目的 探讨容积超声胎儿心脏计算机辅助诊断技术（SonoVCADheart）在胎儿冠状动脉形态评估中的准确性与临床应用价值。 方法 选取2024年7月至2025年9月于珠海市中西医结合医院接受产前检查并具有胎儿冠状动脉异常高危因素的322例中晚期孕妇为研究对象，所有孕妇均接受SonoVCADheart技术与传统胎儿超声心动图检查。以出生后超声心动图检查为金标准，比较两种技术在冠状动脉内径、走行等方面的诊断效能。 结果 共纳入322例中晚期孕妇，根据图像质量质控标准，22例（6.83%）因原始二维四腔心切面图像质量评分≤ 2分（图像模糊、伪影严重或胎儿体位遮挡导致无法满足STIC采集要求）被排除，未纳入后续分析。300例孕妇中，经产后超声心动图确诊冠状动脉正常胎儿278例，冠状动脉异常22例，异常发生率为7.33%。其中冠状动脉起源异常2例（9.09%），包括右冠状动脉起源于左冠窦1例、冠状动脉高位起源1例；冠状动脉扩张20例（90.91%），均为弥漫性扩张，且20例均合并存在胎儿生长受限（FGR）。SonoVCADheart技术诊断胎儿冠状动脉异常的真阳性20例、假阳性2例、真阴性276例、假阴性2例；传统胎儿超声心动图诊断的真阳性16例、假阳性7例、真阴性271例、假阴性6例。2例起源异常中，SonoVCADheart正确识别1例（右冠状动脉起源于左冠窦），漏诊1例（冠状动脉高位起源）；20例扩张中，正确诊断19例，漏诊1例。传统胎儿超声心动图2例起源异常均被漏诊，20例扩张病例中正确诊断16例，漏诊4例。两种方法的假阳性病例中，SonoVCADheart的2例均误判为扩张，传统超声的7例中5例误判为扩张、2例误判为起源异常。SonoVCADheart技术诊断胎儿冠状动脉异常的准确率高于传统胎儿超声心动图（P 0.05），敏感性、特异性、阳性预测值、阴性预测值差异无统计学意义（P 0.05）。两种方法测量左冠状动脉主干、右冠状动脉内径的线性Pearson相关系数为0.865、0.859，均呈正相关（P 0.05），提示两种方法测量结果的相关性较高。ICC分析显示，两种方法测量左冠状动脉主干、右冠状动脉的ICC分别为0.848、0.853，一致性良好。 结论 SonoVCADheart技术对胎儿冠状动脉形态学改变方面具有更高诊断效能，可减少漏诊误诊，在冠状动脉内径测量中稳定性更优，能为病变的检出、分型与量化评估提供精准数据。

Objective To evaluate the diagnostic accuracy and clinical utility of SonoVCADheart， a volume ultrasound-based computer-aided diagnosis system for the fetal heart， in assessing fetal coronary artery morphology compared with conventional fetal echocardiography. Methods A total of 322 high-risk pregnant women in the second and third trimesters were prospectively enrolled between July 2024 and September 2025. All participants underwent both conventional fetal echocardiography and SonoVCADheart imaging. Postnatal echocardiography served as the reference standard. Diagnostic performance for coronary artery origin， course， and luminal diameter was compared between the two modalities. Twenty-two cases were excluded due to inadequate image quality （four-chamber view score ≤ 2）， leaving 300 for final analysis. Results Among the 300 analyzed fetuses， postnatal follow-up confirmed normal coronary anatomy in 278 and abnormalities in 22 （7.33%）， comprising 2 anomalous origins and 20 diffuse dilations （all associated with fetal growth restriction）. SonoVCADheart demonstrated significantly higher overall diagnostic accuracy than conventional echocardiography （P 0.05）. Specifically， SonoVCADheart identified 20 true positives， 2 false positives， 276 true negatives， and 2 false negatives （sensitivity 90.9%， specificity 99.3%）， whereas conventional echocardiography detected 16 true positives， 7 false positives， 271 true negatives， and 6 false negatives （sensitivity 72.7%， specificity 97.5%）. Although sensitivity， specificity， positive predictive value， and negative predictive value were numerically higher for SonoVCADheart， these differences did not reach statistical significance （all P 0.05）. For coronary artery diameter measurements， Pearson correlation coefficients between the two methods were 0.865 for the left main and 0.859 for the right coronary artery （both P 0.05）. Intraclass correlation coefficients （ICCs） were 0.848 and 0.853， respectively， indicating strong inter-method agreement. Conclusions SonoVCADheart provides superior diagnostic accuracy and more reliable quantitative measurements of fetal coronary artery morphology compared with conventional echocardiography. It effectively reduces missed diagnoses and misclassifications， particularly for coronary origin anomalies and secondary dilations， offering valuable support for precise prenatal detection， classification， and clinical management.