Explaining deep learning for ECG analysis: Building blocks for auditing and knowledge discovery.

Deep neural networks have become increasingly popular for analyzing ECG data because of their ability to accurately identify cardiac conditions and hidden clinical factors. However, the lack of transparency due to the black box nature of these models is a common concern. To address this issue, explainable AI (XAI) methods can be employed. In this study, we present a comprehensive analysis of post-hoc XAI methods, investigating the glocal (aggregated local attributions over multiple samples) and global (concept based XAI) perspectives. We have established a set of sanity checks to identify saliency as the most sensible attribution method. We provide a dataset-wide analysis across entire patient subgroups, which goes beyond anecdotal evidence, to establish the first quantitative evidence for the alignment of model behavior with cardiologists' decision rules. Furthermore, we demonstrate how these XAI techniques can be utilized for knowledge discovery, such as identifying subtypes of myocardial infarction. We believe that these proposed methods can serve as building blocks for a complementary assessment of the internal validity during a certification process, as well as for knowledge discovery in the field of ECG analysis.

Using explainable AI to investigate electrocardiogram changes during healthy aging-From expert features to raw signals.

Cardiovascular diseases remain the leading global cause of mortality. Age is an important covariate whose effect is most easily investigated in a healthy cohort to properly distinguish the former from disease-related changes. Traditionally, most of such insights have been drawn from the analysis of electrocardiogram (ECG) feature changes in individuals as they age. However, these features, while informative, may potentially obscure underlying data relationships. In this paper we present the following contributions: (1) We employ a deep-learning model and a tree-based model to analyze ECG data from a robust dataset of healthy individuals across varying ages in both raw signals and ECG feature format. (2) We use explainable AI methods to identify the most discriminative ECG features across age groups.(3) Our analysis with tree-based classifiers reveals age-related declines in inferred breathing rates and identifies notably high SDANN values as indicative of elderly individuals, distinguishing them from younger adults. (4) Furthermore, the deep-learning model underscores the pivotal role of the P-wave in age predictions across all age groups, suggesting potential changes in the distribution of different P-wave types with age. These findings shed new light on age-related ECG changes, offering insights that transcend traditional feature-based approaches.

SonicGuard Sensor-A Multichannel Acoustic Sensor for Long-Term Monitoring of Abdominal Sounds Examined through a Qualification Study.

Auscultation is a fundamental diagnostic technique that provides valuable diagnostic information about different parts of the body. With the increasing prevalence of digital stethoscopes and telehealth applications, there is a growing trend towards digitizing the capture of bodily sounds, thereby enabling subsequent analysis using machine learning algorithms. This study introduces the SonicGuard sensor, which is a multichannel acoustic sensor designed for long-term recordings of bodily sounds. We conducted a series of qualification tests, with a specific focus on bowel sounds ranging from controlled experimental environments to phantom measurements and real patient recordings. These tests demonstrate the effectiveness of the proposed sensor setup. The results show that the SonicGuard sensor is comparable to commercially available digital stethoscopes, which are considered the gold standard in the field. This development opens up possibilities for collecting and analyzing bodily sound datasets using machine learning techniques in the future.

[Artificial intelligence-enhanced electrocardiography : Will it revolutionize diagnosis and management of our patients?] / Durch künstliche Intelligenz verstärkte Elektrokardiographie : Wird sie Diagnostik und Management unserer Patienten revolutionieren?

The use of artificial intelligence (AI) in healthcare has made significant progress in the last 10 years. Many experts believe that utilization of AI technologies, especially deep learning, will bring about drastic changes in how physicians understand, diagnose, and treat diseases. One aspect of this development is AI-enhanced electrocardiography (ECG) analysis. It involves not only optimizing the traditional ECG analysis by the physician and improving the accuracy of automatic interpretation by the ECG device but also introducing entirely new diagnostic options enabled by AI. Examples include assessing left ventricular function, predicting atrial fibrillation, and diagnosing both cardiac and noncardiac conditions. Through AI, the ECG becomes a comprehensive tool for screening, diagnosis, and patient management, potentially revolutionizing clinical practices. This paper provides an overview of the current state of this development, discusses existing limitations, and explores the challenges that may arise for healthcare professionals in this context.

Insights into the inner workings of transformer models for protein function prediction.

MOTIVATION: We explored how explainable artificial intelligence (XAI) can help to shed light into the inner workings of neural networks for protein function prediction, by extending the widely used XAI method of integrated gradients such that latent representations inside of transformer models, which were finetuned to Gene Ontology term and Enzyme Commission number prediction, can be inspected too. RESULTS: The approach enabled us to identify amino acids in the sequences that the transformers pay particular attention to, and to show that these relevant sequence parts reflect expectations from biology and chemistry, both in the embedding layer and inside of the model, where we identified transformer heads with a statistically significant correspondence of attribution maps with ground truth sequence annotations (e.g. transmembrane regions, active sites) across many proteins. AVAILABILITY AND IMPLEMENTATION: Source code can be accessed at https://github.com/markuswenzel/xai-proteins.

