Hyperspectral Imaging Using Flexible Endoscopy for Laryngeal Cancer Detection

Area of Science:

• Medical imaging
• Optical engineering
• Cancer diagnostics

Background:

• Hyperspectral imaging (HSI) is emerging in medicine, primarily with rigid endoscopy for in vivo cancer detection.
• Flexible endoscopy offers improved access to difficult anatomical regions but introduces image artifacts.

Purpose of the Study:

• To develop and validate a hyperspectral imaging system using flexible endoscopy for medical applications.
• To create an effective filtering technique for removing fiber optic artifacts from hyperspectral data.

Main Methods:

• Developed a hyperspectral imaging system integrated with flexible endoscopy for laryngeal imaging.
• Designed a specialized filtering algorithm to eliminate the honeycomb pattern from hyperspectral data.
• Evaluated the filter's performance using objective metrics and classification algorithms (unsupervised and supervised).

Main Results:

• The developed filtering technique successfully removed the honeycomb pattern from hyperspectral images.
• The method minimized information loss, preserving small-area pixel variations crucial for analysis.
• Classification performance was significantly improved on pre-processed data compared to raw data and conventional filtering.

Conclusions:

• The novel filtering technique effectively removes fiber optic artifacts in flexible hyperspectral endoscopy.
• This advancement improves the accuracy of cancer detection by preserving essential image details.
• The system demonstrates feasibility for enhanced in vivo medical diagnostics in challenging anatomical sites.