Enhancing endometrial cancer detection: Blood serum intrinsic fluorescence data processing and machine learning application

Endometrial cancer (EC) is the most prevalent cancer within the female reproductive system in developed countries. Despite its high incidence, there is currently no established laboratory screening test for EC, making early detection challenging. This study introduces an innovative, minimally invasive, and cost-effective method utilizing three-dimensional fluorescence analysis combined with machine learning algorithms to enhance early EC detection. Intrinsic fluorescence of blood serum samples was measured using a luminescence spectrophotometer, which captured fluorescence spectra as synchronous excitation spectra and visualized them through wavelength contour matrices. The spectral data were processed using machine learning algorithms, including Random Forest (RF), Logistic Regression (LR), Support Vector Machine (SVM), and Stochastic Gradient Descent (SGD), along with exploratory techniques such as Principal Component Analysis (PCA) and Partial Least Squares Discriminant Analysis (PLS-DA). Fluorescence ratios R300/330 and R360/490, indicative of altered tryptophan metabolism and redox state changes, were identified as fluorescent spectral markers and represent key metabolic biomarkers. These ratios demonstrated high diagnostic efficacy with AUC values of 0.88 and 0.91, respectively. Among the ML algorithms, LR and RF exhibited high sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV), showing significant promise for clinical application. After optimization, LR achieved a sensitivity of 0.94, specificity of 0.89, and an impressive AUC value of 0.94. The application of this novel approach in laboratory diagnostics has the potential to significantly enhance early detection and improve prognosis for EC patients.