ECG Feature Importance Rankings: Cardiologists vs. Algorithms.

Feature importance methods promise to provide a ranking of features according to importance for a given classification task. A wide range of methods exist but their rankings often disagree and they are inherently difficult to evaluate due to a lack of ground truth beyond synthetic datasets. In this work, we put feature importance methods to the test on real-world data in the domain of cardiology, where we try to distinguish three specific pathologies from healthy subjects based on ECG features comparing to features used in cardiologists' decision rules as ground truth. We found that the SHAP and LIME methods and Chi-squared test all worked well together with the native Random forest and Logistic regression feature rankings. Some methods gave inconsistent results, which included the Maximum Relevance Minimum Redundancy and Neighbourhood Component Analysis methods. The permutation-based methods generally performed quite poorly. A surprising result was found in the case of left bundle branch block, where T-wave morphology features were consistently identified as being important for diagnosis, but are not used by clinicians.

Towards Quantitative Precision for ECG Analysis: Leveraging State Space Models, Self-Supervision and Patient Metadata.

Deep learning has emerged as the preferred modeling approach for automatic ECG analysis. In this study, we investigate three elements aimed at improving the quantitative accuracy of such systems. These components consistently enhance performance beyond the existing state-of-the-art, which is predominantly based on convolutional models. Firstly, we explore more expressive architectures by exploiting structured state space models (SSMs). These models have shown promise in capturing long-term dependencies in time series data. By incorporating SSMs into our approach, we not only achieve better performance, but also gain insights into long-standing questions in the field. Specifically, for standard diagnostic tasks, we find no advantage in using higher sampling rates such as 500 Hz compared to 100 Hz. Similarly, extending the input size of the model beyond 3 seconds does not lead to significant improvements. Secondly, we demonstrate that self-supervised learning using contrastive predictive coding can further improve the performance of SSMs. By leveraging self-supervision, we enable the model to learn more robust and representative features, leading to improved analysis accuracy. Lastly, we depart from synthetic benchmarking scenarios and incorporate basic demographic metadata alongside the ECG signal as input. This inclusion of patient metadata departs from the conventional practice of relying solely on the signal itself. Remarkably, this addition consistently yields positive effects on predictive performance. We firmly believe that all three components should be considered when developing next-generation ECG analysis algorithms.

Diffusion-based conditional ECG generation with structured state space models.

Generating synthetic data is a promising solution for addressing privacy concerns that arise when distributing sensitive health data. In recent years, diffusion models have become the new standard for generating various types of data, while structured state space models have emerged as a powerful approach for capturing long-term dependencies in time series. Our proposed solution, SSSD-ECG, combines these two technologies to generate synthetic 12-lead electrocardiograms (ECGs) based on over 70 ECG statements. As reliable baselines are lacking, we also propose conditional variants of two state-of-the-art unconditional generative models. We conducted a thorough evaluation of the quality of the generated samples by assessing pre-trained classifiers on the generated data and by measuring the performance of a classifier trained only on synthetic data. SSSD-ECG outperformed its GAN-based competitors. Our approach was further validated through experiments that included conditional class interpolation and a clinical Turing test, which demonstrated the high quality of SSSD-ECG samples across a wide range of conditions.

PTB-XL+, a comprehensive electrocardiographic feature dataset.

Machine learning (ML) methods for the analysis of electrocardiography (ECG) data are gaining importance, substantially supported by the release of large public datasets. However, these current datasets miss important derived descriptors such as ECG features that have been devised in the past hundred years and still form the basis of most automatic ECG analysis algorithms and are critical for cardiologists' decision processes. ECG features are available from sophisticated commercial software but are not accessible to the general public. To alleviate this issue, we add ECG features from two leading commercial algorithms and an open-source implementation supplemented by a set of automatic diagnostic statements from a commercial ECG analysis software in preprocessed format. This allows the comparison of ML models trained on clinically versus automatically generated label sets. We provide an extensive technical validation of features and diagnostic statements for ML applications. We believe this release crucially enhances the usability of the PTB-XL dataset as a reference dataset for ML methods in the context of ECG data.

From modern CNNs to vision transformers: Assessing the performance, robustness, and classification strategies of deep learning models in histopathology.

