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.

Association of an impaired GH-IGF-I axis with cardiac wasting in patients with advanced cancer.

BACKGROUND: Growth hormone (GH) resistance is characterized by high GH levels but low levels of insulin-like growth factor-I (IGF-I) and growth hormone binding protein (GHBP) and, for patients with chronic disease, is associated with the development of cachexia. OBJECTIVES: We investigated whether GH resistance is associated with changes in left ventricular (LV) mass (cardiac wasting) in patients with cancer. METHODS: We measured plasma IGF-I, GH, and GHBP in 159 women and 148 men with cancer (83% stage III/IV). Patients were grouped by tertile of echocardiographic LVmass/height2 (women, < 50, 50-61, > 61 g/m2; men, < 60, 60-74, > 74 g/m2) and by presence of wasting syndrome with unintentional weight loss (BMI < 24 kg/m2 and weight loss ≥ 5% in the prior 12 months). Repeat echocardiograms were obtained usually within 3-6 months for 85 patients. RESULTS: Patients in the lowest LVmass/height2 tertile had higher plasma GH (median (IQR) for 1st, 2nd, and 3rd tertile women, 1.8 (0.9-4.2), 0.8 (0.2-2.2), 0.5 (0.3-1.6) ng/mL, p = 0.029; men, 2.1 (0.8-3.2), 0.6 (0.1-1.7), 0.7 (0.2-1.9) ng/mL, p = 0.003). Among women, lower LVmass was associated with higher plasma IGF-I (68 (48-116), 72 (48-95), 49 (35-76) ng/mL, p = 0.007), whereas such association did not exist for men. Patients with lower LVmass had lower log IGF-I/GH ratio (women, 1.60 ± 0.09, 2.02 ± 0.09, 1.88 ± 0.09, p = 0.004; men, 1.64 ± 0.09, 2.14 ± 0.11, 2.04 ± 0.11, p = 0.002). GHBP was not associated with LVmass. Patients with wasting syndrome with unintentional weight loss had higher plasma GH and GHBP, lower log IGF-I/GH ratio, and similar IGF-I. Overall, GHBP correlated inversely with log IGF-I/GH ratio (women, r = - 0.591, p < 0.001; men, r = - 0.575, p < 0.001). Additionally, higher baseline IGF-I was associated with a decline in LVmass during follow-up (r = - 0.318, p = 0.003). CONCLUSION: In advanced cancer, reduced LVmass is associated with increased plasma GH and reduced IGF-I/GH ratio, suggesting increasing GH resistance, especially for patients with wasting syndrome with unintentional weight loss. Higher baseline IGF-I was associated with a decrease in relative LVmass during follow-up.

[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.

Addendum to the German Consensus Recommendations on Ponatinib in the Treatment of Chronic Myeloid Leukemia.

BACKGROUND: Based on the new data from the primary analysis of the OPTIC (Optimizing Ponatinib Treatment in CP-CML) trial on dose optimization of ponatinib in patients with chronic phase (CP)-CML, the German consensus paper on ponatinib published in 2020 (Saussele S et al., Acta Haematol. 2020) has been updated in this addendum. SUMMARY: Focus is on the update of efficacy and safety of ponatinib, reflecting the new data set, as well as the update of the benefit-risk assessment and recommendations for ponatinib starting dose in CP-CML - provided that the decision to use ponatinib has already been made. Furthermore, based on OPTIC and additional empirical data, the expert panel collaborated to develop a decision tree for ponatinib dosing, specifically for intolerant and resistant patients. The recommendations on cardiovascular management have also been updated based on the most recent 2021 guidelines of the European Society of Cardiology (ESC) on cardiovascular disease prevention in clinical practice. KEY MESSAGES: The OPTIC data confirm the high efficacy of ponatinib in patients with CP-CML and provide the basis for individualized dose adjustment during the course of treatment.

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.

Clinical and Prognostic Relevance of Cardiac Wasting in Patients With Advanced Cancer.

