Information‐Distilled Generative Label‐Free Morphological Profiling Encodes Cellular Heterogeneity

Advanced Science, Journal Year: 2024, Volume and Issue: 11(29)

Published: June 12, 2024

Abstract Image‐based cytometry faces challenges due to technical variations arising from different experimental batches and conditions, such as differences in instrument configurations or image acquisition protocols, impeding genuine biological interpretation of cell morphology. Existing solutions, often necessitating extensive pre‐existing data knowledge control samples across batches, have proved limited, especially with complex data. To overcome this, “Cyto‐Morphology Adversarial Distillation” (CytoMAD), a self‐supervised multi‐task learning strategy that distills biologically relevant cellular morphological information batch variations, is introduced enable integrated analysis multiple without assumptions manual annotation. Unique CytoMAD its “morphology distillation”, symbiotically paired deep‐learning image‐contrast translation—offering additional interpretable insights into label‐free The versatile efficacy demonstrated augmenting the power biophysical imaging cytometry. It allows classification human lung cancer types accurately recapitulates their progressive drug responses, even when trained concentration information. also joint tumor heterogeneity, linked epithelial‐mesenchymal plasticity, standard fluorescence markers overlook. can substantiate wide adoption for cost‐effective diagnosis screening.

Language: Английский

Single‐detector multiplex imaging flow cytometry for cancer cell classification with deep learning

Cytometry Part A, Journal Year: 2024, Volume and Issue: 105(9), P. 666 - 676

Published: Aug. 5, 2024

Abstract Imaging flow cytometry, which combines the advantages of cytometry and microscopy, has emerged as a powerful tool for cell analysis in various biomedical fields such cancer detection. In this study, we develop multiplex imaging (mIFC) by employing spatial wavelength division multiplexing technique. Our mIFC can simultaneously obtain brightfield multi‐color fluorescence images individual cells flow, are excited metal halide lamp measured single detector. Statistical results experiments with resolution test lens, magnification fluorescent microspheres validate operation good channel consistency micron‐scale differentiation capabilities. A deep learning method is designed image processing that consists three networks (U‐net, very super resolution, visual geometry group 19). It demonstrated cluster 24 (CD24) more sensitive than brightfield, nucleus, or antigen 125 (CA125) classifying types ovarian lines (IOSE80 normal cell, A2780, OVCAR3 cells). An average accuracy rate 97.1% achieved classification these when all four channels considered. single‐detector promising development future cytometers automatic single‐cell fields.

Language: Английский