While machine learning is currently transforming the field of histopathology, the domain lacks a comprehensive evaluation of state-of-the-art models based on essential but complementary quality requirements beyond a mere classification accuracy. In order to fill this gap, we developed a new methodology to extensively evaluate a wide range of classification models, including recent vision transformers, and convolutional neural networks such as: ConvNeXt, ResNet (BiT), Inception, ViT and Swin transformer, with and without supervised or self-supervised pretraining. We thoroughly tested the models on five widely used histopathology datasets containing whole slide images of breast, gastric, and colorectal cancer and developed a novel approach using an image-to-image translation model to assess the robustness of a cancer classification model against stain variations. Further, we extended existing interpretability methods to previously unstudied models and systematically reveal insights of the models' classification strategies that allow for plausibility checks and systematic comparisons. The study resulted in specific model recommendations for practitioners as well as putting forward a general methodology to quantify a model's quality according to complementary requirements that can be transferred to future model architectures.

New definitions of human lymphoid and follicular cell entities in lymphatic tissue by machine learning.

Histological sections of the lymphatic system are usually the basis of static (2D) morphological investigations. Here, we performed a dynamic (4D) analysis of human reactive lymphoid tissue using confocal fluorescent laser microscopy in combination with machine learning. Based on tracks for T-cells (CD3), B-cells (CD20), follicular T-helper cells (PD1) and optical flow of follicular dendritic cells (CD35), we put forward the first quantitative analysis of movement-related and morphological parameters within human lymphoid tissue. We identified correlations of follicular dendritic cell movement and the behavior of lymphocytes in the microenvironment. In addition, we investigated the value of movement and/or morphological parameters for a precise definition of cell types (CD clusters). CD-clusters could be determined based on movement and/or morphology. Differentiating between CD3- and CD20 positive cells is most challenging and long term-movement characteristics are indispensable. We propose morphological and movement-related prototypes of cell entities applying machine learning models. Finally, we define beyond CD clusters new subgroups within lymphocyte entities based on long term movement characteristics. In conclusion, we showed that the combination of 4D imaging and machine learning is able to define characteristics of lymphocytes not visible in 2D histology.

To pretrain or not? A systematic analysis of the benefits of pretraining in diabetic retinopathy.

There is an increasing number of medical use cases where classification algorithms based on deep neural networks reach performance levels that are competitive with human medical experts. To alleviate the challenges of small dataset sizes, these systems often rely on pretraining. In this work, we aim to assess the broader implications of these approaches in order to better understand what type of pretraining works reliably (with respect to performance, robustness, learned representation etc.) in practice and what type of pretraining dataset is best suited to achieve good performance in small target dataset size scenarios. Considering diabetic retinopathy grading as an exemplary use case, we compare the impact of different training procedures including recently established self-supervised pretraining methods based on contrastive learning. To this end, we investigate different aspects such as quantitative performance, statistics of the learned feature representations, interpretability and robustness to image distortions. Our results indicate that models initialized from ImageNet pretraining report a significant increase in performance, generalization and robustness to image distortions. In particular, self-supervised models show further benefits to supervised models. Self-supervised models with initialization from ImageNet pretraining not only report higher performance, they also reduce overfitting to large lesions along with improvements in taking into account minute lesions indicative of the progression of the disease. Understanding the effects of pretraining in a broader sense that goes beyond simple performance comparisons is of crucial importance for the broader medical imaging community beyond the use case considered in this work.

[Artificial intelligence-based ECG analysis: current status and future perspectives-Part 1 : Basic principles]. / EKG-Diagnostik mithilfe künstlicher Intelligenz: aktueller Stand und zukünftige Perspektiven ­ Teil 1 : Grundlagen.

Even though electrocardiography is a diagnostic procedure that is now more than 100 years old, medicine cannot do without it. On the contrary, interest in the procedure and its clinical significance is even increasing again. Reports on the evaluation of electrocardiograms (ECGs) with the aid of artificial intelligence (AI) are also responsible for this. Using machine learning and in particular deep learning, both AI subfields, completely new perspectives of ECG evaluation and interpretation arise. The weaknesses inherent in classical computer-assisted ECG evaluation appear to be overcome. This two-part overview deals with AI-based ECG analysis. Part 1 introduces basic aspects of the procedure. Part 2, which is published separately, is devoted to the current state of research and discusses the available studies. In addition, possible scenarios of future application of AI in ECG analysis are discussed.

[Artificial intelligence-based ECG analysis: current status and future perspectives-Part 2 : Recent studies and future]. / EKG-Diagnostik mit Hilfe künstlicher Intelligenz: aktueller Stand und zukünftige Perspektiven ­ Teil 2 : Aktuelle Studienlage und Ausblick.

