Early Breast Cancer Prediction Using Thermal Images and Hybrid Feature Extraction-Based System
The aggressive nature of breast cancer makes it the leading cause of cancer-related death in women. Infrared thermal imaging has great potential for early prediction of breast cancer, offering non-invasive, cost-effective, non-radiation exposure and applicability to be used regularly for both young...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10879496/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The aggressive nature of breast cancer makes it the leading cause of cancer-related death in women. Infrared thermal imaging has great potential for early prediction of breast cancer, offering non-invasive, cost-effective, non-radiation exposure and applicability to be used regularly for both young and old women. The infrared energy emitted by the breast surface is highly related to its physiological characteristics. This valuable information is encoded in the spatial and local intensity variations of acquired breast images, revealing hidden patterns. This article introduces a novel hybrid feature extraction approach that combines traditional and deep learning techniques to improve the prediction of breast cancer. The proposed approach makes use of two main methods for feature extraction. The first method generates texture and edge maps based on the Gabor filter bank, Canny edge detector, and holistically nested edge detector to inherit the distinctive information and patterns associated with breast functionality. Then, an image fusion process is provided to construct one composite image retaining all the map’s information from which a highly valuable feature vector is extracted using the histogram of oriented gradients (HOG). The second method extracts deep features from the breast images capturing high-level predictive information using ResNet-50 and MobileNet pre-trained convolution neural networks (CNNs). The principal component analysis (PCA) is utilized for dimensionality reduction. The back end of the methodology uses support vector machine (SVM) and extreme gradient boosting (XGB) classification algorithms to establish the relationship between the retrieved feature vector and breast functionality. The results are impressive with 95.51% and 96.22% of accuracy, 96.62% and 97.19% of sensitivity, and 94.76% and 95.23% of specificity using SVM and XGB, respectively, on the test set. These findings show the significant advantages of the proposed framework for early prediction of breast cancer using thermal images. |
|---|---|
| ISSN: | 2169-3536 |