Uncertainty-Aware Body Composition Analysis with Deep Regression Ensembles on UK Biobank MRI

Abstract

To enable fast and automated analysis of body composition from UK Biobank MRI with accurate estimates of individual measurement errors. Methods: In an ongoing large-scale imaging study the UK Biobank has acquired MRI of over 40,000 men and women aged 44-82. Phenotypes derived from these images, such as body composition, can reveal new links between genetics, cardiovascular disease, and metabolic conditions. In this retrospective study, neural networks were trained to provide six measurements of body composition from UK Biobank neck-to-knee body MRI. A ResNet50 architecture can automatically predict these values by image-based regression, but may also produce erroneous outliers. Predictive uncertainty, which could identify these failure cases, was therefore modeled with a mean-variance loss and ensembling. Its estimates of individual prediction errors were evaluated in cross-validation on over 8,000 subjects, tested on another 1,000 cases, and finally applied for inference. Results: Relative measurement errors below 5% were achieved on all but one target, for intra-class correlation coefficients (ICC) above 0.97 both in validation and testing. Both mean-variance loss and ensembling yielded improvements and provided uncertainty estimates that highlighted some of the worst outlier predictions. Combined, they reached the highest quality, but also exhibited a consistent bias towards high uncertainty in heavyweight subjects. Conclusion: Mean-variance regression and ensembling provided complementary benefits for automated body composition measurements from UK Biobank MRI, reaching high speed and accuracy. These values were inferred for the entire cohort, with uncertainty estimates that can approximate the measurement errors and identify some of the worst outliers automatically.

MIMIR: Deep Regression for Automated Analysis of UK Biobank Body MRI

“…Out-of-domain tasks would likely require hundreds of subjects for renewed training [11]. However, robust performance within UKB can likely be expected, as prior work saw similar systems match or exceed the cross-validation performance on withheld UKB subjects [15]. The predicted uncertainties underestimate the prediction errors, especially since no ensembling was used.…”

“…As a result, it can provide both a point estimate µ for the measurement itself and a heteroscedastic variance σ 2 , which can expresses aleatoric uncertainty and describe confidence-or prediction intervals [14]. This methodology was previously proposed for six measurements of body composition on the same data [15]. Here, neural network instances were not formed into ensembles, however, in favor of speed and convenience.…”

“…Neck-to-knee body MRI was shared by UKB for about 40,000 men and women aged 44-82 (mean 63) years, BMI 14-62 (mean 27) kg/m 2 and 95% self-reported white British ethnicity (see also [1], [2]). Following visual inspection for exclusion of artifacts, severe pathologies, and other anomalies, 38,916 subjects remained for the experiments [15].…”

“…Additionally, the coefficient of determination (R 2 ), mean absolute error (MAE), mean absolute percentage error (MAPE) and the area under curve of the receiver operating characteristic curve (AUC-ROC) are provided. The uncertainty estimates are known to often underestimate the true prediction errors [19], and were therefore calibrated post-hoc with scaling factors determined on the validation data [15].…”

Deep regression for uncertainty-aware and interpretable analysis of large-scale body MRI

“…In this work, regression with convolutional neural networks on UK Biobank neck-to-knee body MRI is examined, with a closer look at recently presented approaches for estimates of age [7], general biometry [8], liver fat [9], and body composition [10]. Whereas these publications provide extensive medical context, this submission aims to provide an overview and motivate the underlying design decisions with ablation experiments and further discussion of a more technical nature.…”