Heinz Eric Krestel, MD

People with epilepsy (PWE) are not only affected by seizures but also be short electric changes in the electroencephalogram (EEG) that occur between seizures, known as interictal epileptiform discharges (IEDs). IEDs can have effects on behavior and cognition that range from negligible impairment to the transient inability to perform an action such as overlooking a stop sign and causing an accident. Although IEDs are much shorter in duration compared to seizures, ranging from half a second to a few seconds, if they occur in series or runs, they can be up to 2,000 times more frequent than seizures in temporal lobe epilepsy (Jansky J Epileptic Disord. 2005). Thus, IEDs can affect daily lives of PWE, for example, during communication, driving a car, while studying and recalling memory. There are experimental tests to visualize IED-induced deficits, but there are no standardized tests that can be routinely used to detect IED-associated effects on daily social functioning, because of a lack of automatic, objective, fast, and versatile methods. In addition, there is currently no consensus in the scientific literature about the ideal test for measuring IED effects in relation to driving ability (Markhus R. Seizure 2020; Spöndlin L. Epilepsia Open 2024).

To better predict the clinical correlates of IEDs, we have investigated the relationship between electrical IED characteristics in the EEG, so called electrophysiological parameters, and their clinical correlates. We studied nearly 100 patients with different types of epilepsy using 2 reaction tests, a simple flash test and a simple driving game on a laptop, and a tracking test, i.e., a realistic driving simulator (Krestel H. Neurology 2023). The measures were reaction times and missed reactions or virtual accidents during normal EEG and IEDs in single recordings. The IEDs were mostly serial spikes and/or spike waves. The visual test stimuli were triggered randomly during normal EEG and as soon as IEDs were perceived by staff. We then calculated both the IED-associated reaction time prolongation and the IED-associated missed responses per person. These values were first grouped according to the IEDs’ electrical field width in values associated with focal IEDs and values associated with generalized IEDs. The latter values were further divided according to IED appearance into those associated with generalized typical IEDs and those associated with generalized atypical IEDs. Typical IEDs had an unequivocal spike or spike-wave morphology and a constant amplitude during the duration of the IED, thus they were well organized. All other generalized epileptiform EEG changes were designated as generalized atypical and their clinical correlates, i.e., reaction time prolongations and missed reactions were pooled in this group. Generalized typical IEDs were associated with longer reaction time prolongations compared with generalized atypical and focal IEDs, as well as with a higher rate of missed reactions. Longer reaction time prolongations and higher rates of missed reactions were also associated with longer durations of IEDs. The mean latency of IEDs associated with missed responses was significantly longer (2481 ms) than the mean latency of IEDs associated with preserved responses (1403 ms) (Krestel H. Neurology 2023). Since IEDs can vary in their appearance and have variable effects, two identically looking IEDs in the EEG may have different effects. We will therefore probably never be able to predict the effects of IEDs with high accuracy based on their electrical characteristics as seen in the EEG. One of the predictive powers of IEDs is that their severe clinical correlates such as missed reactions can be predicted from their mild effects such as reaction time prolongation. By describing the cumulative rate of missed reactions as a non-linear function of IED-associated reaction time prolongation, a risk for missed reactions can be predicted: for example, the risk is 20% at a mean reaction time prolongation of 100ms and rises to 50% at an average reaction time prolongation of 150ms (Krestel H. Neurology 2023). Finally, this work, which was awarded the Alfred-Hauptmann Award for Clinical Research 2025 by the Swiss, German, and Austrian Leagues against Epilepsy (https://www.epi.ch/fr/deux-laureats-de-suisse-pour-le-prix-alfred-hauptmann/) also shows that odds ratios between causing a virtual accident in a realistic driving simulator and causing non-fatal accidents while driving due to sleepiness or low-level blood alcohol are comparable (Krestel H. Neurology 2023). While it is difficult to predict clinical correlates from electrical IED characteristics, we used Artificial Intelligence (AI) that can learn IEDs of variable appearance to then trigger tasks and measure the patients’ responses. We called our AI-based system the Digital Response Test in Epilepsy (DigRTEpi). DigRTEpi was developed in a multi-national collaboration of the Epilepsy Center Frankfurt Rhine-Main (Germany), Von Allmen Engineering (Switzerland), and the Yale School of Medicine (USA), under the leadership of the lead inventor H. Krestel, and funded by the European Union Framework Program for Research and Innovation Horizon 2020, Marie Sklodowska-Curie Grand Agreement No. 99791. DigRTEpi’s AI visualizes the ongoing recording window-by-window, classifies the resulting images using a deep neural network, and triggers stimuli in a driving game measuring reaction times and crashes, or in the interictal Automated Responsiveness Test (iART) ultra-rapidly assessing cognitive functions. When iART is used in the closed electronic circuit of DigRTEpi, the automatic IED detection triggers videos that contain brief instructions in video, audio and written form (subtitles), lasting up to 5 seconds, and test orientation (time, person, place), speech comprehension, word recall, word repetition, knowledge of body parts, left-right discrimination, apraxia, memory (e.g. time), numerical reasoning, and executive functions. (https://www.researchgate.net/publication/362263877_The_interictal_Automated_Responsiveness_Test_iART_analyzes_transient_cognitive_impairment_in_an_international_manner#fullTextFileContent). We have recorded the instructions in 17 different languages, because we believe that the sensitivity and specificity of detecting transitory cognitive impairment due to IEDs are higher when PWE are tested in their native language. DigRTEpi has been tested with PWE in a two-arms prospective study. The manuscript is currently under review. Our AI-based system combines advantageous features that have not been available in previous systems. First, it detects IEDs in real time, which is essential to capture their transient clinical effects. Second, the AI used to detect IEDs has high sensitivity, specificity, and a low generalization error, as shown in the prospective pilot study. This enable the use of DigRTEpi in clinical routine, where all possible types of epilepsy with IEDs of variable appearance are seen. Third, DigRTEpi is versatile and combines testing of behavior (slowed and missed reactions) using its driving game, as well as cognitive functions using its newly developed ultra-rapid neuropsychological testing iART. All of the DigRTEpi features are important because there is still disagreement about a suitable/ideal test to measure IED effects. Thus, DigRTEpi allows laboratory measurements of IED effects with real-world applicability and bridges a gap in current clinical care of PWE by detecting IEDs of variable appearance and measuring their effects in real time.

