Christian Koch

Machine learning (ML) models will soon be found in devices of critical structures or even in high-risk applications. Arranging the Artificial Intelligence Act of European Union, requirements and a methodology for an approval of measurement devices by a regulatory authority or a state institution need to be developed. In this study, using the example of loudness classification from electroencephalography measurement data, phenomenological methods were developed for the evaluation of ML models in the context of an approval process. A schedule of six sequential steps is proposed that can be used as a flowchart in such a process. After identifying the relevant features (step 1) performance measures and benchmarks were defined (step 2). In the following performance evaluation (step 3) the influence of the input data on the model’s output was determined. The range and characteristics of the input data at which the device operates correctly were identified and quantitative descriptors were determined for the full range of operating conditions specified in standards or regulations for approval (step 4). Perturbation-based techniques were developed and adapted to define corner cases and to create new test data that identify the areas of the input data where performance falls below a certain threshold (step 5). The newly found test data were finally applied in step 6, and it could be shown that the data can reveal weak points in the models. All processes can and must be used without knowledge of the models, as an independent test of a finished device is always carried out during a type approval. Thus, this study contributes to the still very nascent field of type approval development for measuring devices, software and methods containing machine or deep learning.

The effects of prolonged infrasound (IS) exposure on brain function and behavior are largely unknown, with only one prior study investigating functional connectivity (FC) changes. In a long-term randomized-controlled trial, 38 participants were exposed to inaudible airborne IS (6 Hz, 80–90 dB) or sham devices for four weeks (8 h/night). We assessed FC changes in resting-state networks (auditory, default mode (DMN), sensorimotor (SMN), and executive control (ECN)), and explored IS ‘sensitivity’ as a predictor of identified significant FC changes. We also examined correlations between somatic symptoms and FC. IS exposure led to decreased FC in the right precuneus (DMN) and increased FC in the Vermis IV and V (SMN). In the ECN, we observed increased FC in the right frontal middle gyrus (BA8) and the right inferior parietal lobe, and decreased FC in another region of the right frontal middle gyrus. Changes in the ECN (right inferior parietal lobe) were negatively associated with self-reported annoyance from IS/low-frequency noise. A significant negative association was found between FC changes in the DMN (right precuneus) and somatic symptoms. Our study is the first to investigate prolonged IS exposure effects on brain FC, revealing changes in the vDMN, SMN, and ECN, but not in the auditory network. Future studies should assess annoyance and sensitivity markers, fine-grained measures of somatic symptoms, and stratify samples by sensitivity to uncover individual differences in response to IS.

With the growing prevalence of infrasound and potential for annoyance comes the need for noise assessment. Performance validation of measuring instruments is an established necessity for reliable measurement data at conventional frequencies. However, infrasound measurements are critically dependent on the integrity of the microphone. A case study is presented showing that errors in excess of 20 dB result if the microphone diaphragm is perforated, and that such a defect cannot be detected by visual examination or with a typical sound calibrator. A further laboratory study validates the findings, and a scheme is proposed for identifying when such an issue exists.

Reliable assessment of airborne infrasound emitted by natural and technical sources has been gaining importance, while technology is limited and regulations are not sufficient in the infrasound frequency range. This paper provides a comparison of measurement microphones and microbarometers used for infrasound measurements in the field near wind power plants. Simultaneous measurements with both sensor systems were conducted at multiple sites in the vicinity of a wind park. Based on these measurements, it was shown that wind-induced noise is a major concern when measuring infrasound in the field. For this reason, additional experiments were conducted in an aeroacoustic wind tunnel to further investigate the sensitivity to wind of the sensors used. A combination of field measurements and laboratory experiments was applied to determine the influence of wind-induced noise in a real-life situation and multiple strategies are discussed for establishing quality criteria and controlling the influence of wind on such measurements.

The demand for reliable and traceable measurements of airborne infrasound has risen, one major application being the International Monitoring System run by the Comprehensive Nuclear-Test-Ban Treaty Organization. However, the current calibration methods do not sufficiently cover the infrasound frequency range. In this paper, we present a calibration method for microphones in this frequency range and its implementation as a measurement setup. The method is based on the vertical gradient of the ambient pressure as a stimulus. A DUT is subjected to an alternating pressure by periodically changing its altitude. The measurement setup realizes such a periodic altitude change by means of a vertically rotating disk and is capable of calibrating measurement microphones in a frequency range from 0.1 Hz to 5 Hz with a planned extension to 10 Hz. A measurement uncertainty of 0.07 dB at maximum could be realized. Particular attention was paid to the mechanics of the measurement setup to ensure that the DUT moves in a precisely determined orbit. We present example calibrations and an uncertainty budget for a Brüel & Kjær 4193 measurement microphone in the frequency range from 0.1 Hz to 5 Hz. Finally, we demonstrate the performance of the calibration method by comparing the acquired results to other calibration techniques showing an agreement better than 0.1 dB.

Concerns about potential health risks from infrasound exposure require objective investigation methods for the perception of infrasound by humans. Neural-network methods were applied to identify weak infrasound evoked potentials and to investigate the efficacy of these techniques. Two simple networks – a fully connected three-layer feed-forward network and a two-layer autoencoder – were designed for the analysis of electroencephalography (EEG) data recorded in response to infrasound stimuli presented. Late auditory evoked potentials (LAEPs) were detected and correlation factors between the target and the output of the networks were determined to assess the differentiation ability of the two networks. Both networks were able to differentiate infrasound evoked potentials from noise but showed different performance levels with respect to a differentiation threshold. The results can be applied to objectively identify perception events by infrasound in humans.

