This preliminary study aims to develop a non-invasive classification framework for estimating cervical dilation stages during labor by leveraging electrohysterographic (EHG) signal features and clinical parameters. Given the discomfort, infection risk, and variability associated with repeated vaginal examinations, alternative approaches based on EHG and machine learning may provide more consistent and patient-friendly labor monitoring. Seventy-three single-lead EHG recordings were segmented into 648 intervals, derived from low-risk pregnancies. Signals were decomposed into three frequency sub-bands, and 21 linear and non-linear features were extracted. These were combined with maternal age (MA), gestational age (GA), and low and high frequency contraction count metrics (LC, HC), yielding 25 predictors. Segments were labeled by dilation stage: Low (1–4 cm), Moderate (5–6 cm), or Advanced (7–10 cm). Data were split into training (70%) and test (30%) sets. Five key predictors were selected using ranking algorithms, and 33 classifiers were trained and evaluated. The macro-average F₁-score guided classifier selection. The best-performing model was the Bagged Tree, which achieved the maximum performance possible in classification on both the validation and test sets using only the MA, GA, LC, and HC features. These findings introduce a novel diagnostic paradigm combining uterine signal analysis with clinical data. This breakthrough redefines cervical dilation assessment by replacing subjective examinations with an objective, patient-safe computational framework.