A study is ongoing at the Yale Comprehensive Epilepsy Center, in which children and adults with epilepsy are tested for the clinical correlates of their IEDs. An NIH application has been submitted with the title: “Digital Response Testing in Epilepsy (DigRTEpi): Development of an Intelligent System for Assessing the Impact of Epileptiform Discharges on Cognition, Attention, and Driving Fitness”. Its aims are to personalize the AI algorithms, optimize the system’s usability and stability, expand DigRTEpi’s compatibility with the most common commercial EEG systems, increase the testing options, and automate reporting.

DigRTEpi will also be introduced into clinical routine at the Cantonal Hospital Winterthur and the University Hospital Zürich in Switzerland. Swiss driving regulations require a so-called EEG compatibility for PWE in addition to certain periods of seizure freedom. DigRTEpi will help find clinically relevant effects of IEDs and thus contribute to increased road safety. A longitudinal treatment trial is planned in which PWE, who were found to have clinically relevant IED effects, will undergo repeated adjustments to their antiseizure medication and measurements of IED effects using DigRTEpi. This study will help to determine in the future, if clinically relevant IED effects such as missed responses to stimuli can be suppressed by treatment adjustments and if patient’s quality of life can be improved. Finally, a web-based application is planned that uses DigRTEpi's AI to review anonymized EEGs and make recommendations on which patients should be clinically tested.