BACKGROUND: Body wasting in patients with cancer can affect the heart. OBJECTIVES: The frequency, extent, and clinical and prognostic importance of cardiac wasting in cancer patients is unknown. METHODS: This study prospectively enrolled 300 patients with mostly advanced, active cancer but without significant cardiovascular disease or infection. These patients were compared with 60 healthy control subjects and 60 patients with chronic heart failure (ejection fraction <40%) of similar age and sex distribution. RESULTS: Cancer patients presented with lower left ventricular (LV) mass than healthy control subjects or heart failure patients (assessed by transthoracic echocardiography: 177 ± 47 g vs 203 ± 64 g vs 300 ± 71 g, respectively; P < 0.001). LV mass was lowest in cancer patients with cachexia (153 ± 42 g; P < 0.001). Importantly, the presence of low LV mass was independent of previous cardiotoxic anticancer therapy. In 90 cancer patients with a second echocardiogram after 122 ± 71 days, LV mass had declined by 9.3% ± 1.4% (P < 0.001). In cancer patients with cardiac wasting during follow-up, stroke volume decreased (P < 0.001) and resting heart rate increased over time (P = 0.001). During follow-up of on average 16 months, 149 patients died (1-year all-cause mortality 43%; 95% CI: 37%-49%). LV mass and LV mass adjusted for height squared were independent prognostic markers (both P < 0.05). Adjustment of LV mass for body surface area masked the observed survival impact. LV mass below the prognostically relevant cutpoints in cancer was associated with reduced overall functional status and lower physical performance. CONCLUSIONS: Low LV mass is associated with poor functional status and increased all-cause mortality in cancer. These findings provide clinical evidence of cardiac wasting-associated cardiomyopathy in cancer.

[Alternative treatment strategies for atrial fibrillation : Benefits from yoga?] / Alternative Behandlungsverfahren bei Vorhofflimmern : Was bringt Yoga?

Atrial fibrillation, the most common sustained cardiac arrhythmia, is associated with significant morbidity, mortality, and healthcare utilization. Since the procedures used to treat atrial fibrillation have a number of limitations and risks, there is a growing interest in alternative treatment strategies for patients with atrial fibrillation. One such option is yoga. To date, only a few studies are available on its effect on atrial fibrillation. However, these suggest that yoga may indeed be able to reduce the frequency of the arrhythmia and its progression. The risk factors for atrial fibrillation and quality of life in affected patients are also positively affected. As adverse effects and complications are extremely rare with competent guidance, yoga may already be recommended now. However, further clinical studies are needed to provide recommendations that meet evidence-based criteria.

Detection of Patients with Congenital and Often Concealed Long-QT Syndrome by Novel Deep Learning Models.

INTRODUCTION: The long-QT syndrome (LQTS) is the most common ion channelopathy, typically presenting with a prolonged QT interval and clinical symptoms such as syncope or sudden cardiac death. Patients may present with a concealed phenotype making the diagnosis challenging. Correctly diagnosing at-risk patients is pivotal to starting early preventive treatment. OBJECTIVE: Identification of congenital and often concealed LQTS by utilizing novel deep learning network architectures, which are specifically designed for multichannel time series and therefore particularly suitable for ECG data. DESIGN AND RESULTS: A retrospective artificial intelligence (AI)-based analysis was performed using a 12-lead ECG of genetically confirmed LQTS (n = 124), including 41 patients with a concealed LQTS (33%), and validated against a control cohort (n = 161 of patients) without known LQTS or without QT-prolonging drug treatment but any other cardiovascular disease. The performance of a fully convolutional network (FCN) used in prior studies was compared with a different, novel convolutional neural network model (XceptionTime). We found that the XceptionTime model was able to achieve a higher balanced accuracy score (91.8%) than the associated FCN metric (83.6%), indicating improved prediction possibilities of novel AI architectures. The predictive accuracy prevailed independently of age and QTc parameters. CONCLUSIONS: In this study, the XceptionTime model outperformed the FCN model for LQTS patients with even better results than in prior studies. Even when a patient cohort with cardiovascular comorbidities is used. AI-based ECG analysis is a promising step for correct LQTS patient identification, especially if common diagnostic measures might be misleading.

[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.

Rare Variation in Drug Metabolism and Long QT Genes and the Genetic Susceptibility to Acquired Long QT Syndrome.

