Cardiotocography (CTG) is essential for fetal monitoring but is frequently compromised by diverse artefacts, including beat halving and doubling, maternal heart rate interference, missing signal and isolated spikes. These obscure true fetal heart rate (FHR) patterns and can lead to misdiagnosis or delayed intervention. Current deep‐learning approaches typically bypass comprehensive artefact handling, applying minimal preprocessing or focusing solely on downstream classification, while traditional methods rely on simple interpolation or rule‐based filtering that addresses only missing samples and fail to correct complex artefact types. We present CleanCTG, an end‐to‐end dual‐stage model that first identifies multiple artefact types via multi‐scale convolution and context‐aware cross‐attention, then reconstructs corrupted segments through artefact‐specific correction branches. Training utilised over 800,000 min of physiologically realistic, synthetically corrupted CTGs derived from expert‐verified “clean” recordings. External validation on 10,190 min of clinician‐annotated segments yielded AU-ROC = 0.95 (sensitivity = 83.44%, specificity 94.22%), surpassing six comparator classifiers. When integrated with the Dawes-Redman system on 933 clinical CTG recordings, denoised traces increased specificity (from 80.70% to 82.70%) and shortened median time to decision by 33%. On synthetic data, CleanCTG achieved perfect artefact detection (AU-ROC = 1.00) and reduced mean squared error (MSE) on corrupted segments to 2.74 × 10⁻4 (clean‐segment MSE = 2.40 × 10−⁶), outperforming the next best method by more than 60%. These findings demonstrate that explicit artefact removal and signal reconstruction can improve the reliability and efficiency of CTG interpretation, offering a practical pathway to faster and more robust clinical decision-making.