arXiv subjects

Tobias Becher

Publications and source records attributed to Tobias Becher.

11 recordsLinked to original sources

arXiv2026

GPU-accelerated superiorization on constrained physical problems with SupPy

The superiorization method (SM) is situated between feasibility-seeking and constrained optimization. Instead of aiming at the minimum of a given objective function over a constraint set, it seeks a feasible point at which the objective function value is reduced - though not necessarily minimal - rather than hard targets, or in which a mathematically optimal solution is not strictly required. While the method has been investigated for several applications in physics, its broader use has been limited, in part due to the lack of openly available software for researchers wishing to explore it. In this work we apply superiorization to three problems from applied physics: seismic image reconstruction, low-dose CT reconstruction and intensity-modulated radiotherapy treatment planning. These experiments are conducted with SupPy, an open-source modularized Python toolbox developed for this work, which supports execution of feasibility-seeking algorithms and their superiorized version on both the CPU and the GPU. In all three cases the superiorized algorithms achieve favorable results compared to feasibility-seeking alone, with reduced noise in the imaging examples and lowered body dose in the radiotherapy plans. For the radiotherapy case we further observe that superiorization produces clinically viable plans on infeasible constraint sets.

physics.comp-ph↗

arXiv2026

Tipping the Balance: Impact of Class Imbalance Correction on the Performance of Clinical Risk Prediction Models

Objective: ML-based clinical risk prediction models are increasingly used to support decision-making in healthcare. While class-imbalance correction techniques are commonly applied to improve model performance in settings with rare outcomes, their impact on probabilistic calibration remains insufficiently understood. This study evaluated the effect of widely used resampling strategies on both discrimination and calibration across real-world clinical prediction tasks. Methods: Ten clinical datasets spanning diverse medical domains and including 605,842 patients were analyzed. Multiple machine-learning model families, including linear models and several non-linear approaches, were evaluated. Models were trained on the original data and under three commonly used 1:1 class-imbalance correction strategies (SMOTE, RUS, ROS). Performance was assessed on held-out data using discrimination and calibration metrics. Results: Across all datasets and model families, resampling had no positive impact on predictive performance. Changes in the Receiver Operating Characteristic Area Under Curve (ROC-AUC) relative to models trained on the original data were small and inconsistent (ROS: -0.002, p<0.05; RUS: -0.004, p>0.05; SMOTE: -0.01, p<0.05), with no resampling strategy demonstrating a systematic improvement. In contrast, resampling in general degraded the calibration performance. Models trained using imbalance correction exhibited higher Brier scores (0.029 to 0.080, p<0.05), reflecting poorer probabilistic accuracy, and marked deviations in calibration intercept and slope, indicating systematic distortions of predicted risk despite preserved rank-based performance. Conclusion: In a diverse set of real-world clinical prediction tasks, commonly used class-imbalance correction techniques did not provide generalizable improvements in discrimination and were associated with degraded calibration.

q-bio.QM↗

arXiv2025

Trading robustness: a scenario-free approach to robust Multi-Criteria Optimization for Treatment Planning

Treatment planning in radiotherapy is inherently a multi-criteria optimization (MCO) problem. Traditionally, the treatment's robustness is not formulated as a part of this decision making problem, but dealt with separately through margins or robust optimization. This work facilitates integration of robustness into multi-criteria optimization using a recently proposed efficient scenario-free (s-f) robust optimization approach: The s-f approach relies on the fast evaluation of the expected dose distribution and mean variance during optimization. This is achieved by precomputation of probabilistic quantities, which can then be used for repeated solving of subproblems in the two explored MCO approaches: Lexicographic Ordering (LO) and Pareto Front (PF) approximation. Different prioritization strategies within the LO approach are used to assess the impact of variance reduction while a 3-objective PF approximation, including a variance reduction objective, is generated to visualize and analyze trade-offs between the competing objectives. The robust optimization is performed including 100 scenarios modeling setup and range errors, as well as organ motion, on 3D- and 4DCT lung cancer patient datasets. Robustness analysis is performed to assess and explore the efficacy of all optimization strategies. The s-f approach enabled robust optimization in MCO with computational times comparable to nominal MCO. Both MCO strategies highlighted the interplay between dosimetric and variance reduction objectives. The LO approach showed how prioritization affects plan quality and robustness, while the PF analysis revealed a clear trade-off between robustness and organ-at-risk sparing. The reported analysis highlighted the conflicting trade-off nature of plan robustness and dosimetric quality, demonstrating how robust MCO supports a more informed and flexible decision-making process in treatment planning.

physics.med-ph↗

arXiv2025

GPU-accelerated FREDopt package for simultaneous dose and LETd proton radiotherapy plan optimization via superiorization methods

