Eike Petersen

Eike Petersen

Auto-generated profile, based on publicly available information

Eike Petersen

Technical University of Denmark

Research Focus & Metrics

Academic Career Length

11Years (2016-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
27 May 2022 in IEEE Access
Verified
Journal Article
Open access
EPEike Petersen
Eike Petersen
YPYannik Potdevin
Yannik Potdevin
EMEsfandiar Mohammadi
Esfandiar Mohammadi
SZStephan Zidowitz
Stephan Zidowitz
SBSabrina Breyer
Sabrina Breyer
CHChristian Herzog
Christian Herzog
Machine learning is expected to fuel significant improvements in medical care. To ensure that fundamental principles such as beneficence, respect for human autonomy, prevention of harm, justice, privacy, and transparency are respected, medical machine learning systems must be developed responsibly. Many high-level declarations of ethical principles have been put forth for this purpose, but there is a severe lack of technical guidelines explicating the practical consequences for medical machine learning. Similarly, there is currently considerable uncertainty regarding the exact regulatory requirements placed upon medical machine learning systems. This survey provides an overview of the technical and procedural challenges involved in creating medical machine learning systems responsibly and in conformity with existing regulations, as well as possible solutions to address these challenges. First, a brief review of existing regulations affecting medical machine learning is provided, showing that properties such as safety, robustness, reliability, privacy, security, transparency, explainability, and nondiscrimination are all demanded already by existing law and regulations—albeit, in many cases, to an uncertain degree. Next, the key technical obstacles to achieving these desirable properties are discussed, as well as important techniques to overcome these obstacles in the medical context. We notice that distribution shift, spurious correlations, model underspecification, uncertainty quantification, and data scarcity represent severe challenges in the medical context. Promising solution approaches include the use of large and representative datasets and federated learning as a means to that end, the careful exploitation of domain knowledge, the use of inherently transparent models, comprehensive out-of-distribution model testing and verification, as well as algorithmic impact assessments.

14 July 2023 in Patterns
Verified
Journal Article
Open access
EPEike Petersen
Eike Petersen
SHSune Holm
Sune Holm
MGMelanie Ganz
Melanie Ganz
AFAasa Feragen
Aasa Feragen
To ensure equitable quality of care, differences in machine learning model performance between patient groups must be addressed. Here, we argue that two separate mechanisms can cause performance differences between groups. First, model performance may be worse than theoretically achievable in a given group. This can occur due to a combination of group underrepresentation, modeling choices, and the characteristics of the prediction task at hand. We examine scenarios in which underrepresentation leads to underperformance, scenarios in which it does not, and the differences between them. Second, the optimal achievable performance may also differ between groups due to differences in the intrinsic difficulty of the prediction task. We discuss several possible causes of such differences in task difficulty. In addition, challenges such as label biases and selection biases may confound both learning and performance evaluation. We highlight consequences for the path toward equal performance, and we emphasize that leveling up model performance may require gathering not only more data from underperforming groups but also better data. Throughout, we ground our discussion in real-world medical phenomena and case studies while also referencing relevant statistical theory.

8 March 2019 in Frontiers in Physiology
Verified
Journal Article
Open access
PDF file
EPEike Petersen
Eike Petersen
PRPhilipp Rostalski
Philipp Rostalski
Neuromuscular physiology is a vibrant research field that has recently seen exciting advances. Previous publications have focused on thorough analyses of particular aspects of neuromuscular physiology, yet an integration of the various novel findings into a single, comprehensive model is missing. In this article, we provide a unified description of a comprehensive mathematical model of surface electromyographic (EMG) measurements and the corresponding force signal in skeletal muscles, both consolidating and extending the results of previous studies regarding various components of the neuromuscular system. The model comprises motor unit (MU) pool organization, recruitment and rate coding, intracellular action potential generation and the resulting EMG measurements, as well as the generated muscular force during voluntary isometric contractions. Mathematically, it consists of a large number of linear PDEs, ODEs, and various stochastic nonlinear relationships, some of which are solved analytically, others numerically. A parameterization of the electrical and mechanical components of the model is proposed that ensures a physiologically meaningful EMG-force relation in the simulated signals, in particular taking the continuous, size-dependent distribution of MU parameters into account. Moreover, a novel nonlinear transformation of the common drive model input is proposed, which ensures that the model force output equals the desired target force. On a physiological level, this corresponds to adjusting the rate coding model to the force generating capabilities of the simulated muscle, while from a control theoretic point of view, this step is equivalent to an exact linearizing transformation of the controlled neuromuscular system. Finally, an alternative analytical formulation of the EMG model is proposed, which renders the physiological meaning of the model more clear and facilitates a mathematical proof that muscle fibers in this model at no point in time represent a net current source or sink. A consistent description of a complete physiological model as presented here, including thorough justification of model component choices, will facilitate the use of these advanced models in future research. Results of a numerical simulation highlight the model's capability to reproduce many physiological effects observed in experimental measurements, and to produce realistic synthetic data that are useful for the validation of signal processing algorithms.

