基于人工智能的自动应变定量技术评价慢性肾病患者早期心肌损伤的临床价值

Clinical Value of Automatic Strain Quantitative Technology Based on Artificial Intelligence in Evaluating Early Myocardial Injury in Patients with Chronic Kidney Disease.

目的探讨基于人工智能的自动应变定量技术评价慢性肾病(CKD)患者早期心肌损伤的临床价值。方法选取2021年7月至2023年7月我院收治的CKD患者125例为研究对象,根据CKD分期分别纳入A组(2~3期,67例)、B组(4~5期,58例),另选取同期45例健康体检者为对照组。比较3组二维超声心动图参数[左室射血分数(LVEF)、舒张末期室间隔厚度(IVSd)、舒张末期左室后壁厚度(LVPWd)、左心室舒张末期内径(LVEDD)、右心室内径(RVD)]、血清N末端脑钠肽前体(NT-pro BNP)水平、自动及手动应变定量技术应变参数[左心室整体纵向收缩期峰值应变(LVGLS)、右心室整体纵向收缩期峰值应变(RVGLS)、左房储备期纵向应变(LASr)、左房管道期纵向应变(LAScd)、左房收缩期纵向应变(LASct)],采用Pearson分析自动应变参数与手动应变参数、二维超声心动图参数及血清NT-pro BNP水平的相关性,比较LVGLS、RVGLS、左房应变分析的手动、自动应变分析时间,采用受试者工作特征(ROC)曲线分析自动应变参数联合检测对CKD患者早期心肌损伤的诊断价值。结果B组LVEF水平低于A组、对照组,B组IVSd、LVPWd、LVEDD、RVD及血清NT-pro BNP水平高于A组、对照组,且A组高于对照组,差异有统计学意义(P<0.05);B组自动及手动LVGLS、RVGLS、LAScd、LASct均高于A组、对照组,B组LASr低于A组、对照组,且A组自动LVGLS、RVGLS、LAScd、LASct高于对照组,LASr低于对照组,A组手动LAScd、LASct高于对照组,LASr低于对照组,差异有统计学意义(P<0.05);自动分析各应变参数与手动分析各参数均呈正相关(r=0.852,0.721,0.605,0.630,0.627,P<0.05);自动LVGLS、RVGLS、LAScd、LASct与LVEF呈负相关(r=-0.659,-0.598,-0.612,-0.601,P<0.05),与NT-pro BNP呈正相关(r=0.643,0.638,0.622,0.647,P<0.05);自动LASr与LVEF呈正相关(r=0.629,P<0.05),与NT-pro BNP呈负相关(r=-0.658,P<0.05)。自动病变分析时间较手动分析时间明显减少,差异有统计学意义(P<0.05)。LVGLS诊断CKD患者早期心肌损伤的曲线下面积(AUC)为0.928(95%CI:0.889~0.957),约登指数为0.736,最佳诊断敏感度为87.20%,特异度为86.40%,差异有统计学意义(P<0.05)。结论自动应变定量技术应变参数联合检测对CKD患者早期心肌损伤具有一定诊断价值,可作为临床评估早期心肌损伤的有效检查方式,并对后续治疗具有一定指导价值。

Objective To explore the clinical value of automatic strain quantitative technology based on artificial intelligence in evaluating early myocardial injury in patients with chronic kidney disease(CKD).Methods A total of 125 CKD patients admitted to the our hospital from July 2021 to July 2023 were selected as the research subjects,and were divided into group A(stage 2~3,67 cases)and group B(stage 4~5,58 cases)according to CKD staging.In addition,45 healthy physical examination subjects were selected as the control group.The two-dimensional echocardiographic parameters[left ventricular ejection fraction(LVEF),interventricular septal thickness at end diastole(IVSd),left ventricular posterior wall thickness at end diastole(LVPWd),left ventricular end-diastolic internal diameter(LVEDD),right ventricular internal diameter(RVD)],serum N-terminal pro-brain natriuretic peptide(NT-pro BNP)levels,automatic and manual strain quantitative technology strain parameters[left ventricular global longitudinal systolic peak strain(LVGLS),right ventricular global longitudinal systolic peak strain(RVGLS),left atrial reserve period longitudinal strain(LASr),left atrial conduit period longitudinal strain(LAScd),left atrial systolic longitudinal strain(LASct)]were compared among the three groups.Pearson analysis was used to analyze the correlation between automatic strain parameters and manual strain parameters,two-dimensional echocardiographic parameters,and serum NT-pro BNP levels.The manual and automatic strain analysis times of LVGLS,RVGLS,and left atrial strain analysis were compared,and the diagnostic value of the combined automatic strain parameter detection for early myocardial injury in CKD patients was analyzed by using the subject's work characteristic(ROC)curve.Results The LVEF level in group B was lower than that in group A and the control group.The levels of IVSd,LVPWd,LVEDD,RVD,and serum NT-pro BNP in group B were higher than those in group A and the control group,and the levels in group A were higher than those in the control group,the differences were statistically significant(P<0.05).The automatic and manual LVGLS,RVGLS,LAScd,and LASct in group B were higher than those in group A and the control group.The LASr in group B was lower than that in group A and the control group,and the automatic LVGLS,RVGLS,LAScd,LASct in group A were higher than those in the control group,while LASr was lower than the control group.The differences were statistically significant(P<0.05).Both automatic and manual analysis of strain parameters showed positive correlation(r=0.852,0.721,0.605,0.630,0.627,P<0.05).Automatic LVGLS,RVGLS,LAScd,LASct were negatively correlated with LVEF(r=-0.659,-0.598,-0.612,-0.601,P<0.05),and positively correlated with NT-pro BNP(r=0.643,0.638,0.622,0.647,P<0.05).Automatic LASr was positively correlated with LVEF(r=0.629,P<0.05)and negatively correlated with NT-pro BNP(r=-0.658,P<0.05).The time for automatic lesion analysis was significantly shorter than that for manual analysis,and the difference was statistically significant(P<0.05).The area under the curve(AUC)of LVGLS for diagnosing early myocardial injury in CKD patients was 0.928(95%CI 0.889~0.957),with a Youden's index of 0.736,an optimal diagnostic sensitivity of 87.20%,and a specificity of 86.40%,and the differences were statistically significant(P<0.05).Conclusion Automatic strain quantification technique combined strain parameter detection has a certain diagnostic value for early myocardial injury in patients with CKD,can be used as an effective examination method for clinical evaluation of early myocardial injury,and has a certain guiding value for subsequent treatment.