Differentiating single cervical cells by mitochondrial fluorescence imaging and deep learning-based label-free light...

Area of Science:

• Biomedical Engineering
• Cell Biology
• Medical Imaging

Background:

• Cervical cancer poses a significant global health risk to women.
• Accurate and early detection is crucial for effective treatment and improved patient outcomes.

Purpose of the Study:

• To develop and validate a multi-modal static cytometry technique for classifying normal human cervical epithelial cells (H8) and cervical cancer cells (HeLa).
• To investigate the potential of intracellular mitochondria as biomarkers for cervical cancer diagnosis.
• To evaluate the efficacy of label-free light scattering patterns combined with deep learning for automated cervical cancer cell classification.

Main Methods:

• Utilized light-sheet static cytometry to acquire brightfield (BF) images, fluorescence (FL) images, and 2D light scattering (LS) patterns of single cervical cells.
• Employed three distinct feature extraction methods to analyze multi-modal data based on varying data characteristics.
• Applied deep learning algorithms for automatic feature extraction from label-free LS patterns.

Main Results:

• Morphological and textural feature analysis highlighted the significance of intracellular mitochondria in distinguishing between normal and cancerous cervical cells.
• The deep learning approach applied to label-free LS patterns achieved a classification accuracy of 76.16%.
• This accuracy surpassed the performance of single-mode analyses using BF and FL data.

Conclusions:

• Multi-modal static cytometry, incorporating diverse feature analysis, effectively identifies intracellular mitochondria as promising biomarkers for cervical cancer diagnosis.
• Label-free 2D light scattering combined with deep learning demonstrates significant potential for automated, non-invasive early cervical cancer classification.