Related Experiment Video
Updated: Aug 25, 2025
08:56
Terahertz Imaging and Characterization Protocol for Freshly Excised Breast Cancer Tumors
Published on: April 5, 2020
11.1K
Deep Learning Classification of Breast Cancer Tissue from Terahertz Imaging Through Wavelet Synchro-Squeezed
Haoyan Liu1, Nagma Vohra2, Keith Bailey3
1Department of Computer Science and Computer Engineering, University of Arkansas, Fayetteville, AR, 72701, USA.
Journal of Infrared, Millimeter and Terahertz Waves
|October 17, 2022
Summary
Deep learning enhances terahertz imaging for medical applications. This study improves tissue classification and cancer segmentation in fresh tissue using wavelet transforms and convolutional neural networks.
Area of Science:
- Medical imaging
- Biomedical engineering
- Artificial intelligence in medicine
Background:
- Terahertz imaging and spectroscopy offer potential for medical insights.
- Statistical methods for terahertz image classification face challenges with fresh tissue and invasive tumors.
- Deep learning shows promise in improving medical imaging tasks.
Purpose of the Study:
- To adapt deep learning methods for classifying tissue regions in terahertz images of fresh murine xenograft tissue.
- To improve the segmentation of cancerous tissue, particularly in cases of muscle invasion.
- To enhance the accuracy of terahertz imaging analysis for medical diagnosis.
Main Methods:
- Preprocessing terahertz data using wavelet synchronous-squeezed transformation (WSST) to create spectrograms.
- Utilizing deep convolution neural networks (CNNs) for pixel-wise classification of spectrograms.
- Implementing a leave-one-sample-out cross-validation strategy for robust evaluation.
Main Results:
- Achieved improved classification accuracy compared to traditional statistical methods.
- Demonstrated enhanced segmentation of cancerous tissue from surrounding muscle regions.
- Identified specific areas for future improvements in terahertz imaging and classification methodologies.
Conclusions:
- Deep learning, combined with WSST preprocessing, significantly advances terahertz medical imaging analysis.
- The proposed method offers a more accurate approach to segmenting cancerous tissues in challenging fresh tissue samples.
- Further research can refine this technique for broader clinical applications in cancer detection.