15 September 2022
Verified
Book Chapter
Closed access
PDF file
EPEike Petersen
Eike Petersen
AFAasa Feragen
Aasa Feragen
MZMaria Luise da Costa Zemsch
Maria Luise da Costa Zemsch
AHAnders Henriksen
Anders Henriksen
OCOskar Eiler Wiese Christensen
Oskar Eiler Wiese Christensen
FIFor The Alzheimer’s Disease Neuroimaging Initiative
For The Alzheimer’s Disease Neuroimaging Initiative

10 February 2020 in IEEE Access
Verified
Journal Article
Open access
EPEike Petersen
Eike Petersen
JSJulia Sauer
Julia Sauer
JGJan Graßhoff
Jan Graßhoff
PRPhilipp Rostalski
Philipp Rostalski
Electromyographic (EMG) measurements of the respiratory muscles provide a convenient and noninvasive way to assess respiratory muscle function and detect patient activity during assisted mechanical ventilation. However, surface EMG measurements of the diaphragm and intercostal muscles are substantially contaminated by cardiac activity due to the vicinity of the cardiac muscles. Many algorithmic solutions to this problem have been proposed, yet a conclusive performance comparison of the most promising candidates currently is missing. The objective of this work is to provide a quantitative performance comparison of six previously proposed cardiac artifact removal algorithms operating on singlechannel EMG measurements, and two newly proposed, improved versions of these algorithms. Algorithmic performance is evaluated quantitatively based on four different measures of separation success, using both synthetic validation signals and electromyographic measurements of the respiratory muscles in eight subjects. The compared algorithms are two versions of the empirical template subtraction algorithm, two model-based Bayesian filtering procedures, a wavelet denoising approach, an empirical mode decomposition (EMD) based approach, and classical high-pass filtering. Different algorithms perform well with respect to different performance measures. Template subtraction algorithms yield the best results if the characteristics of the raw signal are of interest, while filtering algorithms like simple high-pass filtering, wavelet denoising, and EMD-based denoising show superior performance for calculating a cleaned envelope signal. No algorithm completely removes the cardiac interference, but the characteristic errors introduced by the considered algorithms differ. Hence, the choice of the algorithm to use should be made depending on the target application. Finally, we also demonstrate that our empirical SNR measure, which can be calculated without knowledge of the true, undisturbed signals, correlates strongly with the exact reconstruction error. Thus, it represents a reliable indicator for algorithm performance on real measurement data.

12 June 2023
Verified
Conference Paper
Open access
EPEike Petersen
Eike Petersen
MGMelanie Ganz
Melanie Ganz
SHSune Holm
Sune Holm
AFAasa Feragen
Aasa Feragen
Recent work on algorithmic fairness has largely focused on the fairness of discrete decisions, or classifications. While such decisions are often based on risk score models, the fairness of the risk models themselves has received considerably less attention. Risk models are of interest for a number of reasons, including the fact that they communicate uncertainty about the potential outcomes to users, thus representing a way to enable meaningful human oversight. Here, we address fairness desiderata for risk score models. We identify the provision of similar epistemic value to different groups as a key desideratum for risk score fairness, and we show how even fair risk scores can lead to unfair risk-based rankings. Further, we address how to assess the fairness of risk score models quantitatively, including a discussion of metric choices and meaningful statistical comparisons between groups. In this context, we also introduce a novel calibration error metric that is less sample size-biased than previously proposed metrics, enabling meaningful comparisons between groups of different sizes. We illustrate our methodology – which is widely applicable in many other settings – in two case studies, one in recidivism risk prediction, and one in risk of major depressive disorder (MDD) prediction.