MRUNet-3D: A multi-stride residual 3D UNet for lung nodule segmentation.

Obtaining an accurate segmentation of the pulmonary nodules in computed tomography (CT) images is challenging. This is due to: (1) the heterogeneous nature of the lung nodules; (2) comparable visual characteristics between the nodules and their surroundings. A robust multi-scale feature extraction mechanism that can effectively obtain multi-scale representations at a granular level can improve segmentation accuracy. As the most commonly used network in lung nodule segmentation, UNet, its variants, and other image segmentation methods lack this robust feature extraction mechanism. In this study, we propose a multi-stride residual 3D UNet (MRUNet-3D) to improve the segmentation accuracy of lung nodules in CT images. It incorporates a multi-slide Res2Net block (MSR), which replaces the simple sequence of convolution layers in each encoder stage to effectively extract multi-scale features at a granular level from different receptive fields and resolutions while conserving the strengths of 3D UNet. The proposed method has been extensively evaluated on the publicly available LUNA16 dataset. Experimental results show that it achieves competitive segmentation performance with an average dice similarity coefficient of 83.47 % and an average surface distance of 0.35 mm on the dataset. More notably, our method has proven to be robust to the heterogeneity of lung nodules. It has also proven to perform better at segmenting small lung nodules. Ablation studies have shown that the proposed MSR and RFIA modules are fundamental to improving the performance of the proposed model.

On the variability and dependence of human leg stiffness across strides during running and some consequences for the analysis of locomotion data.

Typically, animal locomotion studies involve consecutive strides, which are frequently assumed to be independent with parameters that do not vary across strides. This assumption is often not tested. However, failing in particular to account for dependence across strides may cause an incorrect estimate of the uncertainty of the measurements and thereby lead to either missing (overestimating variance) or over-evaluating (underestimating variance) biological signals. In turn, this impacts replicability of the results because variability is accounted for differently across experiments. In this paper, we analyse the changes of a couple of measures of human leg stiffness across strides during running experiments, using a publicly available dataset. A major finding of this analysis is that the time series of these measurements of stiffness show autocorrelation even at large lags and so there is dependence between individual strides, even when separated by many intervening strides. Our results question the practice in biomechanics research of using each stride as an independent observation or of sub-selecting strides at small lags. Following the outcome of our analysis, we strongly recommend caution in doing so without first confirming the independence of the measurements across strides and without confirming that sub-selection does not produce spurious results.