Prediction of Idiopathic Recurrent Spontaneous Miscarriage using Machine Learning

Recurrent spontaneous miscarriage (RSM) is defined as the spontaneous loss of two or more clinically diagnosed pregnancies within 20 weeks of gestation. Despite extensive research, etiology remains undefined in 50% of RSM cases, and are classified as idiopathic. Thus, further study is warranted to understand molecular mechanism associated with the disease pathogenesis. In the present study, we aim to identify Raman fingerprints in endometrial/uterine tissues of women with history of idiopathic recurrent spontaneous miscarriage (IRSM) and controls by performing Raman spectroscopy with chemometric analysis and spectral classification models. Unsupervised analysis such as principal component analysis (PCA), hierarchical cluster analysis (HCA) and supervised analysis such as orthogonal projections to latent structures discriminant analysis (OPLS-DA) showed a distinct separation between IRSM and controls. The principal component loading plots indicated that proteins, amino acids, cholesterol and glutamate were responsible for the separation between the two groups. The pre-processed Raman spectral data were subjected to eight different machine learning (ML) classifiers with hyperparameter optimization to develop prediction models. Comparing the various algorithms, support vector machine (SVM), decision tree (DT), Extreme Gradient Boosting (XGBoost), convolutional neural network (CNN), and artificial neural network (ANN) outperform the other models based on accuracy (< 85%). Next, grid search and Bayesian optimization was used for tuning the hyperparameters of all methods. Further, 10-fold cross-validation was done to validate the model performances.