Djordje Slijepcevic

Fabian Horst, Djordje Slijepcevic, Matthias Zeppelzauer et al.

State-of-the-art machine learning (ML) models are highly effective in classifying gait analysis data, however, they lack in providing explanations for their predictions. This "black-box" characteristic makes it impossible to understand on which input patterns, ML models base their predictions. The present study investigates whether Explainable Artificial Intelligence methods, i.e., Layer-wise Relevance Propagation (LRP), can be useful to enhance the explainability of ML predictions in gait classification. The research question was: Which input patterns are most relevant for an automated gender classification model and do they correspond to characteristics identified in the literature? We utilized a subset of the GAITREC dataset containing five bilateral ground reaction force (GRF) recordings per person during barefoot walking of 62 healthy participants: 34 females and 28 males. Each input signal (right and left side) was min-max normalized before concatenation and fed into a multi-layer Convolutional Neural Network (CNN). The classification accuracy was obtained over a stratified ten-fold cross-validation. To identify gender-specific patterns, the input relevance scores were derived using LRP. The mean classification accuracy of the CNN with 83.3% showed a clear superiority over the zero-rule baseline of 54.8%.

Djordje Slijepcevic, Fabian Horst, Marvin Simak et al.

Machine learning (ML) models have proven effective in classifying gait analysis data, e.g., binary classification of young vs. older adults. ML models, however, lack in providing human understandable explanations for their predictions. This "black-box" behavior impedes the understanding of which input features the model predictions are based on. We investigated an Explainable Artificial Intelligence method, i.e., Layer-wise Relevance Propagation (LRP), for gait analysis data. The research question was: Which input features are used by ML models to classify age-related differences in walking patterns? We utilized a subset of the AIST Gait Database 2019 containing five bilateral ground reaction force (GRF) recordings per person during barefoot walking of healthy participants. Each input signal was min-max normalized before concatenation and fed into a Convolutional Neural Network (CNN). Participants were divided into three age groups: young (20-39 years), middle-aged (40-64 years), and older (65-79 years) adults. The classification accuracy and relevance scores (derived using LRP) were averaged over a stratified ten-fold cross-validation. The mean classification accuracy of 60.1% was clearly higher than the zero-rule baseline of 37.3%. The confusion matrix shows that the CNN distinguished younger and older adults well, but had difficulty modeling the middle-aged adults.

Armin Kirchknopf, Djordje Slijepcevic, Ilkay Wunderlich et al.

We investigate the problem of explainability for visual object detectors. Specifically, we demonstrate on the example of the YOLO object detector how to integrate Grad-CAM into the model architecture and analyze the results. We show how to compute attribution-based explanations for individual detections and find that the normalization of the results has a great impact on their interpretation.

Armin Kirchknopf, Djordje Slijepcevic, Matthias Zeppelzauer

Social media is accompanied by an increasing proportion of content that provides fake information or misleading content, known as information disorder. In this paper, we study the problem of multimodal fake news detection on a largescale multimodal dataset. We propose a multimodal network architecture that enables different levels and types of information fusion. In addition to the textual and visual content of a posting, we further leverage secondary information, i.e. user comments and metadata. We fuse information at multiple levels to account for the specific intrinsic structure of the modalities. Our results show that multimodal analysis is highly effective for the task and all modalities contribute positively when fused properly.

Thomas Baumhauer, Djordje Slijepcevic, Matthias Zeppelzauer

Explainable artificial intelligence is the attempt to elucidate the workings of systems too complex to be directly accessible to human cognition through suitable side-information referred to as "explanations". We present a trainable explanation module for convolutional image classifiers we call bounded logit attention (BLA). The BLA module learns to select a subset of the convolutional feature map for each input instance, which then serves as an explanation for the classifier's prediction. BLA overcomes several limitations of the instancewise feature selection method "learning to explain" (L2X) introduced by Chen et al. (2018): 1) BLA scales to real-world sized image classification problems, and 2) BLA offers a canonical way to learn explanations of variable size. Due to its modularity BLA lends itself to transfer learning setups and can also be employed as a post-hoc add-on to trained classifiers. Beyond explainability, BLA may serve as a general purpose method for differentiable approximation of subset selection. In a user study we find that BLA explanations are preferred over explanations generated by the popular (Grad-)CAM method.

