Cell-MICS: Detecting Immune Cells With Label-Free Two-Photon Autofluorescence and Deep Learning
Clinical Snapshot
PICO Framework
| P — Population | Isolated immune cells (T cells, neutrophils, and four other cell types) imaged ex vivo |
| I — Intervention | Two-photon autofluorescence imaging with convolutional neural network (SqueezeNet) classification |
| C — Comparator | Conventional antibody-based cell identification markers |
| O — Outcomes | Diagnostic accuracy metrics: ROC-AUC, PR-AUC, F1 score, precision, and recall for immune cell classification |
Bottom Line
This proof-of-concept study demonstrates that machine learning can classify immune cells using label-free two-photon autofluorescence imaging, achieving reasonable accuracy for binary classification (ROC-AUC 0.89) but modest performance for multi-class scenarios (F1 score 0.689). However, the study's validity is significantly limited by its ex vivo design using isolated cells, which may not reflect the complexity of in vivo tissue environments. The lack of traditional diagnostic accuracy measures, confidence intervals, and insufficient methodological detail further limit clinical interpretation. While the approach shows promise for future label-free tissue imaging, substantial validation in clinical specimens and comparison with established diagnostic methods would be required before considering clinical implementation. Australian clinicians should view this as early-stage research requiring significant further development rather than a clinically ready diagnostic tool.
Key Findings
P Value: Not reported
Effect Size: ROC-AUC 0.89, PR-AUC 0.95 for binary classification
Primary Outcome: Binary classification accuracy between T cells and neutrophils
Nnt Or Sensitivity: Multi-class: F1 score 0.689, precision 0.697, recall 0.748
Confidence Interval: Not reported
Clinical Application
Requires specialized two-photon microscopy equipment and computational resources for CNN implementation Technology not yet available through PBS; would require TGA approval for clinical use; potential future application in specialized pathology services Limited applicability - study conducted on isolated cells rather than clinical specimens
Abstract
Multiphoton imaging has been widely used for deep-tissue imaging. Although its label-free, metabolic contrast is ideal for investigating inflammation, the label-free two-photon induced autofluorescence is often regarded as less specific compared to conventional antibody markers. In this work, we investigate the potential for multiphoton imaging with computational specificity (MICS) by training a convolutional neural network on images of different immune cells. A low-complexity squeezeNet architecture was able to achieve reliable immune cell classification results (0.89 ROC-AUC, 0.95 PR-AUC for binary classification between T cells and neutrophils; 0.689 F1 score, 0.697 precision, 0.748 recall for multi-class classification between six isolated cell types). Perturbation tests confirmed that the model was not confused by the extracellular environment and that 2P-AF from NADH and FAD is equally important for the classification. In the future, deep learning could provide computational specificity for specific immune cells in unstained tissues, with great potential for label-free in vivo endomicroscopy.
References
- 1.Kreiss, L., Chaware, A., Roohian, M., Lemire, S., Thoma, O. M., Carlé, B., Waldner, M., Schürmann, S., Friedrich, O., & Horstmeyer, R. (2026). Cell-MICS: Detecting immune cells with label-free two-photon autofluorescence and deep learning. Journal of Biophotonics. https://doi.org/10.1109/ACCESS.2025.3547711
Related Research
The American journal of pathology
An Integrated Machine Learning and Genomic Framework for Precise Detection of Gastric Cancer
3 Aug 2026
Clinical & translational oncology : official publication of the Federation of Spanish Oncology Societies and of the National Cancer Institute of Mexico
Research progress of machine learning applications in gastric cancer diagnosis and therapy
2 Aug 2026
Scientific reports
Safety design guidelines for clinician-AI interaction in computer-aided diagnosis systems using system-theoretic framework with explainability validation
31 July 2026
This content is for educational purposes for healthcare professionals only and does not constitute clinical advice. Clinical decisions should be based on individual patient assessment, current guidelines, and appropriate specialist consultation. Editorial Standards · Privacy Policy · Terms of Service