While fundamental aspects of the application of artificial intelligence (AI) to electrocardiogram (ECG) analysis were discussed in part 1 of this review, the present work (part 2) provides a review of recent studies on the practical application of this new technology. The number of published articles on the topic of AI-based ECG analysis has been increasing rapidly since 2017. This is especially true for studies that use deep learning (DL) with artificial neural networks. The aim is not only to overcome the weaknesses of classical ECG diagnostics, but also to extend the functionality of the ECG. This involves the detection of cardiological and noncardiological diseases and the prediction for clinical events, e.g., the future development of left ventricular dysfunction and future clinical manifestation of atrial fibrillation. This is made possible by AI using DL to find subclinical patterns in giant ECG datasets and using them for algorithm development. AI-assisted ECG analysis is becoming a screening tool; it goes far beyond just being "better" than a cardiologist. The progress that has been made is remarkable and is generating much attention and also euphoria among experts and the public. However, most studies are proof-of-concept studies. Often, private (institution-owned) data are used, the quality of which is unclear. To date, clinical validation of the developed algorithms in other collectives and scenarios has been rare. Particularly problematic is that the way AI finds a solution so far mostly remains hidden from humans (black-box character of AI). Overall, AI-based electrocardiography is still in its infancy. However, it is already foreseeable that the ECG, as a diagnostic procedure that is easy to use and can be repeated as often as desired, will not only continue to be indispensable in the future, but will also gain in clinical importance.

Self-supervised representation learning from 12-lead ECG data.

Clinical 12-lead electrocardiography (ECG) is one of the most widely encountered kinds of biosignals. Despite the increased availability of public ECG datasets, label scarcity remains a central challenge in the field. Self-supervised learning represents a promising way to alleviate this issue. This would allow to train more powerful models given the same amount of labeled data and to incorporate or improve predictions about rare diseases, for which training datasets are inherently limited. In this work, we put forward the first comprehensive assessment of self-supervised representation learning from clinical 12-lead ECG data. To this end, we adapt state-of-the-art self-supervised methods based on instance discrimination and latent forecasting to the ECG domain. In a first step, we learn contrastive representations and evaluate their quality based on linear evaluation performance on a recently established, comprehensive, clinical ECG classification task. In a second step, we analyze the impact of self-supervised pretraining on finetuned ECG classifiers as compared to purely supervised performance. For the best-performing method, an adaptation of contrastive predictive coding, we find a linear evaluation performance only 0.5% below supervised performance. For the finetuned models, we find improvements in downstream performance of roughly 1% compared to supervised performance, label efficiency, as well as robustness against physiological noise. This work clearly establishes the feasibility of extracting discriminative representations from ECG data via self-supervised learning and the numerous advantages when finetuning such representations on downstream tasks as compared to purely supervised training. As first comprehensive assessment of its kind in the ECG domain carried out exclusively on publicly available datasets, we hope to establish a first step towards reproducible progress in the rapidly evolving field of representation learning for biosignals.

Generative neural samplers for the quantum Heisenberg chain.

Generative neural samplers offer a complementary approach to Monte Carlo methods for problems in statistical physics and quantum field theory. This paper tests the ability of generative neural samplers to estimate observables for real-world low-dimensional spin systems. It maps out how autoregressive models can sample configurations of a quantum Heisenberg chain via a classical approximation based on the Suzuki-Trotter transformation. We present results for energy, specific heat, and susceptibility for the isotropic XXX and the anisotropic XY chain are in good agreement with Monte Carlo results within the same approximation scheme.

Inferring Respiratory and Circulatory Parameters from Electrical Impedance Tomography With Deep Recurrent Models.

Electrical impedance tomography (EIT) is a noninvasive imaging modality that allows a continuous assessment of changes in regional bioimpedance of different organs. One of its most common biomedical applications is monitoring regional ventilation distribution in critically ill patients treated in intensive care units. In this work, we put forward a proof-of-principle study that demonstrates how one can reconstruct synchronously measured respiratory or circulatory parameters from the EIT image sequence using a deep learning model trained in an end-to-end fashion. For this purpose, we devise an architecture with a convolutional feature extractor whose output is processed by a recurrent neural network. We demonstrate that one can accurately infer absolute volume, absolute flow, normalized airway pressure and within certain limitations even the normalized arterial blood pressure from the EIT signal alone, in a way that generalizes to unseen patients without prior calibration. As an outlook with direct clinical relevance, we furthermore demonstrate the feasibility of reconstructing the absolute transpulmonary pressure from a combination of EIT and absolute airway pressure, as a way to potentially replace the invasive measurement of esophageal pressure. With these results, we hope to stimulate further studies building on the framework put forward in this work.