Investigating the potential of combining Raman spectroscopy and machine learning for gestational age classification in rat cervical tissue.

Spontaneous preterm birth (SPB) is a leading cause of neonatal morbidity and mortality worldwide. It occurs when the uterine cervix (UC) opens prematurely due to various biological factors. Early detection is crucial to reducing its adverse outcomes. Current prediction methods, such as cervical length measurement and fetal fibronectin testing, often suffer from high false-positive rates due to the heterogeneous nature of SPB. Prior to labor, significant biochemical changes occur in the cervical tissue. Raman spectroscopy (RS), a non-invasive and label-free technique, can detect these molecular alterations without harming the mother or fetus. This study represents the first application of RS combined with machine learning (ML) algorithms for gestational age classification in rat cervical tissue. Ex vivo Raman spectra were collected from Wistar-Albino rats on gestational days 15, 17, 18, 19, and 20, yielding a dataset of 600 spectra. Several ML algorithms were evaluated, including decision trees, random forests, support vector machines (SVM), K-nearest neighbors (K-NN), and Artificial neural networks (ANN). Among these, the ANN model achieved the highest classification accuracy (87.5 %), outperforming other ML models (58-69 %). Notably, spectra from gestational day 19 showed increased misclassification, which likely reflects heightened biochemical variability in the cervix during late pregnancy. These findings demonstrate that integrating RS with ML, particularly ANN, provides a promising approach for classifying gestational stage based on cervical biochemical composition. This study supports the potential of RS as a non-invasive tool for monitoring pregnancy progression and lays the groundwork for future clinical applications in early SPB risk assessment.