Non-Invasive Endometrial Cancer Screening through Urinary Fluorescent Metabolome Profile Monitoring and Machine...

Area of Science:

• Biomedical diagnostics
• Analytical chemistry
• Metabolomics

Background:

• Endometrial cancer incidence is rising, necessitating advanced, non-invasive diagnostic tools.
• Current diagnostic methods for endometrial cancer can be invasive and may lack sufficient accuracy.
• Urinary fluorescence spectroscopy offers a potential non-invasive approach to analyze metabolic profiles.

Purpose of the Study:

• To evaluate urinary fluorescence spectroscopy as a diagnostic method for endometrial cancer.
• To develop spectral markers for differentiating endometrial cancer from benign conditions and controls.
• To assess the efficacy of machine learning models in conjunction with spectroscopy for endometrial cancer detection.

Main Methods:

• Synchronous fluorescence spectroscopy was used to analyze urine samples from endometrial cancer patients, benign uterine tumor patients, and controls.
• Specific fluorescence ratios were calculated to distinguish between sample groups.
• Partial Least Squares Discriminant Analysis (PLS-DA) and various machine learning models (Random Forest, Logistic Regression, SVM, SGD) were applied for data analysis and classification.

Main Results:

• Spectral markers showed potential clinical applicability with an Area Under the Curve (AUC) of up to 80% for differentiating sample groups.
• PLS-DA achieved an overall accuracy of 79% and an AUC of 90% in distinguishing controls from endometrial cancer patients.
• Machine learning models demonstrated effectiveness in classifying samples based on urinary fluorescence profiles.

Conclusions:

• Urinary fluorescence spectroscopy is a promising non-invasive technique for endometrial cancer diagnosis.
• The integration of spectroscopy with machine learning models can significantly enhance diagnostic accuracy.
• This approach offers a potential rapid, accurate, and non-invasive alternative to existing endometrial cancer diagnostic methods.