Machine learning-enabled cancer diagnostics with widefield polarimetric second-harmonic generation microscopy.

Area of Science:

• Biomedical Optics
• Cancer Research
• Materials Science

Background:

• Extracellular matrix (ECM) collagen remodeling is critical in tumorigenesis.
• ECM alterations are underutilized in cancer diagnostics due to limitations of standard imaging.
• Standard hematoxylin and eosin (H&E) staining lacks sensitivity to collagen ultrastructural changes.

Purpose of the Study:

• To demonstrate whole slide imaging of breast tissue using high-throughput widefield polarimetric second-harmonic generation (P-SHG) microscopy.
• To characterize tissue collagen ultrastructure using multiparameter texture analysis of P-SHG images.
• To evaluate the efficacy of P-SHG parameters in differentiating tumor from normal breast tissue and predicting tumor status.

Main Methods:

• Whole slide imaging of breast tissue microarrays utilizing high-throughput widefield P-SHG microscopy.
• Label-free visualization and ultrastructural investigation of non-centrosymmetric molecules (collagen).
• Texture analysis of P-SHG parameters for multiparameter characterization of tissue collagen.

Main Results:

• P-SHG parameters achieved 94.2% accuracy and F1-score in differentiating tumor from normal tissue, with a 6.3% false discovery rate.
• The trained classifier predicted tumor tissue with 91.3% accuracy and 90.7% F1-score, and a 13.8% false omission rate.
• Widefield P-SHG microscopy effectively revealed collagen ultrastructure across large tissue regions.

Conclusions:

• Widefield P-SHG microscopy provides label-free, ultrastructural insights into tissue collagen.
• P-SHG parameters serve as sensitive biomarkers for cancer diagnostics and prognostics.
• This technique enhances the potential for integrating ECM analysis into routine cancer assessment.