Rapid Hyperspectral Photothermal Mid-Infrared Spectroscopic Imaging from Sparse Data for Gynecologic Cancer Tissue...

Area of Science:

• Biomedical Optics
• Computational Pathology
• Medical Imaging

Background:

• Traditional ovarian cancer detection relies on time-consuming, qualitative, and staining-dependent histopathology.
• Mid-infrared (MIR) hyperspectral photothermal imaging offers label-free, quantitative analysis but is typically slow.

Purpose of the Study:

• To develop a novel, accelerated MIR photothermal imaging approach for enhanced ovarian cancer tissue characterization.
• To overcome the resolution-speed trade-off in MIR hyperspectral imaging.

Main Methods:

• Implemented a sparse imaging methodology for MIR photothermal imaging, enabling high-resolution image reconstruction from undersampled data.
• Utilized random forest and convolutional neural network (CNN) models for tissue classification and segmentation.
• Assessed performance using metrics like MSE, SSIM, ROC curves, and segmentation accuracy.

Main Results:

• Achieved a 10X improvement in data acquisition time for MIR hyperspectral photothermal imaging.
• Demonstrated superior image quality and accurate distinction between gynecological tissue types with >95% segmentation accuracy.
• Validated findings on 100 ovarian cancer patient samples and over 65 million data points.

Conclusions:

• The rapid MIR hyperspectral photothermal imaging method is feasible for enhanced ovarian cancer tissue characterization.
• This technology integrates label-free imaging with machine learning for quantitative, automated histopathology.
• Paves the way for faster and more accurate diagnostic tools in gynecologic oncology.