This study presents FREDopt, a newly developed GPU-accelerated open-source optimization software for simultaneous proton dose and dose-averaged LET (LETd) optimization in IMPT treatment planning. FREDopt was implemented entirely in Python, leveraging CuPy for GPU acceleration and incorporating fast Monte Carlo (MC) simulations from the FRED code. The treatment plan optimization workflow includes pre-optimization and optimization, the latter equipped with a novel superiorization of feasibility-seeking algorithms. Feasibility-seeking requires finding a point that satisfies prescribed constraints. Superiorization interlaces computational perturbations into iterative feasibility-seeking steps to steer them toward a superior feasible point, replacing the need for costly full-fledged constrained optimization. The method was validated on two treatment plans of patients treated in a clinical proton therapy center, with dose and LETd distributions compared before and after reoptimization. Simultaneous dose and LETd optimization using FREDopt led to a substantial reduction of LETd and (dose)x(LETd) in organs at risk (OARs) while preserving target dose conformity. Computational performance evaluation showed execution times of 14-50 minutes, depending on the algorithm and target volume size-satisfactory for clinical and research applications while enabling further development of the well-tested, documented open-source software.

physics.med-ph↗

arXiv2025

Anatomically Informed GREIT Reconstruction: Improving EIT Imaging for Lung Monitoring

Objective: Time-difference electrical impedance tomography (EIT) is gaining widespread use for bedside lung monitoring in intensive care patients suffering from lung-related diseases. It involves collecting voltage measurements from electrodes placed on the patient's thorax, which are then used to reconstruct impedance images. This study investigates how incorporating anatomical information from CT data into the widely used GREIT reconstruction algorithm affects EIT images and improves their interpretability. Approach: Based on clinically motivated lung state scenarios, we simulated EIT measurements to assess how the GREIT parameters influence the result of EIT image reconstruction, particularly with respect to noise performance and image accuracy. We introduce quality measures that allow us to perform a quantitative assessment of reconstruction quality. Anatomical features from CT data were included by customizing background conductivity and GREIT training target distribution. Main results: Our analysis confirmed that unphysiological background conductivity assumptions can lead to misleading EIT images, whereas physiological values, although more accurate, come with higher noise sensitivity. By increasing the number of GREIT training targets inside the lung and adapting the respective weighting radius, we significantly improved the anatomical accuracy of the EIT images. When applied to clinical EIT data from a representative ARDS patient, these adjustments in the reconstruction setup substantially enhanced the interpretability of the resulting EIT images. Significance: Integrating CT-based anatomical data in the GREIT reconstruction significantly enhances the clinical applicability of EIT in lung monitoring. The improved interpretability of EIT images facilitates better-informed clinical decisions and the individualized adjustment of ventilation strategies for critically ill patients.

physics.med-ph↗

arXiv2024

Exploring AI-Enabled Cybersecurity Frameworks: Deep-Learning Techniques, GPU Support, and Future Enhancements

Traditional rule-based cybersecurity systems have proven highly effective against known malware threats. However, they face challenges in detecting novel threats. To address this issue, emerging cybersecurity systems are incorporating AI techniques, specifically deep-learning algorithms, to enhance their ability to detect incidents, analyze alerts, and respond to events. While these techniques offer a promising approach to combating dynamic security threats, they often require significant computational resources. Therefore, frameworks that incorporate AI-based cybersecurity mechanisms need to support the use of GPUs to ensure optimal performance. Many cybersecurity framework vendors do not provide sufficiently detailed information about their implementation, making it difficult to assess the techniques employed and their effectiveness. This study aims to overcome this limitation by providing an overview of the most used cybersecurity frameworks that utilize AI techniques, specifically focusing on frameworks that provide comprehensive information about their implementation. Our primary objective is to identify the deep-learning techniques employed by these frameworks and evaluate their support for GPU acceleration. We have identified a total of \emph{two} deep-learning algorithms that are utilized by \emph{three} out of 38 selected cybersecurity frameworks. Our findings aim to assist in selecting open-source cybersecurity frameworks for future research and assessing any discrepancies between deep-learning techniques used in theory and practice.

cs.CR↗

arXiv2024

Patient-specific prediction of regional lung mechanics in ARDS patients with physics-based models: a validation study

The choice of lung protective ventilation settings for mechanical ventilation has a considerable impact on patient outcome, yet identifying optimal ventilatory settings for individual patients remains highly challenging due to the inherent inter- and intra-patient pathophysiological variability. In this validation study, we demonstrate that physics-based computational lung models tailored to individual patients can resolve this variability, allowing us to predict the otherwise unknown local state of the pathologically affected lung during mechanical ventilation. For seven ARDS patients undergoing invasive mechanical ventilation, physics-based, patient-specific lung models were created using chest CT scans and ventilatory data. By numerically resolving the interaction of the pathological lung with the airway pressure and flow imparted by the ventilator, we predict the time-dependent and heterogeneous local state of the lung for each patient and compare it against the regional ventilation obtained from bedside monitoring using Electrical Impedance Tomography. Excellent agreement between numerical simulations and experimental data was obtained, with the model-predicted anteroposterior ventilation profile achieving a Pearson correlation of 96% with the clinical reference data. Even when considering the regional ventilation within the entire transverse chest cross-section and across the entire dynamic ventilation range, an average correlation of more than 81% and an average root mean square error of less than 15% were achieved. The results of this first systematic validation study demonstrate the ability of computational models to provide clinically relevant information and thereby open the door for a truly patient-specific choice of ventilator settings on the basis of both individual anatomy and pathophysiology.

