Related Experiment Video
Updated: Jun 29, 2025
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
42.9K
A novel breast cancer image classification model based on multiscale texture feature analysis and dynamic learning.
Jia Guo1,2,3, Hao Yuan1,2,3, Binghua Shi4,5,6
1Hubei Key Laboratory of Digital Finance Innovation, Hubei University of Economics, Wuhan, 430205, Hubei, China.
Scientific Reports
|March 28, 2024
Summary
This study introduces an unsupervised model for breast cancer image classification using multiscale texture analysis. The novel method achieves high accuracy, offering a valuable tool for medical diagnosis.
Area of Science:
- Medical Imaging
- Machine Learning
- Computational Biology
Background:
- Traditional machine learning for medical image classification demands extensive labeled data and long training times, increasing costs and limiting practical use.
- Developing efficient and cost-effective automated diagnostic tools is crucial for reducing healthcare professional workload and improving patient outcomes.
Purpose of the Study:
- To propose a novel unsupervised breast cancer image classification model for mammograms.
- To overcome limitations of traditional methods by employing multiscale texture analysis and a dynamic learning strategy.
Main Methods:
- Images were transformed into multiscale texture feature vectors using gray-level cooccurrence matrix and Tamura coarseness.
- An unsupervised dynamic learning mechanism was utilized for the classification of these feature vectors.
Main Results:
- The proposed unsupervised model achieved high performance metrics in simulation experiments.
- Specific results include 91.500% accuracy, 92.780% precision, 91.370% F1-score, and 91.500% AUC at a 40-pixel resolution.
Conclusions:
- The developed unsupervised breast cancer classification model demonstrates significant potential for aiding medical personnel.
- The method offers an effective and efficient reference for breast cancer diagnosis, particularly from mammograms.
