Related Experiment Video
Updated: Jun 12, 2026
10:39
A Label-Free Segmentation Approach for Intravital Imaging of Mammary Tumor Microenvironment
Published on: May 24, 2022
Leveraging Pretrained Neural Network Models for the Classification of Tumor Cells Analyzed by Label-Free Phase
Leonor V C Losa1, Temple A Douglas1, Lia Santos1
1INL-International Iberian Nanotechnology Laboratory, Nieder Group on Quantum-,Bio- and Nanophotonics, 4719-330 Braga, Portugal.
Computational and Structural Biotechnology Journal
|June 11, 2026
Summary
This study introduces label-free phase holotomographic microscopy and neural networks for cancer cell classification. EfficientNet-B0 achieved 96.9% accuracy, demonstrating potential for personalized cancer diagnostics.
Area of Science:
- Biomedical Optics
- Computational Biology
- Cancer Research
Background:
- Accurate cancer cell classification is crucial for effective treatment.
- Label-free imaging methods reduce sample preparation complexity and cost.
- Deep learning models show promise in analyzing complex biological images.
Purpose of the Study:
- To develop and validate a label-free imaging and deep learning pipeline for cancer cell classification.
- To assess the performance of various convolutional neural networks in identifying cancer cell treatment status.
- To demonstrate the potential for classifying different grades of urothelial cancer cells.
Main Methods:
- 3D phase holotomographic microscopy was used for label-free imaging of live A549 lung cancer cells.
- Images were processed into 2D maximum-intensity projections for analysis.
- Pretrained convolutional neural networks (VGG16, ResNet18, DenseNet121, EfficientNet-B0) were evaluated for classification tasks.
Main Results:
- EfficientNet-B0 achieved 96.9% accuracy in classifying lung cancer cells based on paclitaxel treatment.
- Refractive index analysis revealed drug-induced biophysical changes in cells.
- The pipeline accurately classified high- vs. low-grade urothelial cancer cells with 90.6% accuracy.
Conclusions:
- Integrating label-free holotomographic imaging with deep learning offers a powerful approach for cancer cell classification.
- This method enables rapid, label-free detection of drug effects and tumor cell grading.
- The findings support the development of advanced, personalized diagnostic and treatment strategies.