In vivo multi spectral colonoscopy in mice

Area of Science:

• Gastroenterology
• Medical Imaging
• Computational Biology

Background:

• Multispectral and hyperspectral endoscopy can enhance neoplastic lesion detection during colonoscopy.
• Previous research often used ex vivo samples, raising questions about in vivo applicability.

Purpose of the Study:

• To evaluate the effectiveness of multispectral endoscopy for in vivo detection of colonic neoplastic lesions.
• To determine if incorporating spatial features improves detection rates.

Main Methods:

• Utilized machine learning for data analysis on a dataset of 25 mice with colonic neoplastic lesions.
• Investigated the impact of adding spatial features derived from Gauss-Laguerre polynomials.

Main Results:

• Achieved a Matthews Correlation Coefficient (MCC) of 0.47 for neoplastic lesion detection.
• Demonstrated comparable classification accuracy between multispectral and hyperspectral colonoscopy when spatial features were included.

Conclusions:

• Multispectral endoscopy is effective for in vivo colonic neoplastic lesion detection.
• Results suggest that challenges in transferring ex vivo findings to in vivo settings can be overcome.
• Supports the advancement of multi/hyperspectral endoscopy towards clinical implementation.