Publications, topic-related since 2019

• Krestel H, Rosenow F, Blumenfeld H, von Allmen A. Real-time EEG classification with convolutional networks and ResNet. Conference proceeding American Epilepsy Society meeting 2019; https://cms.aesnet.org/abstractslisting/real-time-eeg-classification-with-convolutional-networks-and-resnet
• Cohen E, Antwi P, Banz BC, Vincent P, Saha R, Arencibia CA, Ryu JH, Atac E, Saleem N, Tomatsu S, Swift K, Hu C, Krestel H, Farooque P, Levy S, Wu J, Crowley M, Vaca FE, Blumenfeld H. Realistic driving simulation during generalized epileptiform discharges to identify electroencephalographic features related to motor vehicle safety: Feasibility and pilot study. Epilepsia 2020;61:19-28; DOI: 10.1111/epi.16356
• Markhus R, Henning O, Molteberg E, Hećimović H, Ujvari A, Hirsch E, Rheims S, Surges R, Malmgren K, Rüegg S, Gil-Nagel A, Roivainen R, Picard F, Steinhoff B, Marusic P, Mostacci B, Kimiskidis VK, Mindruta I, Jagella C, Mameniškienė R, Schulze-Bonhage A, Rosenow F, Kelemen A, Fabo D, Walker MC, Seeck M, Krämer G, Arsene OT, Krestel H, Lossius M. EEG in fitness to drive evaluation in people with epilepsy - variations across Europe. Seizure: European Journal of Epilepsy 2020;79:56-60. DOI: 10.1016/j.seizure.2020.04.013
• Kumar A, Martin R, Chen W, Bauerschmidt A, Youngblood MW, Cunningham C, Si Y, Ezeani C, Kratochvil Z, Bronen J, Thomson J, Riordan K, Yoo JY, Shirka R, Manganas L, Krestel H, Hirsch LJ, Blumenfeld H Simulated driving in the epilepsy monitoring unit: Effects of seizure type, consciousness, and motor impairment. Epilepsia 2022;63:e30-e34. DOI: 10.1111/epi.17136
• Kumar A, Martin R, Chen W, et al. Simulated driving in the epilepsy monitoring unit: Effects of seizure type, consciousness, and motor impairment. Erratum. Epilepsia. 2022 Feb 25. DOI: 10.1111/epi.17203
• Abukhadra Y, Li J, Springer M, Khalaf A, Roethlisberger S, Krestel H, Blumenfeld H. EEG and Machine Learning in Prediction of Impaired Responses to Visual Stimuli During Interictal Epileptiform Discharges. Conference proceeding American Epilepsy Society meeting 2021; https://cms.aesnet.org/abstractslisting/eeg-and-machine-learning-in-prediction-of-impaired-responses-to-visual-stimuli-during-interictal-epileptiform-discharges
• Springer M, Khalaf A, Vincent P, Ryu JH, Abukhadra Y, Beniczky S, Glauser T, Krestel H, Blumenfeld H. A Machine Learning Approach for Predicting Impaired Consciousness in Absence Epilepsy. Annals of Clinical and Translational Neurology 2022. DOI: https://doi.org/10.1002/acn3.51647
• Krestel H, Rackauskaite J, Khoueiry M, von Allmen A, Li YY, Pereira Marcal G, Quiroz A, Erceg N, Panos L, Pelczar M, Schoretsanitis G, Cicek M, Zedka M, Jagella C, Markhus C, Rosenow F, Blumenfeld H. The interictal automated responsiveness test (iART) analyzes transient cognitive impairment in an international manner. Conference proceeding European Epilepsy Congress 2022; https://www.researchgate.net/publication/362263877_The_interictal_Automated_Responsiveness_Test_iART_analyzes_transient_cognitive_impairment_in_an_international_manner#fullTextFileContent
• European Epilepsy Congress 2022 Forum, Title: "Real-time processing of brain biosignals; applied to epileptiform discharges, behavior, and cognition" Chair: Heinz Krestel (Switzerland) State of the art of machine learning algorithms for real-time biosignal processing, advantages, disadvantages, potential applications - Speaker: Diyuan Lu (Germany);Brain-computer interfaces and its latencies - Speaker: Heinz Krestel (Switzerland);Detection of transient cognitive impairment in dementia evaluation - Speaker: Justina Rackauskaite (Switzerland); Detection of transient cognitive impairment in fitness-to-drive evaluation - Speaker: Heinz Krestel (Switzerland)
• Krestel H, Schreier DR, Sakiri E, von Allmen A, Abukhadra Y, Nirkko A, Steinlin M, Rosenow F, Markhus R, Schneider G, Jagella C, Mathis J, Blumenfeld H. Predictive Power of Interictal Epileptiform Discharges in Fitness-to-Drive Evaluation. Neurology. 2023
  Aug 29;101(9):e866-e878. doi: 10.1212/WNL.0000000000207531
• Kleen JK, Davis KA. Is It Reasonable to Drive When There Is a (Spike) Train? Neurology. 2023 Aug 29;101(9):377-379.
  doi: 10.1212/WNL.0000000000207651
• von Allmen A, Lu D, Jagella C, Abukhadra Y, Markhus R, Triesch J, Gopaul M, Hirsch LJ, Rosenow F, Blumenfeld H, and Krestel H. AI-based System to assess clinical impact of interictal epileptiform discharges in real time. Submitted.