physics.med-ph↗

arXiv2023

Pressure- and time-dependent alveolar recruitment/derecruitment in a spatially resolved patient-specific computational model for injured human lungs

We present a novel computational model for the dynamics of alveolar recruitment/derecruitment (RD), which reproduces the underlying characteristics typically observed in injured lungs. The basic idea is a pressure- and time-dependent variation of the stress-free reference volume in reduced dimensional viscoelastic elements representing the acinar tissue. We choose a variable reference volume triggered by critical opening and closing pressures in a time-dependent manner from a straightforward mechanical point of view. In the case of (partially and progressively) collapsing alveolar structures, the volume available for expansion during breathing reduces and vice versa, eventually enabling consideration of alveolar collapse and reopening in our model. We further introduce a method for patient-specific determination of the underlying critical parameters of the new alveolar RD dynamics when integrated into the tissue elements, referred to as terminal units, of a spatially resolved physics-based lung model that simulates the human respiratory system in an anatomically correct manner. Relevant patient-specific parameters of the terminal units are herein determined based on medical image data and the macromechanical behavior of the lung during artificial ventilation. We test the whole modeling approach for a real-life scenario by applying it to the clinical data of a mechanically ventilated patient. The generated lung model is capable of reproducing clinical measurements such as tidal volume and pleural pressure during various ventilation maneuvers. We conclude that this new model is an important step toward personalized treatment of ARDS patients by considering potentially harmful mechanisms - such as cyclic RD and overdistension - and might help in the development of relevant protective ventilation strategies to reduce ventilator-induced lung injury (VILI).

physics.med-ph↗

arXiv2023

An approach to study recruitment/derecruitment dynamics in a patient-specific computational model of an injured human lung

We present a new approach for physics-based computational modeling of diseased human lungs. Our main object is the development of a model that takes the novel step of incorporating the dynamics of airway recruitment/de-recruitment into an anatomically accurate, spatially resolved model of respiratory system mechanics, and the relation of these dynamics to airway dimensions and the biophysical properties of the lining fluid. The importance of our approach is that it potentially allows for more accurate predictions of where mechanical stress foci arise in the lungs, since it is at these locations that injury is thought to arise and propagate from. We match the model to data from a patient with Acute Respiratory Distress Syndrome (ARDS) to demonstrate the potential of the model for revealing the underlying derangements in ARDS in a patient-specific manner. To achieve this, the specific geometry of the lung and its heterogeneous pattern of injury are extracted from medical CT images. The mechanical behavior of the model is tailored to the patient's respiratory mechanics using measured ventilation data. In retrospective simulations of various clinically performed, pressure-driven ventilation profiles, the model adequately reproduces clinical quantities measured in the patient such as tidal volume and change in pleural pressure. The model also exhibits physiologically reasonable lung recruitment dynamics and has the spatial resolution to allow the study of local mechanical quantities such as alveolar strains. This modeling approach advances our ability to perform patient-specific studies in silico, opening the way to personalized therapies that will optimize patient outcomes.

cs.CE↗

arXiv2020

Inferring respiratory and circulatory parameters from electrical impedance tomography with deep recurrent models

Electrical impedance tomography (EIT) is a noninvasive imaging modality that allows a continuous assessment of changes in regional bioimpedance of different organs. One of its most common biomedical applications is monitoring regional ventilation distribution in critically ill patients treated in intensive care units. In this work, we put forward a proof-of-principle study that demonstrates how one can reconstruct synchronously measured respiratory or circulatory parameters from the EIT image sequence using a deep learning model trained in an end-to-end fashion. We demonstrate that one can accurately infer absolute volume, absolute flow, normalized airway pressure and within certain limitations even the normalized arterial blood pressure from the EIT signal alone, in a way that generalizes to unseen patients without prior calibration. As an outlook with direct clinical relevance, we furthermore demonstrate the feasibility of reconstructing the absolute transpulmonary pressure from a combination of EIT and absolute airway pressure, as a way to potentially replace the invasive measurement of esophageal pressure. With these results, we hope to stimulate further studies building on the framework put forward in this work.

eess.IV↗

arXiv2018

A parametric model for the changes in the complex valued conductivity of a lung during tidal breathing

Classical homogenization theory based on the Hashin-Shtrikman coated ellipsoids is used to model the changes in the complex valued conductivity (or admittivity) of a lung during tidal breathing. Here, the lung is modeled as a two-phase composite material where the alveolar air-filling corresponds to the inclusion phase. The theory predicts a linear relationship between the real and the imaginary parts of the change in the complex valued conductivity of a lung during tidal breathing, and where the loss cotangent of the change is approximately the same as of the effective background conductivity and hence easy to estimate. The theory is illustrated with numerical examples, as well as by using reconstructed Electrical Impedance Tomography (EIT) images based on clinical data from an ongoing study within the EU-funded CRADL project. The theory may be potentially useful for improving the imaging algorithms and clinical evaluations in connection with lung EIT for respiratory management and monitoring in neonatal intensive care units.

physics.bio-ph↗