Search

Automated cardiac image interpretation has the potential to transform clinical practice in multiple ways, including enabling serial assessment of cardiac function by nonexperts in primary care and rural settings. We hypothesized that advances in computer vision could enable building a fully automated, scalable analysis pipeline for echocardiogram interpretation, including (1) view identification, (2) image segmentation, (3) quantification of structure and function, and (4) disease detection.Using 14 035 echocardiograms spanning a 10-year period, we trained and evaluated convolutional neural network models for multiple tasks, including automated identification of 23 viewpoints and segmentation of cardiac chambers across 5 common views. The segmentation output was used to quantify chamber volumes and left ventricular mass, determine ejection fraction, and facilitate automated determination of longitudinal strain through speckle tracking. Results were evaluated through comparison to manual segmentation and measurements from 8666 echocardiograms obtained during the routine clinical workflow. Finally, we developed models to detect 3 diseases: hypertrophic cardiomyopathy, cardiac amyloid, and pulmonary arterial hypertension.Convolutional neural networks accurately identified views (eg, 96% for parasternal long axis), including flagging partially obscured cardiac chambers, and enabled the segmentation of individual cardiac chambers. The resulting cardiac structure measurements agreed with study report values (eg, median absolute deviations of 15% to 17% of observed values for left ventricular mass, left ventricular diastolic volume, and left atrial volume). In terms of function, we computed automated ejection fraction and longitudinal strain measurements (within 2 cohorts), which agreed with commercial software-derived values (for ejection fraction, median absolute deviation=9.7% of observed, N=6407 studies; for strain, median absolute deviation=7.5%, n=419, and 9.0%, n=110) and demonstrated applicability to serial monitoring of patients with breast cancer for trastuzumab cardiotoxicity. Overall, we found automated measurements to be comparable or superior to manual measurements across 11 internal consistency metrics (eg, the correlation of left atrial and ventricular volumes). Finally, we trained convolutional neural networks to detect hypertrophic cardiomyopathy, cardiac amyloidosis, and pulmonary arterial hypertension with C statistics of 0.93, 0.87, and 0.85, respectively.Our pipeline lays the groundwork for using automated interpretation to support serial patient tracking and scalable analysis of millions of echocardiograms archived within healthcare systems.

BackgroundMeCP2, methyl-CpG-binding protein 2, binds to methylated cytosines at CpG dinucleotides, as well as to unmethylated DNA, and affects chromatin condensation.MECP2mutations in females lead to Rett syndrome, a neurological disorder characterized by developmental stagnation and regression, loss of purposeful hand movements and speech, stereotypic hand movements, deceleration of brain growth, autonomic dysfunction and seizures. Most mutations occurde novoduring spermatogenesis. Located at Xq28,MECP2is subject to X inactivation, and affected females are mosaic. Rare hemizygous males suffer from a severe congenital encephalopathy.MethodsTo identify the pathways mis-regulated by MeCP2 deficiency, microarray-based global gene expression studies were carried out in cerebellum ofMecp2mutant mice. We compared transcript levels in mutant/wildtype male sibs of two different MeCP2-deficient mouse models at 2, 4 and 8 weeks of age. Increased transcript levels were evaluated by real-time quantitative RT-PCR. Chromatin immunoprecipitation assays were used to documentin vivoMeCP2 binding to promoter regions of candidate target genes.ResultsOf several hundred genes with altered expression levels in the mutants, twice as many were increased than decreased, and only 27 were differentially expressed at more than one time point. The number of misregulated genes was 30% lower in mice with the exon 3 deletion (Mecp2tm1.1Jae) than in mice with the larger deletion (Mecp2tm1.1Bird). Between the mutants, few genes overlapped at each time point. Real-time quantitative RT-PCR assays validated increased transcript levels for four genes:Irak1, interleukin-1 receptor-associated kinase 1;Fxyd1, phospholemman, associated with Na, K-ATPase;Reln, encoding an extracellular signaling molecule essential for neuronal lamination and synaptic plasticity; andGtl2/Meg3, an imprinted maternally expressed non-translated RNA that serves as a host gene for C/D box snoRNAs and microRNAs. Chromatin immunoprecipitation assays documentedin vivoMeCP2 binding to promoter regions ofFxyd1, Reln, andGtl2.ConclusionTranscriptional profiling of cerebellum failed to detect significant global changes inMecp2-mutant mice. Increased transcript levels ofIrak1, Fxyd1, Reln, andGtl2may contribute to the neuronal dysfunction in MeCP2-deficient mice and individuals with Rett syndrome. Our data provide testable hypotheses for future studies of the regulatory or signaling pathways that these genes act on.

Rett syndrome (RTT) is a neurodevelopmental disorder characterized by cognitive regression, loss of purposeful hand movements and speech, stereotypies, ataxia, seizures, mental retardation and acquired microcephaly. Mutations in MECP2, encoding methyl-CpG-binding protein 2, are responsible for approximately 90% of classic RTT cases. RTT displays phenotypic overlap with Angelman syndrome, a disorder caused by loss of expression of the imprinted gene UBE3A. MeCP2 binds to methylated DNA and may alter the expression of imprinted genes, thereby suggesting a mechanistic link between the two disorders. Here, we tested the hypothesis that MeCP2 deficiency affects expression of Ube3a in mouse models of RTT. As Ube3a is only imprinted in brain, we evaluated Ube3a expression in brains of 15 different litters of neonatal or 8-week-old male Mecp2 mutant mice by real-time quantitative RT-PCR and western blot analysis. We found no significant differences between Mecp2(tm1.1Bird/Y) or Mecp2(tm1.1Jae/Y) mutants and their wild-type male siblings that served as negative controls. In positive control mice carrying a maternally inherited Ube3a deletion, Ube3a sense transcript and protein levels were drastically reduced. Our data contrast with two recent reports of substantially decreased Ube3a expression in brain tissues of MeCP2-deficient mice. We, therefore, challenge the conclusion that decreased UBE3A/Ube3a expression contributes to the pathophysiology of RTT.