Probability-based differential normalized fluorescence bivariate analysis for the classification of tissue...

Area of Science:

• Biomedical Optics
• Medical Diagnostics
• Cancer Research

Background:

• Differential normalized fluorescence (DNF) is an established method for distinguishing cancerous from normal tissue based on fluorescence spectra.
• Current DNF methods extract diagnostic features from spectral differences for tissue classification.

Purpose of the Study:

• To introduce a novel probability-based DNF bivariate analysis method.
• To enhance the accuracy of cancer diagnosis by utilizing two differentiation features simultaneously.

Main Methods:

• Developed a probability-based DNF bivariate analysis building upon the univariate DNF approach.
• Employed Bayes decision theory to determine the probability of each sample belonging to a disease state.
• Utilized two differentiation features concurrently for improved classification.

Main Results:

• Demonstrated improved accuracy in colonic cancer diagnosis using a dataset of 57 tissue sites.
• The probability approach provides classification based on disease states and includes uncertainty information.
• The bivariate DNF analysis requires minimal spectral data points, suggesting potential for faster acquisition.

Conclusions:

• Probability-based DNF bivariate analysis offers a more accurate method for cancer diagnosis compared to previous DNF techniques.
• This novel approach enhances tissue classification and provides valuable uncertainty metrics.
• The method shows promise for accelerating tissue imaging acquisition speeds in clinical settings.