Establishment Of Automatic Identification System Of Cardiac Malformations In Mouse

BackgroundAs the major cause of morbidity/mortality in infancy and childhood,1.4 million congenital heart diseases(CHD)are diagnosed every year worldwide,with a frequency of 0.8-1.2%among live births.Uncovering the underlying molecular mechanisms of CHD is crucial for developing effective prevention and intervention strategies.Mouse model is ideal for investigating CHD genetics,developmental etiology and mechanisms.However,for CHD and heart development researchers,phenotyping mouse cardiac malformations remains challenging.In previous studies,researchers mainly use analysis of histopathological sections to identify the cardiac structural phenotype.This method relies on long-term training and consumes a lot of time.On the other hand,two-dimensional(3D)section is hard to accurately reflect the cardiac malformations at 3D level.A rapid and stable identification system of cardiac malformations has become a fundamental need for the mechanical study of CHD based on mouse models.ResultsⅠ.Establishment of an automatic identification system for cardiac structural malformations in mice:CACCT.We used micro-computed tomography(micro-CT)to obtained consecutive cardiac images of C57BL6/J embryonic mice at 17.5 days of gestation(E17.5).To improve the identification efficiency of cardiac structural malformations in mice,we developed a 3D image analysis system named as CACCT(computer-assisted cardiac cavity tracking system).CACCT used threshold segmentation and region growth algorithm to extract the cardiac cavity regions in CT images,and then used the 3D relationship of cardiac cavity regions in different images to construct the 3D graphical data trove.The 3D graphical data trove was used to automatically identify the ventricular cavities and the great arterial lumens and judge the connection between the ventricle and the great artery.The results of chamber connection identified by CACCT were consistent with those of pediatric cardiac surgeons with 20 years of working experience(n=11).It took only 0.08±0.005 hours for CACCT to analyze one heart,and 29.90±1.33h per heart for manual labeling(n=5,p<0.001),suggesting CACCT could significantly improve the efficiency of cardiac structural phenotypic identification.By examining the CACCT cardiac cavity segmentation results section by section,we found that the cardiac cavity recognition accuracy of CACCT was more than 99.77%,and the coefficient of variation of the recognition result was less than 0.38%,indicating that CACCT had accurate and stable cardiac cavity segmentation ability.Ⅱ.CACCT identifies cardiac outflow tract malformation and ventricular septal defect.We administrated 60 mg/kg all-trans retinoic acid(RA)to pregnant mouse on E8.5 to construct cardiac structural abnormalities,including transposition of great arteries(TGA),double-outlet right ventricle(DORV),ventricular septal defect(VSD).The RA-treated hearts were harvested on E17.5 and scanned by micro-CT to obtain consecutive cardiac CT images.We used CACCT to analyze the CT images and found CACCT could accurately judge the connecting relationship of ventricles and the great arteries.In addition,CACCT could distinguish the normal heart(n=11),TGA(n=8)and DORV(n=22)with 100%accuracy.CACCT was able to automatically detect abnormal connections between the left and right ventricles and show the location and shape of VSD in the heart.By manually checking the VSD identified by CACCT section by section,we found that the accuracy of CACCT was 100%(n=22 hearts)for RA-induced VSD.Ⅲ.CACCT identifies semi-lunar valve malformations.We found that all Isl-1Cre/+;Ezh2fl/fl mice died within 1.5 days(P1.5)after birth,but we did not find obvious cardiac malformations by traditional two-dimensional phenotypic methods.We used CACCT to analyze the micro-CT images of Isl-1Cre/+;Ezh2fl/fl(n=4),Isl-1Cre/+;Ezh2fl/fl(n=3)and wildtype mice(n=3).We found oblique VSD in all Isl-1Cre/+;Ezh2fl/fl hearts(n=4),however no VSD in Isl-1Cre/+;Ezh2fl/fl(n=3)and wildtype mice(n=3)was detected.We applied semilunar valve detecting module of CACCT to automatically segment semilunar valve regions in cardiac CT images and obtained the 3D images of the semilunar valve of Isl-1Cre/+;Ezh2fl/fl mice.We found bicuspid aortic valve in all the hearts of Isl-1Cre/+;Ezh2fl/fl mice(n=4).Conclusion1)CACCT can automatically and accurately segment the ventricular cavity and large artery lumen in normal heart images and automatically judge the connecting relationship of cardiac cavities.2)CACCT can accurately identify normal heart,TGA and DORV induced by all-trans RA.3)CACCT can identify multiple VSDs at one time,including oblique VSD which is hard to be identified by traditional methods.4)CACCT can accurately identify the semilunar valve regions in cardiac CT images and identify bicuspid aortic valve.