Concept-based AI interpretability in physiological time-series data: Example of abnormality detection in electroencephalography

Computer Methods and Programs in Biomedicine, vol. 257, pp. 108448

Abstract

• Testing with concept activation vectors (TCAV) employs concept-based explanations of deep neural networks built for the classification of electroencephalography (EEG) data. • TCAV analysis confirms the medical expectation with high concept sensitivity towards typical epileptogenic signal patterns for the classification of abnormal EEGs. • Concept samples can be created from labeled signal sections, using metadata, as well as by extracting or simulating signal characteristics in the form of frequencies. • TCAV analysis yields consistent results for concept samples from different sources. Despite recent performance advancements, deep learning models are not yet adopted in clinical practice on a wide scale. The intrinsic intransparency of such systems is commonly cited as one major reason for this reluctance. This has motivated methods that aim to provide explanations of model functioning. Known limitations of feature-based explanations have led to an increased interest in concept-based interpretability. Testing with Concept Activation Vectors (TCAV) employs human-understandable, abstract concepts to explain model behavior. The method has previously been applied to the medical domain in the context of electronic health records, retinal fundus images and magnetic resonance imaging. We explore the usage of TCAV for building interpretable models on physiological time series, using an example of abnormality detection in electroencephalography (EEG). For this purpose, we adopt the XceptionTime model, which is suitable for multi-channel physiological data of variable sizes. The model provides state-of-the-art performance on raw EEG data and is publically available. We propose and test several ideas regarding concept definition through metadata mining, using additional labeled EEG data and extracting interpretable signal characteristics in the form of frequencies. By including our own hospital data with analog labeling, we further evaluate the robustness of our approach. The tested concepts show a TCAV score distribution that is in line with the clinical expectations, i.e. concepts known to have strong links with EEG pathologies (such as epileptiform discharges) received higher scores than the neutral concepts (e.g. sex). The scores were consistent across the applied concept generation strategies. TCAV has the potential to improve interpretability of deep learning applied to multi-channel signals as well as to detect possible biases in the data. Still, further work on developing the strategies for concept definition and validation on clinical physiological time series is needed to better understand how to extract clinically relevant information from the concept sensitivity scores.

Authors 9

1. Alexander Brenner corresponding

   University of Münster

   Affiliation as printed

   Institute of Medical Informatics, University of Münster, Münster, Germany. Electronic address: alexander.brenner@uni-muenster.de

2. Felix Knispel Aachen Institute of Medical Informatics Medical Faculty

   RWTH Aachen University

   Affiliation as printed

   Institute of Medical Informatics, Medical Faculty, RWTH Aachen University, Aachen, Germany

3. Florian P. Fischer Aachen Department of Epileptology and Neurology Medical Faculty

   Universitätsklinikum Aachen · RWTH Aachen University

   Affiliation as printed

   Department of Epileptology and Neurology, Medical Faculty, RWTH Aachen University Hospital, Aachen, Germany

4. Peter Rossmanith Aachen Theoretical Computer Science Department of Computer Science

   RWTH Aachen University

   Affiliation as printed

   Theoretical Computer Science, Department of Computer Science, RWTH Aachen University, Aachen, Germany

5. Yvonne G. Weber Aachen Department of Epileptology and Neurology Medical Faculty

   Universitätsklinikum Aachen · RWTH Aachen University

   Affiliation as printed

   Department of Epileptology and Neurology, Medical Faculty, RWTH Aachen University Hospital, Aachen, Germany

6. Henner Koch Aachen Department of Epileptology and Neurology Medical Faculty

   Universitätsklinikum Aachen · RWTH Aachen University

   Affiliation as printed

   Department of Epileptology and Neurology, Medical Faculty, RWTH Aachen University Hospital, Aachen, Germany

7. Rainer Röhrig Aachen Institute of Medical Informatics Medical Faculty

   RWTH Aachen University

   Affiliation as printed

   Institute of Medical Informatics, Medical Faculty, RWTH Aachen University, Aachen, Germany

8. Julian Varghese

   University of Münster

   Affiliation as printed

   Institute of Medical Informatics, University of Münster, Münster, Germany

9. Ekaterina Kutafina

   University of Cologne · University Hospital Cologne

   Affiliation as printed

   Institute for Biomedical Informatics, Faculty of Medicine, University Hospital Cologne, University of Cologne, Cologne, Germany