As airborne ultrasound can be found in many technical applications and everyday situations, the question as to whether sounds at these frequencies can be heard by human beings or whether they present a risk to their hearing system is of great practical relevance. To objectively study these issues, the monaural hearing threshold in the frequency range from 14 to 24 kHz was determined for 26 test subjects between 19 and 33 years of age using pure tone audiometry. The hearing threshold values increased strongly with increasing frequency up to around 21 kHz, followed by a range with a smaller slope toward 24 kHz. The number of subjects who could respond positively to the threshold measurements decreased dramatically above 21 kHz. Brain activation was then measured by means of magnetoencephalography (MEG) and functional magnetic resonance imaging (fMRI) and with acoustic stimuli at the same frequencies, with sound pressure levels (SPLs) above and below the individual threshold. No auditory cortex activation was found for levels below the threshold. Although test subjects reported audible sounds above the threshold, no brain activity was identified in the above-threshold case under current experimental conditions except at the highest sensation level, which was presented at the lowest test frequency.

Hintergrund: Nach einer Phakoemulsifikation können Zellschädigungen am Hornhautendothel auftreten, deren Ursache bislang unbekannt ist. Die Temperatursteigerung in der Vorderkammer während der Linsenemulsifikation gilt als eine potenzielle Schädigungsursache. Ziel dieser Studie war, den Zusammenhang zwischen der Temperatursteigerung und den Zellschäden zu prüfen. Material und Methoden: Die Studie wurde an 86 Schweineaugen durchgeführt, die in 6 Versuchsgruppen A–F aufgeteilt wurden. Gruppe A diente als unbehandelte Kontrollgruppe. In den Gruppen B–E wurde eine Phakoemulsifikation mit verschiedenen gebräuchlichen Geräteeinstellungen simuliert. Gleichzeitig wurde die Temperatursteigerung in der Vorderkammer mit Sensoren gemessen. In Gruppe F wurden lediglich die Instrumente positioniert und Irrigation und Aspiration aktiviert. Bei allen 86 Hornhäuten wurde der Zellschaden am Endothel mittels Rasterelektronenmikroskopie (REM) nach einem standardisierten Protokoll quantifiziert. Der Zellschaden wurde auf eine Korrelation mit den zuvor gemessenen Temperatursteigerungen hin überprüft. Ergebnisse: In Gruppe E mit einer Leistungseinstellung von 50 % und blockierter Aspiration wurden Temperatursteigerungen von bis zu 17,77 °C gemessen. Signifikant niedrigere Temperatursteigerungen wurden in Gruppe D mit 100 % und unblockierter Aspiration gemessen (p = 0,006), hier betrug die höchste Temperatursteigerung 8,89 °C. Dagegen wurde der Zustand der Zellen in Gruppe D rasterelektronenmikroskopisch signifikant schlechter bewertet als in Gruppe E (p < 0,001). Unter Berücksichtigung aller Messdaten konnte keine signifikante Korrelation zwischen der Temperatursteigerung und der Zellschädigung festgestellt werden. Mit der Steigerung der Leistungseinstellung nahm jedoch der Zellverlust nicht linear zu. Schlussfolgerung: Die Temperatursteigerung war in den Messungen nicht die entscheidende Schädigungsursache. Es kann daher davon ausgegangen werden, dass die Temperatursteigerung in der Vorderkammer bei einer unter Alltagsbedingungen durchgeführten Phakoemulsifikation keine ursächliche Rolle für den Endothelschaden spielt. Aufgrund des starken Zusammenhangs zwischen der Leistungseinstellung und den auftretenden Zellschäden sollte in der klinischen Praxis stets die geringstmögliche Leistungseinstellung verwendet werden, welche die Fragmentierung der Linse ermöglicht. Purpose: Endothelial cell damage can be found after a phacoemulsification. The reason for this cell damage is not yet known. The temperature rise inside the anterior chamber during the emulsification of the lens is considered as a potential damage mechanism. The aim of this study was to investigate the relationship between the temperature increase and the cell damage. Materials and Methods: This study was performed with 86 enucleated porcine eyes, which were divided into six groups A–F. Group A served as control group. Phacoemulsification was simulated in the groups B–E with different surgically relevant system settings. The temperature change was measured simultaneously with temperature sensors inside the anterior chamber. Inside the eyes of group F, the instruments were positioned and irrigation and aspiration were activated, but no ultrasound was applied. For all 86 corneas, the endothelial cell damage was quantified with a standardised evaluation protocol using scanning electron microscopy (SEM). A potential correlation was checked between the observed cell damage and the measured temperature rise. Results: The highest temperature increase in group E with an obstructed aspiration line and an output power setting of 50 % was 17.77 °C. The temperature increase in group D with an output power setting of 100 % and unblocked fluid flow was significantly lower (p = 0.006). The highest temperature increase in group D was 8.89 °C. In contrast, the cells in group D were rated with a significantly lower score value than the cells in group E (p < 0.001). No significant correlation between the temperature rise and the cell damage could be obtained in due consideration of all measured data. However, an increase of the output power setting was connected with a non-linear increase of the cell loss. Conclusion: The temperature rise was not the main reason for the endothelial cell damage in the performed experiments. Therefore, there seems to be no causal relationship between the temperature increase inside the anterior chamber during a common phacoemulsification and the endothelial cell damage. Due to the strong dependence between the output power setting and the cell damage, the lowest output power setting should be used during surgery which ensures the emulsification of the nucleus.