Objective To compare the diagnostic ef?cacy of the four-chamber view (FCV) and three-vessel view (3VV) in prenatal screening for congenital heart disease (CHD) in fetuses, and to construct an auxiliary identi?cation system for optimizing CHD screening strategies. Methods A retrospective analysis was conducted on 70 fetuses that underwent prenatal ultrasound screening at People’s Hospital of Tangyin County between January 2023 and December 2024. All of the fetuses received a de?nitive cardiac structural diagnosis postpartum. The detection rates of FCV and 3VV were analyzed separately across different types of CHD, and the sensitivity, speci?city, positive predictive value, negative predictive value and Kappa value of the two methods were evaluated. The distribution of false positives and false negatives was also analyzed. A deep learning-based fetal cardiac image-assisted identi?cation system was developed and the sensitivity, speci?city and area under the curve (AUC) of the image recognition model were assessed. Results A total of 35 fetuses were diagnosed with CHD. FCV detected 22 cases and 3VV detected 28 cases. FCV was more accurate at identifying structural abnormalities such as ventricular septal defects, single ventricle and hypoplastic left heart syndrome. In contrast, 3VV was better at identifying major vascular malformations such as Tetralogy of Fallot and Transposition of the Great Arteries. The sensitivity of 3VV was higher than that ofFCV (80.00% vs. 62.86%, P < 0.05), as was the Kappa value (0.752 vs. 0.584, P <0.05). Misdiagnosis analysis showed that false negatives in FCV were primarily due to vascular abnormalities, whereas false positives in 3VV were mainly caused by misinterpreting veins or image interference. The image-assisted identi?cation system demonstrated superior screening accuracy and consistency, with an overall sensitivity of 85.71%, a speci?city of 91.43%, a positive predictive value of 88.89%, a negative predictive value of 89.74%, and an AUC of 0.883. Conclusion Both FCV and 3VV have its own advantages in fetal CHD screening, but their use alone has certain limitations. It is recommended that both standard sections be used in combination in prenatal ultrasound to improve the detection rate and screening accuracy of different types of CHD. A deep learning-based fetal heart image-assisted identi?cation system has high sensitivity and speci?city and has certain clinical application value.