The proposed hybrid attention-based deep learning model achieved a classification accuracy of 97.94% in detecting fetal hypoxia from cardiotocography images.
Does a deep-learning-based computer vision approach combining attention mechanisms and CNNs accurately classify fetal health states from CTG images?
A hybrid attention-based deep learning approach directly processing raw CTG images achieved 97.94% accuracy in classifying fetal health states, offering a potential tool for early detection of fetal hypoxia.
Absolute Event Rate: 97.94% vs 93.88%
Objective To develop a deep-learning-based computer vision approach for fetal heart rate (FHR) monitoring that can efficiently detect fetal hypoxia without relying on complex feature extraction methods. Methods A hybrid attention mechanism was proposed for direct processing of fetal monitoring images (cardiotocography, CTG), eliminating the need for manual feature extraction. The method leverages deep learning to classify fetal health states based on real-time CTG images. Results Experiments on a real-world clinical dataset demonstrated that the proposed method achieved a classification accuracy of 97.94%, indicating its high efficiency in detecting fetal hypoxia. Conclusion The proposed hybrid attention-based deep learning approach provides reliable support for the early detection of fetal hypoxia, overcoming the limitations of traditional machine learning methods that rely on complex feature extraction.
Wang et al. (2026) studied Fetal hypoxia (n=489). Hybrid attention-based deep learning model (ResNet18-HA) vs. Baseline ResNet-18 model without hybrid attention was evaluated on Classification accuracy for fetal health states. The proposed hybrid attention-based deep learning model achieved a classification accuracy of 97.94% in detecting fetal hypoxia from cardiotocography images.