Machine Learning in Body Composition Analysis

Body composition analysis (BCA) generates objective anthropometric data that can inform prognostication and treatment decisions across a wide variety of urologic conditions. A patient’s body composition, specifically muscle and adipose tissue mass, may be characterized via segmentation of cross-sectional images (computed tomography, magnetic resonance imaging) obtained as part of routine clinical care. Unfortunately, conventional semi-automated segmentation techniques are time- and resource-intensive, precluding translation into clinical practice. Machine learning (ML) offers the potential to automate and scale rapid and accurate BCA. To date, ML for BCA has relied on algorithms called convolutional neural networks designed to detect and analyze images in ways similar to human neuronal connections. This mini review provides a clinically oriented overview of ML and its use in BCA. We address current limitations and future directions for translating ML and BCA into clinical practice. Body composition analysis is the measurement of muscle and fat in your body based on analysis of computed tomography or magnetic resonance imaging scans. We discuss the use of machine learning to automate body composition analysis. The information provided can be used to guide shared decision-making and to help in identifying the best therapy option. Body composition analysis (BCA) generates objective anthropometric data that can inform prognostication and treatment decisions across a wide variety of urologic conditions. Machine learning offers the potential to automate and scale rapid and accurate BCA using convolutional neural networks designed to detect and analyze images in ways similar to human neuronal connections. This mini review provides a clinically oriented overview of machine learning and how it has been operationalized for the purposes of body composition analysis.