BACKGROUND: Acquired long QT syndrome (aLQTS) is a serious unpredictable adverse drug reaction. Pharmacogenomic markers may predict risk. METHODS: Among 153 aLQTS patients (mean age 58 years [range, 14-88], 98.7% White, 85.6% symptomatic), computational methods identified proteins interacting most significantly with 216 QT-prolonging drugs. All cases underwent sequencing of 31 candidate genes arising from this analysis or associating with congenital LQTS. Variants were filtered using a minor allele frequency <1% and classified for susceptibility for aLQTS. Gene-burden analyses were then performed comparing the primary cohort to control exomes (n=452) and an independent replication aLQTS exome sequencing cohort. RESULTS: In 25.5% of cases, at least one rare variant was identified: 22.2% of cases carried a rare variant in a gene associated with congenital LQTS, and in 4% of cases that variant was known to be pathogenic or likely pathogenic for congenital LQTS; 7.8% cases carried a cytochrome-P450 (CYP) gene variant. Of 12 identified CYP variants, 11 (92%) were in an enzyme known to metabolize at least one culprit drug to which the subject had been exposed. Drug-drug interactions that affected culprit drug metabolism were found in 19% of cases. More than one congenital LQTS variant, CYP gene variant, or drug interaction was present in 7.8% of cases. Gene-burden analyses of the primary cohort compared to control exomes (n=452), and an independent replication aLQTS exome sequencing cohort (n=67) and drug-tolerant controls (n=148) demonstrated an increased burden of rare (minor allele frequency<0.01) variants in CYP genes but not LQTS genes. CONCLUSIONS: Rare susceptibility variants in CYP genes are emerging as potentially important pharmacogenomic risk markers for aLQTS and could form part of personalized medicine approaches in the future.

[Second-degree AV block type Mobitz I (Wenckebach) in a 49-year-old man with dilative cardiomyopathy]. / AV-Block 2. Grades Typ Wenckebach bei einem 49-jährigen Patienten mit dilatativer Kardiomyopathie.

A 49-year-old man with dilated cardiomyopathy (left ventricular ejection fraction 50%, unremarkable left ventricular biopsy) developed atrioventricular conduction abnormalities (AV block II type Wenckebach) during exercise testing.

Spontaneous Non-Sustained Ventricular Tachycardia and Premature Ventricular Contractions and Their Prognostic Relevance in Patients with Cancer in Routine Care.

Aims: It is largely unknown whether cancer patients seen in routine care show ventricular arrhythmias in 24 h electrocardiograms (ECGs), and whether when they are detected they carry prognostic relevance. Methods and Results: We included 261 consecutive cancer patients that were referred to the department of cardiology for 24 h ECG examination and 35 healthy controls of similar age and sex in the analysis. To reduce selection bias, cancer patients with known left ventricular ejection fraction <45% were not included in the analysis. Non-sustained ventricular tachycardia (NSVT) episodes of either ≥3 and ≥4 beats duration were more frequent in cancer patients than controls (17% vs. 0%, p = 0.0008; 10% vs. 0%, p = 0.016). Premature ventricular contractions (PVCs)/24 h were not more frequent in cancer patients compared to controls (median (IQR), 26 (2-360) vs. 9 (1-43), p = 0.06; ≥20 PVCs 53% vs. 37%, p = 0.07). During follow-up, (up to 7.2 years, median 15 months) of the cancer patients, 158 (61%) died (1-/3-/5-year mortality rates: 45% [95%CI 39-51%], 66% [95%CI 59-73%], 73% [95%CI 64-82%]). Both non-sustained ventricular tachycardia of ≥4 beats and ≥20 PVCs/24 h independently predicted mortality in univariate and multivariate survival analyses, adjusted for all other univariate predictors of mortality as well as relevant clinical factors, including cancer stage and type, performance status (ECOG), prior potentially cardiotoxic anti-cancer drug therapy, coronary artery disease, potassium concentration, and haemoglobin (multivariate adjusted hazard ratios: NSVT ≥4 beats [HR 1.76, p = 0.022], ≥20 PVCs/24 h [HR 1.63, p < 0.0064]). Conclusions: NSVT ≥4 beats and ≥20 PVCs/day seen in routine 24 h ECGs of patients with cancer carry prognostic relevance.