Raman fiber-optical method for colon cancer detection: Cross-validation and outlier identification approach

Area of Science:

• Medical Technology
• Gastroenterology
• Spectroscopy

Background:

• Endoscopy is crucial for early cancer detection and survival rates.
• Raman spectroscopy offers objective, cost-effective tissue characterization during endoscopy.
• Accurate differentiation of precancerous lesions and risk stratification is vital in gastroenterology.

Purpose of the Study:

• To evaluate Raman spectroscopic fiber-optical approaches for tissue characterization in gastroenterology.
• To optimize bioinformatic analysis and experimental design for endoscopic Raman measurements.
• To compare different cross-validation and outlier identification methods for spectral data analysis.

Main Methods:

• Raman fiber-optical measurements on colon biopsy samples (adenocarcinoma, tubular adenomas, hyperplastic polyps, normal tissue) from 151 patients.
• Analysis of spectral data using leave-one-patient-out, leave-one-sample-out, and leave-one-spectrum-out cross-validation.
• Investigation of three outlier identification methods to minimize systematic errors and spectral dilution.

Main Results:

• High-risk lesions differentiated from low-risk lesions with 79% sensitivity, 74% specificity, and 77% accuracy.
• Cancer distinguished from normal tissue with 79% sensitivity, 83% specificity, and 81% accuracy.
• Outlier identification improved the recognition of neoplastic biopsy samples.

Conclusions:

• Raman spectroscopy fiber-optical methods show promise for objective and efficient tissue characterization in endoscopic procedures.
• Optimized data analysis, including cross-validation and outlier identification, is essential for accurate spectral differentiation of gastrointestinal tissues.
• This approach can aid in early detection and risk stratification of gastrointestinal cancers.