The risk of bias in denoising methods: Examples from neuroimaging

Abstract

Experimental datasets are growing rapidly in size, scope, and detail, but the value of these datasets is limited by unwanted measurement noise. It is therefore tempting to apply analysis techniques that attempt to reduce noise and enhance signals of interest. In this paper, we draw attention to the possibility that denoising methods may introduce bias and lead to incorrect scientific inferences. To present our case, we first review the basic statistical concepts of bias and variance. Denoising techniques typically reduce variance observed across repeated measurements, but this can come at the expense of introducing bias to the average expected outcome. We then conduct three simple simulations that provide concrete examples of how bias may manifest in everyday situations. These simulations reveal several findings that may be surprising and counterintuitive: (i) different methods can be equally effective at reducing variance but some incur bias while others do not, (ii) identifying methods that better recover ground truth does not guarantee the absence of bias, (iii) bias can arise even if one has specific knowledge of properties of the signal of interest. We suggest that researchers should consider and possibly quantify bias before deploying denoising methods on important research data.

Robust high spatio-temporal line-scanning fMRI in humans at 7T using multi-echo readouts, denoising and prospective motion correction

“…T-stats increased after denoising, with respect to the standard reconstruction (no denoising), for every participant. The "hard-thresholding" we used in the adapted NORDIC denoising is very liberal (due to the large number of components removed from the data), hence, as for any denoising techniques, one should be wary of potential biases that can be introduced (Kay, 2022;Vizioli et al, 2021). This is particularly important when less strong stimuli (leading to smaller responses) are used.…”

Improving the accuracy of single-trial fMRI response estimates using GLMsingle

“…It is important to consider whether denoising comes at the potential cost of introducing bias ( Kay, 2022 ). Considering each component of GLMsingle, we believe that the risk of bias is minimal for most use cases.…”

Improved Quantitative Parameter Estimation for Prostate T2 Relaxometry using Convolutional Neural Networks

“…The median value of the absolute error was interpreted as a measure of overall accuracy , which is dependent on both bias and precision, with smaller errors indicating higher accuracy. Decomposing accuracy into separate contributions of bias and precision is important for understand potential sources of bias in quantitative analyses (49,50).…”