As the CMIV researchers are also affiliated to a home department at Linköping University or another university and their research is primarily registered there it can be difficult to overview. Here you will find a selection of the latest publications registered in the DiVA database.
Publications Show/Hide content
Recent publications
28 Nov, 2022
Predictive uncertainty estimation for out-of-distribution detection in digital pathology
Machine learning model deployment in clinical practice demands real-time risk assessment to identify situations in which the model is uncertain. Once deployed, models should be accurate for classes seen during training while providing informative estimates of uncertainty to flag abnormalities and unseen classes for further analysis. Although recent developments in uncertainty estimation have resulted in an increasing number of methods, a rigorous empirical evaluation of their performance on large-scale digital pathology datasets is lacking. This work provides a benchmark for evaluating prevalent methods on multiple datasets by comparing the uncertainty estimates on both in-distribution and realistic near and far out-of-distribution (OOD) data on a whole-slide level. To this end, we aggregate uncertainty values from patch-based classifiers to whole-slide level uncertainty scores. We show that results found in classical computer vision benchmarks do not always translate to the medical imaging setting. Specifically, we demonstrate that deep ensembles perform best at detecting far-OOD data but can be outperformed on a more challenging near-OOD detection task by multi-head ensembles trained for optimal ensemble diversity. Furthermore, we demonstrate the harmful impact OOD data can have on the performance of deployed machine learning models. Overall, we show that uncertainty estimates can be used to discriminate in-distribution from OOD data with high AUC scores. Still, model deployment might require careful tuning based on prior knowledge of prospective OOD data.
Publication in DiVA
06 Dec, 2021
Affective and discriminative touch: a reappraisal
The role of CT afferents in affective touch is often viewed in contrast with the more well-established neural and functional systems supporting discriminative touch. However, a recent groundswell of evidence suggests that a categorical affective versus-discriminative contrast may not bear scrutiny at all levels of the nervous system, especially when applied to finer grained anatomical and functional relationships. Discrepancies in this evidence can be addressed by taking the layered phylogenetic history of specialized afferent systems into account, and how this history may have influenced functional integration within complex spinal circuits and brain networks in generating bodily states and behavior. This perspective inspires four proposed body-behavior reference frames, within which somatosensory-behavior relationships can be schematized in the nervous system of the behaving human: (1) proximal-distal (regarding the body axis and limbs, (2) somaticskeletomotor (regarding efferent effectors) (3) reactive predictive (regarding responses to external events); and (4) passive/receptive-active/motivated (with particular application to socially interactive behavior). Affective and discriminative functions can be dissociated at the extremes of these frame spaces without necessarily existing as discrete categories.
Publication in DiVA
11 Jan, 2022
The direction-dependence of apparent water exchange rate in human white matter
Transmembrane water exchange is a potential biomarker in the diagnosis and understanding of cancers, brain disorders, and other diseases. Filter-exchange imaging (FEXI), a special case of diffusion exchange spectroscopy adapted for clinical applications, has the potential to reveal different physiological water exchange processes. However, it is still controversial whether modulating the diffusion encoding gradient direction can affect the apparent exchange rate (AXR) measurements of FEXI in white matter (WM) where water diffusion shows strong anisotropy. In this study, we explored the diffusion-encoding direction dependence of FEXI in human brain white matter by performing FEXI with 20 diffusion-encoding directions on a clinical 3T scanner in-vivo. The results show that the AXR values measured when the gradients are perpendicular to the fiber orientation (0.77 +/- 0.13 s - 1 , mean +/- standard deviation of all the subjects) are significantly larger than the AXR estimates when the gradients are parallel to the fiber orientation (0.33 +/- 0.14 s - 1 , p < 0.001) in WM voxels with coherently-orientated fibers. In addition, no significant correlation is found between AXRs measured along these two directions, indicating that they are measuring different water exchange processes. Whats more, only the perpendicular AXR rather than the parallel AXR shows dependence on axonal diameter, indicating that the perpendicular AXR might reflect transmembrane water exchange between intra-axonal and extra-cellular spaces. Further finite difference (FD) simulations having three water compartments (intra-axonal, intra-glial, and extra-cellular spaces) to mimic WM micro-environments also suggest that the perpendicular AXR is more sensitive to the axonal water transmembrane exchange than parallel AXR. Taken together, our results show that AXR measured along different directions could be utilized to probe different water exchange processes in WM.
Publication in DiVA