摘要目的:以心脏磁共振(cardiac magnetic resonance，CMR)为金标准，探讨加入人工智能技术的全自动三维超声右心室定量软件(3D Auto RV)评估心脏移植(heart transplantation，HT)术后右心室容积和右心室射血分数(right ventricular ejection fraction，RVEF)的可行性、准确性及可重复性。方法:前瞻性纳入2018年10月至2019年6月于华中科技大学同济医学院附属协和医院行超声心动图复查，并且同意于超声心动图复查后24 h内行CMR检查的HT术后患者46例。分别应用CMR技术、3D Auto RV和常规半自动三维超声右心室量化软件(Tomtec 4D RV function 2.0)获取右心室舒张末期容积(RVEDV)、右心室收缩末期容积(RVESV)、右心室每搏量(RVSV)及RVEF。分别将3D Auto RV、常规半自动Tomtec的测量结果与CMR的测量结果进行比较，比较方法采用配对样本 t检验、Pearson相关分析和一致性检验。 结果:3D Auto RV的可分析率为87%，该软件实现了在27例(59%)患者进行全自动分析，整个分析过程无需调节，分析时间仅需要(12±1)s；另外19例(41%)患者的分析结果需要手动调节，平均分析时间在2 min内，短于常规半自动Tomtec量化软件分析时间[(108±15)s对(160±34)s， P<0.001]。对于右心室容积：3D Auto RV和常规半自动Tomtec分析的RVEDV、RVESV和RVSV，均与CMR分析的相应测量值具有较高的相关性( r＝0.77～0.84，均 P<0.001)。与CMR测量值比较，3D Auto RV和常规半自动Tomtec技术均低估HT术后患者的RVEDV、RVESV和RVSV，但是3D Auto RV较常规半自动Tomtec的负性偏倚值更小。对于RVEF：3D Auto RV获得的RVEF与CMR获得的RVEF具有很高的相关性与一致性( r＝0.84， P<0.001；偏倚值＝-1.1%，一致性界限＝-8.1%～6.0%)。另外，3D Auto RV手动调节获取的右心室容积和RVEF与CMR相应测量值的相关性( r＝0.63～0.72，均 P<0.001)低于全自动分析获取的右心室容积和RVEF与CMR相应测量值的相关性( r＝0.76～0.82，均 P<0.001)。重复性分析显示3D Auto RV获取的RVEDV、RVESV、RVSV和RVEF均具有很好的重复性。 结论:3D Auto RV轻度低估HT术后右心室容积，但是其低估程度低于常规半自动TomTec。3D Auto RV可以准确量化HT术后RVEF，分析过程快速，且分析结果可重复性好，有望实现在HT术后患者随访中的常规临床应用。

abstractsObjective:To explore the feasibility, accuracy and reproducibility of a novel, fully automated three-dimensional echocardiography right ventricular(RV) quantification software(3D Anto RV) to evaluate the RV volume and RV ejection fraction (RVEF) using artificial intelligence in patients after heart transplantation (HT) comparing with the gold reference-cardiac magnetic resonance (CMR).Methods:Forty-six patients after HT who were scheduled for echocardiogram at their routine follow-up examinations and also agreed to undergo CMR examination within the following 24 hours in Union Hospital, Tongji Medical College, Huazhong University of Science and Technology from October 2018 to June 2019 were prospectively included. The right ventricular end-diastolic volume (RVEDV), right ventricular end-systolic volume (RVESV), right ventricular stroke volume (RVSV) and RVEF of HT patients were measured by CMR 3D Auto RV and conventional semi-automated three-dimensional echocardiography RV quantification software (Tomtec 4D RV function 2.0). The results of the 3D Auto RV and conventional semi-automated Tomtec were respectively compared with CMR using paired two-tailed student′s t-tests, Pearson correlation coefficients and Bland-Altman analyses. Results:The feasibility of the 3D Auto RV was 87%.The fully automated analysis realized in 27 (59%) patients by 3D Auto RV and the analysis time required only (12±1)s. The results of the remaining 19 (41%) patients needed manual adjustment and the mean analysis time in manual adjustment was also <2 min that was shorter than the conventional semi-automated three-dimensional echocardiography RV quantification software[(108±15)s vs (160±34)s, P<0.001]. For the results of RV volumes: There were good correlations between the 3D Auto RV and CMR, conventional semi-automated Tomtec and CMR for the measurements of RVEDV, RVESV and RVSV ( r=0.77-0.84, all P<0.001). In addition, compared with CMR, although there were significantly underestimated RV volumes by the 3D Auto RV and conventional semi-automated Tomtec, the negative bias was smaller in the 3D Auto RV than the conventional semi-automated Tomtec. For the results of RVEF: the corresponding RVEF derived from 3D Auto RV and CMR showed an excellent correlation and consistency ( r=0.84, P<0.001; bias=-1.1%, Limit of agreement=-8.1%-6.0%). In addition, the correlations between the manual adjustment by 3D Auto RV and the CMR ( r=0.63-0.72, all P<0.001) was lower than the correlations between the 3D Auto RV and the CMR ( r=0.76-0.82, all P<0.001) for RV volumes and RVEF.Finally, 3D Auto RV had a good reproducibility. Conclusions:The new fully 3D Auto RV quantification software underestimate RV volumes that less than the conventional semi-automated Tomtec. And the 3D Auto RV quantification software can accurately evaluate the RVEF in patients after HT with rapid analysis and higher reproducibility, which may also support the routine adoption of this method during follow-ups of HT patients in the daily clinical workflow.