基于深度学习的冠状动脉CTA人工智能后处理对疑似冠心病患者诊断价值的初步研究

The Value of Artificial Intelligence of Coronary CTA Based On Deep Learning in Suspected Coronary Arteriosclerotic Heart Disease Patients

目的探讨基于深度学习的冠状动脉CT血管造影(CTA)人工智能后处理对疑似冠心病患者的诊断价值。方法对112例患者行双源双能冠状动脉CTA检查,将原始数据同时传输到数坤Coronary Doc工作站和Siemens Syngo View后处理工作站进行图像重组,前者设为AI组,后者设为对照组。对两组图像完成耗时、诊断耗时、图像Likert评分、诊断效能、狭窄符合率进行统计学比较。结果AI组与对照组图像完成耗时、诊断耗时均存在统计学差异(P<0.05),冠状动脉各主要分支Likert评分亦存在统计学差异(P<0.05)。以冠状动脉造影(CAG)为金标准,AI组与对照组敏感度分别为95.38%、95.31%,特异度分别为63.63%、16.67%,阳性预测率分别为93.94%、85.92%,阴性预测率分别为70.00%、40.00%,准确率分别为90.79%、82.89%,其中特异度有统计学差异(P<0.05)。AI组与对照组在冠状动脉轻度、重度狭窄的诊断上一致性较好,对中度狭窄的诊断一致性一般。结论冠状动脉CTA人工智能模式能获得高效稳定的图像重组及诊断同时也能获得较高质量的图像,在筛查及诊断疑似冠心病患者方面起到了精准可靠的辅助作用。

Objective To explore the value of coronary CTA artificial intelligence post-processing based on deep learning in the diagnosis of patients with suspected coronary heart disease.Methods 112 patients underwent dual-source dualenergy coronary CTA examination,and the original data were simultaneously transmitted to the Shukun Coronary Doc workstation and the Siemens Syngo View post-processing workstation for image reconstruction.The former was set as the AI group and the latter was set as the control group.A statistical comparison was made between the time-consuming image completion,time-consuming diagnosis,image Likert score,diagnostic efficiency,and stenosis coincidence rate between the two groups of images Results The AI group and the control group had statistical differences in image completion time and diagnosis time(P<0.05).There were also statistical differences in the Likert scores of the main branches of the coronary artery(P<0.05).Taking CAG as the gold standard,the sensitivity of the AI group and the control group were 95.38%and 95.31%,the specificity was 63.63%,16.67%,the positive prediction rates were 93.94%,85.92%,and the negative prediction rates were 70.00%,40.00%,the accuracy rates were 90.79%and 82.89%respectively,and the specificity was statistically different(P<0.05).The AI group and the control group have good consistency in the diagnosis of mild and severe coronary stenosis,while the consistency of moderate stenosis is general.Conclusion The coronary CTA artificial intelligence model can acquire efficient and stable image reconstruction and time-consuming diagnosis in screening and diagnosing patients with suspected coronary heart disease,while also obtaining high-quality images.It also plays an accurate and reliable complimentary role in diagnostic performance.