 012071. https://doi.org/10.1088/1757-899X/1022/1/012071 [4] Adam, R., Dell’Aquila, K., Hodges, L., Maldjian, T., & Duong, T. Q. (2023). Deep Learning Applications to Breast Cancer Detection by Magnetic Resonance Imaging: A Literature Review. Breast Cancer Research, 25(87). https://doi.org/10.1186/s13058-023-01687-4 [5] Bou Nassif, A., Talib, M. A., Nasir, Q., Afadar, Y., Elgendy, O. (2022). Breast Cancer Detection Using Artificial Intelligence Techniques: A Systematic Literature Review. Artificial Intelligence in Medicine, 127, 102139. https://doi.org/10.1016/j.artmed.2022.102139 [6] Carriero, A., & Groenhoff, L. (2024). Deep Learning in Breast Cancer Imaging: State of the Art and Recent Advancements. Diagnostics, 14(8), 848. https://doi.org/10.3390/diagnostics14080848 [7] Islam, T., & Sheakh, M. A. (2024). Predictive Modeling for Breast Cancer Classification Using Machine Learning and Explainable AI. Scientific Reports. https://doi.org/10.1038/s41598-024-57740-5 [8] Redmon, J., & Farhadi, A. (2018). YOLOv3: An Incremental Improvement. arXiv preprint arXiv:1804.02767. [9] Howard, A. G., et al. (2017). MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications. arXiv preprint arXiv:1704.04861. [10] Ribeiro, M., & Paiva, A. C. (2021). Deep Learning-Based Firearm Detection for Public Safety Applications. Sensors, 21(4), 1345.

Copyright

Copyright © 2025 Rohan Berad, Jyoti Mali, Aniket Kadam, Sagar Pandule , Shrikrishna Nimbekar. 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.