Answers are grounded in this paper and cite their sources.
A minimally invasive detection method is highly desirable for endometriosis due to its long diagnostic delay, non-specific symptoms, and invasive diagnostic procedures. This study explored the synergetic application of vibrational spectroscopy (ATR-FTIR) and protein profile (HPLC-FLD) analysis of blood plasma samples for the detection of endometriosis. The different feature selection tools, such as genetic algorithm (GA), variable importance in projection (VIP) and random forest variable importance (RFVI), were adapted to screen the most significant variables among the whole data. The machine learning models, such as support vector machines (SVM), K-Nearest Neighbours (KNN) and shallow neural network (NN), were applied to the data for the classification of endometriosis from control. The protein profile data with common features selected out of feature selection tools such as GA and VIP, and KNN as a classifier resulted in a sensitivity of 73% an area under the ROC curve of 0.77. Furthermore, the explainable AI tool SHAP (Shapley additive explanations) values was used to identify the blood plasma fractions responsible for classifying endometriosis relative to controls. The untargeted mass spectrometry analysis of these blood plasma fractions revealed a panel of proteins that were responsible for the classification of endometriosis and probable protein biomarkers of endometriosis. The Mid-level data fusion (extracting PLS-DA score) was applied on the ATR-FTIR and HPLC-FLD data set with feature selection algorithms, which resulted in an 80% sensitivity and AUC of 0.80. This study suggests that a multimodal analytical approach with suitable data fusion can be a promising tool for minimally invasive endometriosis detection.
No takes yet. Share an insight, caveat, or question.
M et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: