Related Experiment Video
Updated: Nov 3, 2025
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
43.2K
Computer Aided Breast Cancer Detection Using Ensembling of Texture and Statistical Image Features
Soumya Deep Roy1, Soham Das1, Devroop Kar2
1Department of Metallurgical and Material Engineering, Jadavpur University, Kolkata 700032, India.
Sensors (Basel, Switzerland)
|June 2, 2021
Summary
Machine learning accurately classifies Invasive Ductal Carcinoma (IDC) breast cancer from histopathology images. This approach aids early detection, improving patient survival rates and offering results comparable to deep learning models.
Area of Science:
- Oncology
- Medical Imaging
- Computational Pathology
Background:
- Breast cancer is a leading cause of cancer-related deaths globally, with early detection crucial for improved survival.
- Histopathological slide analysis is standard for diagnosing breast cancer subtypes like Invasive Ductal Carcinoma (IDC) and Ductal Carcinoma In Situ (DCIS).
- Distinguishing IDC from non-IDC on histopathology slides is challenging and time-consuming for pathologists.
Purpose of the Study:
- To develop and evaluate a machine learning model for accurate classification of Invasive Ductal Carcinoma (IDC) from breast cancer histopathology images.
- To investigate the efficacy of combining textural and statistical features for IDC detection.
- To compare the performance of various machine learning classifiers in identifying IDC.
Main Methods:
- Utilized a large dataset of 277,524 breast cancer histopathology images (78,786 IDC positive, 198,738 IDC negative).
- Extracted features using SIFT, SURF, ORB, and Haralick texture analysis, resulting in 782 features.
- Ensembled features via stacking with classifiers (Random Forest, Extra Trees, XGBoost, AdaBoost, CatBoost, MLP) and applied feature selection (Pearson Correlation Coefficient) to reduce features to four for final classification.
Main Results:
- The CatBoost classifier achieved the highest accuracy of 92.55% in classifying IDC positive and negative images.
- The developed machine learning approach demonstrated performance comparable to state-of-the-art deep learning methods.
- Feature engineering and ensemble methods proved effective for IDC classification.
Conclusions:
- Machine learning, particularly using the CatBoost algorithm with selected features, offers a highly accurate and efficient method for diagnosing Invasive Ductal Carcinoma (IDC).
- This computational approach can significantly aid pathologists in breast cancer diagnosis, potentially leading to earlier detection and better patient outcomes.
- The study highlights the potential of integrating traditional feature extraction techniques with advanced machine learning for medical image analysis.