Djordje Slijepcevic, Fabian Horst, Sebastian Lapuschkin et al.

Machine learning (ML) is increasingly used to support decision-making in the healthcare sector. While ML approaches provide promising results with regard to their classification performance, most share a central limitation, namely their black-box character. Motivated by the interest to understand the functioning of ML models, methods from the field of Explainable Artificial Intelligence (XAI) have recently become important. This article investigates the usefulness of XAI methods in clinical gait classification. For this purpose, predictions of state-of-the-art classification methods are explained with an established XAI method, i.e., Layer-wise Relevance Propagation (LRP). We propose to evaluate the obtained explanations with two complementary approaches: a statistical analysis of the underlying data using Statistical Parametric Mapping and a qualitative evaluation by a clinical expert. A gait dataset comprising ground reaction force measurements from 132 patients with different lower-body gait disorders and 62 healthy controls is utilized. We investigate several gait classification tasks, employ multiple classification methods, and analyze the impact of data normalization and different signal components for classification performance and explanation quality. Our experiments show that explanations obtained by LRP exhibit promising statistical properties concerning inter-class discriminativity and are also in line with clinically relevant biomechanical gait characteristics.

Djordje Slijepcevic, Matthias Zeppelzauer, Anna-Maria Gorgas et al.

This article proposes a comprehensive investigation of the automatic classification of functional gait disorders based solely on ground reaction force (GRF) measurements. The aim of the study is twofold: (1) to investigate the suitability of stateof-the-art GRF parameterization techniques (representations) for the discrimination of functional gait disorders; and (2) to provide a first performance baseline for the automated classification of functional gait disorders for a large-scale dataset. The utilized database comprises GRF measurements from 279 patients with gait disorders (GDs) and data from 161 healthy controls (N). Patients were manually classified into four classes with different functional impairments associated with the "hip", "knee", "ankle", and "calcaneus". Different parameterizations are investigated: GRF parameters, global principal component analysis (PCA)-based representations and a combined representation applying PCA on GRF parameters. The discriminative power of each parameterization for different classes is investigated by linear discriminant analysis (LDA). Based on this analysis, two classification experiments are pursued: (1) distinction between healthy and impaired gait (N vs. GD) and (2) multi-class classification between healthy gait and all four GD classes. Experiments show promising results and reveal among others that several factors, such as imbalanced class cardinalities and varying numbers of measurement sessions per patient have a strong impact on the classification accuracy and therefore need to be taken into account. The results represent a promising first step towards the automated classification of gait disorders and a first performance baseline for future developments in this direction.

Markus Wagner, Djordje Slijepcevic, Brian Horsak et al.

In 2014, more than 10 million people in the US were affected by an ambulatory disability. Thus, gait rehabilitation is a crucial part of health care systems. The quantification of human locomotion enables clinicians to describe and analyze a patient's gait performance in detail and allows them to base clinical decisions on objective data. These assessments generate a vast amount of complex data which need to be interpreted in a short time period. We conducted a design study in cooperation with gait analysis experts to develop a novel Knowledge-Assisted Visual Analytics solution for clinical Gait analysis (KAVAGait). KAVAGait allows the clinician to store and inspect complex data derived during clinical gait analysis. The system incorporates innovative and interactive visual interface concepts, which were developed based on the needs of clinicians. Additionally, an explicit knowledge store (EKS) allows externalization and storage of implicit knowledge from clinicians. It makes this information available for others, supporting the process of data inspection and clinical decision making. We validated our system by conducting expert reviews, a user study, and a case study. Results suggest that KAVAGait is able to support a clinician during clinical practice by visualizing complex gait data and providing knowledge of other clinicians.