Scientific publications

We present a new data augmentation method to address the data scarcity problem in deep learning based automatic segmentation of liver tumors. A non-rigid registration algorithm was used to generate new anatomical variations of liver tumors from a publicly available benchmark dataset. Additionally, a conditional image-to-image generative adversarial network (GAN) was trained to translate the generated segmentation masks into corresponding textured CT volumes. We used a state-of the-art segmentation model to investigate the benefit of our data augmentation method. Experiments with varying amounts of synthetic data were conducted and the results show that our novel augmentation method improves tumor segmentation performance by approximately 3 %, outperforming similar data augmentation techniques.

We investigate the impact of melanin concentration on the accuracy of oxygen saturation SpO2 estimation using reflective photoplethysmography (PPG) measurements. Monte Carlo (MC) simulations were used to analyse photon-tissue interactions, and detection of reflected light during systole and diastole. We model a reflective pulse oximeter and analyse the perfusion index (PI), to represent the signal change due varying blood flow during systole and diastole. Moreover, we derive the ratio-of-ratios (RoR), to represent the absorption difference between PIs of different light wavelengths, and SpO2 depending on skin tone. We derive calibration models for SpO2 estimation based on MC simulations tailored to specific melanin concentrations ranging from 2.55% to 30.5%, as well as a generalised population model, to study reflective pulse oximeter performance. While current pulse oximeters are often calibrated for low melanin concentrations, our results show that calibrating for an appropriate skin tone range substantially increases the pulse oximeter performance. Skin tone-adapted calibration reduced the root mean square error Arms of SpO2 estimation by 44% compared to a generalised population model. SpO2 estimation error was minimal at a distance of 4 mm to 5 mm between light source and photodetector of a reflective pulse oximeter. We conclude that skin tone-adapted calibration should be applied to make reflective pulse oximeter applicable to any human, independent of skin pigmentation.

https://ieeexplore.ieee.org/document/10870095

Goal: We investigate the effect of source-detector geometry, including distance and angle, on the reflective photoplethysmography (PPG) signal. Methods: A porcine skin phantom was used for laboratory measurements and replicated by Monte Carlo simulations. Variations in sensor geometry were analysed. Results: Laboratory measurements and Monte Carlo simulations showed agreement for various geometry settings. With decreasing negative sensor angle, the differential path length factor and the average maximum penetration depth increases. Conclusions: Our analyses highlight the influence of source-detector geometry on the PPG DC signal. Based on our analysis of penetration depth and optical path length, the geometry effects can be transferred to the PPG AC signal too. MC simulations provide an important tool to optimise PPG performance.

https://ieeexplore.ieee.org/document/10908070

The combination of shape memory alloys and micro-electro-mechanical systems possesses to rethink and downscale the size of actuator systems to MEMS level. Thus, the heterogeneous integration of shape memory alloy actuators is a focus of current research to enable the use of standardized semi-finished products such as wires and to exploit their good actuator properties. To integrate the actuator into the mechanical structure, it must first be deformed to fit into the intended installation space and to realize a form-fit connection after the first activation. This pre-deformation has a significant influence on the actuator performance. In this paper, the influence of preforming on the resulting actuator behaviour will be investigated. For this purpose, the change in the transformation temperatures and the resulting stroke are investigated. The experiments carried out were based on experimental plans that were created by means of design of experiments.

 

https://www.vde-verlag.de/proceedings-de/456391023.html

In conventional hydrogen pressure sensors, the stress-sensitive diaphragm is the weakest point in terms of construction. Leading to signal drift of the sensor as hydrogen causes material embrittlement and permeation. We address this issue by using a miniaturized gas pressure sensor without a me-chanical transducer. Our sensor structure consists of a heating element and thermopiles arranged in a thin perforated silicon nitride membrane. In the thermal domain, the structure exhibits a characteristic low-pass behavior whose phase shift is dependent on the hydrogen pressure. A dynamically excited heating element and the evaluation of the thermopile responses are used to measure pressures rang-ing from 100 and 2850 kPa. So far, this is the highest measurement range detected with a thermal pressure sensor that has ever been published in literature. Compared to diaphragm-based MEMS pressure sensors, this chip design offers an intrinsic overload protection as well as low-cost fabrication and simple packaging requirements.

 

https://www.researchgate.net/profile/S-Billat/publication/381631502_73_-_Novel_Miniaturized_Thermoelectric_Hydrogen_Pressure_Sensor/links/66ac8da88f7e1236bc2eda7b/73-Novel-Miniaturized-Thermoelectric-Hydrogen-Pressure-Sensor.pdf

The production of 10 nm thin silicon nitride (SiNx) membranes for solid-state nanopore sensors is a challenge, as numerous factors have to be taken into account. Instead of etching thicker membranes, research is focussing on the direct deposition of thin SiNx films using a modified LPCVD process at lower temperature. A SiO₂ underlayer reduces parasitic capacitance and compensates for stresses. KOH wet etching is used for membrane structuring, whereby protection is provided by a layer combination of PECVD-SiO₂ and LPCVD-SiNx.

Technical poster

We investigate combined, personalised biomechanical dynamics models and human surface models to synthesise IMU sensor time series data and to improve Human Activity Recognition (HAR) model performance for activities of daily living (ADLs). We analyse two model training scenarios: (1) data fusion of synthetic and measurement IMU data to train HAR models directly, and (2) pretraining HAR models with synthetic and measured IMU data and subsequent transfer learning with public benchmark datasets. Furthermore, we analyse how the synthetic IMU data helps in configurations with scarce measurement data by limiting the number of participants and IMUs in both training scenarios. We evaluate three state-of-the-art HAR models to determine the benefit of our approach. Our results show that the IMU data synthesis approach improves performance across all HAR models in both training scenarios. Depending on the HAR model, synthetic data increased the macroF1 score on average by 8% for configurations with reduced data and by up to 7.5% for transfer learning. In the transfer learning scenario, combining synthetic data with measurement data during pretraining outperformed the results obtained by pretraining with measurement data only, by an average of 5.2% across the public datasets. We conclude that augmenting HAR models with synthetic IMU data provides clear performance improvements for HAR and a versatile approach to accurately reflect human movements

Schlussbericht "Fortschrittliche Glas-Interposer mit Carbon-Kupfer-Verbundmetallisierung" zu IGF-Vorhaben Nr. 337 EBG