Novel Z Scores to Correct Biases Due to Ventricular Volume Indexing to Body Surface Area in Adolescents and Young Adults.

BACKGROUND: Reference values for cardiac magnetic resonance imaging (cMRI) in children and young adults are scarce. This leads to risk stratification of patients with congenital heart diseases being based on volumes indexed to body surface area (BSA). We aimed to produce cMRI Z score equations for ventricular volumes in children and young adults and to test whether indexing to BSA resulted in an incorrect assessment of ventricular dilation according to sex, body composition, and growth. METHODS: We retrospectively included 372 subjects aged < 26 years with either normal hearts or conditions with no impact on ventricular volumes (reference group), and 205 subjects with repaired tetralogy of Fallot (TOF) aged < 26 years. We generated Z score equations by means of multivariable regression modelling. Right ventricular dilation was assessed with the use of Z scores and compared with indexing to BSA in TOF subjects. RESULTS: Ventricular volume Z scores were independent from age, sex, and anthropometric measurements, although volumes indexed to BSA showed significant residual association with sex and body size. In TOF subjects, indexing overestimated dilation in growing children and underestimated dilation in female compared with male subjects, and in overweight compared with lean subjects. CONCLUSIONS: Indexed ventricular volumes measured with cMRI did not completely adjust for body size and resulted in a differential error in the assessment of ventricular dilation according to sex and body size. Our proposed Z score equations solved this problem. Future studies should evaluate if ventricular volumes expressed as Z scores have a better prognostic value than volumes indexed to BSA.

Deep learning workflow in radiology: a primer.

Interest for deep learning in radiology has increased tremendously in the past decade due to the high achievable performance for various computer vision tasks such as detection, segmentation, classification, monitoring, and prediction. This article provides step-by-step practical guidance for conducting a project that involves deep learning in radiology, from defining specifications, to deployment and scaling. Specifically, the objectives of this article are to provide an overview of clinical use cases of deep learning, describe the composition of multi-disciplinary team, and summarize current approaches to patient, data, model, and hardware selection. Key ideas will be illustrated by examples from a prototypical project on imaging of colorectal liver metastasis. This article illustrates the workflow for liver lesion detection, segmentation, classification, monitoring, and prediction of tumor recurrence and patient survival. Challenges are discussed, including ethical considerations, cohorting, data collection, anonymization, and availability of expert annotations. The practical guidance may be adapted to any project that requires automated medical image analysis.