Breast Tumor Classification: Feature Extraction and Machine Learning Performance
Open AccessThe purpose of this study was to investigate the performance of machine learning (ML) algorithms in classifying benign and cancerous breast tumors from mammograms. The mammograms are provided by the Digital Database for Screening Mammography.The significance of this study is that accurate ML algorithms could potentially reduce the need for biopsies. Though biopsies are important for breast cancer diagnosis, they can also be a high-risk invasive procedure for patients depending on their international normalized ratio (INR) and if they have skin infections [1]. Mammography is not always accurate and can lead to false negatives when examined by a radiologist, with a rate of 1 in 8 missed cases [2]. There clearly is an opportunity for improvement in cancer diagnosis. Studies are beginning to be conducted to see the efficiency and accuracy of using ML to classify breast tumors [3][4]. In this study, Histogram, Grey Level Run Length (GLRL), and Grey Level Co-Occurrence Matrix (GLCM) features were extracted from the images. These features were inputs to various ML algorithms: AdaBoost, Random Forest, Naïve Bayes, K-Nearest Neighbors, and Deep Neural Networks. Further feature analysis of the data using correlation matrices and Principal Component Analysis (PCA) was conducted to look deeper into the relationship between data points that could potentially hinder the creation of generalizable models, (other than Convolutional Neural Networks), that have proven to be successful in these types of image classification problems.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.
| Thumbnail | Title | Date Uploaded | Visibility | Actions |
|---|---|---|---|---|
|
|
Shoukeir_gwu_0075M_16426.pdf | 2023-11-14 | Open Access |
|