Automated Label-Free Classification of Circulating Tumor Cells and White Blood Cells Using Hyperspectral Imaging and
Shun-Chi Wu1, Jon-Nan Chiu1, Yi-Wen Chen1
1Department of Engineering and System Science, National Tsing Hua University, Hsinchu 30013, Taiwan.
Micromachines
|May 4, 2026
Summary
This study introduces a novel label-free method for detecting circulating tumor cells (CTCs) using microfluidics and deep learning. The advanced system achieves high accuracy for cancer biomarker detection in liquid biopsies.
Area of Science:
- Biomedical Engineering
- Cancer Research
- Microfluidics
Background:
- Circulating tumor cells (CTCs) are crucial cancer biomarkers.
- Their rarity and heterogeneity challenge label-free detection methods.
Purpose of the Study:
- To develop an automated, non-invasive framework for CTC detection.
- To integrate microfluidics, hyperspectral imaging, and deep learning for enhanced classification.
Main Methods:
- A self-assembly cell array (SACA) microfluidic chip organized cells into a monolayer.
- Two deep learning pipelines (A2S2K-ResNet and ResNet50) were employed.
- A multi-band ensemble strategy with majority voting was developed to overcome spectral overlap.
Main Results:
- Initial spectral classification showed ~80% accuracy on cell lines.
- The hybrid approach achieved >93.5% accuracy and >92% precision for patient-derived CTCs.
- The method demonstrated robustness against spectral overlap with white blood cells.
Conclusions:
- The integrated microfluidic and multi-band deep learning approach provides reliable, label-free CTC detection.
- This pipeline supports clinical liquid biopsy and real-time cancer monitoring.
- The study highlights the potential of combining spatial control with advanced computational analysis for biomarker discovery.
