Dreaming of electrical waves: Generative modeling of cardiac excitation waves using diffusion models.

Electrical waves in the heart form rotating spiral or scroll waves during life-threatening arrhythmias, such as atrial or ventricular fibrillation. The wave dynamics are typically modeled using coupled partial differential equations, which describe reaction-diffusion dynamics in excitable media. More recently, data-driven generative modeling has emerged as an alternative to generate spatio-temporal patterns in physical and biological systems. Here, we explore denoising diffusion probabilistic models for the generative modeling of electrical wave patterns in cardiac tissue. We trained diffusion models with simulated electrical wave patterns to be able to generate such wave patterns in unconditional and conditional generation tasks. For instance, we explored the diffusion-based (i) parameter-specific generation, (ii) evolution, and (iii) inpainting of spiral wave dynamics, including reconstructing three-dimensional scroll wave dynamics from superficial two-dimensional measurements. Furthermore, we generated arbitrarily shaped bi-ventricular geometries and simultaneously initiated scroll wave patterns inside these geometries using diffusion. We characterized and compared the diffusion-generated solutions to solutions obtained with corresponding biophysical models and found that diffusion models learn to replicate spiral and scroll wave dynamics so well that they could be used for data-driven modeling of excitation waves in cardiac tissue. For instance, an ensemble of diffusion-generated spiral wave dynamics exhibits similar self-termination statistics as the corresponding ensemble simulated with a biophysical model. However, we also found that diffusion models produce artifacts if training data are lacking, e.g., during self-termination, and "hallucinate" wave patterns when insufficiently constrained.

Identifications of Similarity Metrics for Patients With Cancer: Protocol for a Scoping Review.

BACKGROUND: Understanding the similarities of patients with cancer is essential to advancing personalized medicine, improving patient outcomes, and developing more effective and individualized treatments. It enables researchers to discover important patterns, biomarkers, and treatment strategies that can have a significant impact on cancer research and oncology. In addition, the identification of previously successfully treated patients supports oncologists in making treatment decisions for a new patient who is clinically or molecularly similar to the previous patient. OBJECTIVE: The planned review aims to systematically summarize, map, and describe existing evidence to understand how patient similarity is defined and used in cancer research and clinical care. METHODS: To systematically identify relevant studies and to ensure reproducibility and transparency of the review process, a comprehensive literature search will be conducted in several bibliographic databases, including Web of Science, PubMed, LIVIVIVO, and MEDLINE, covering the period from 1998 to February 2024. After the initial duplicate deletion phase, a study selection phase will be applied using Rayyan, which consists of 3 distinct steps: title and abstract screening, disagreement resolution, and full-text screening. To ensure the integrity and quality of the selection process, each of these steps is preceded by a pilot testing phase. This methodological process will culminate in the presentation of the final research results in a structured form according to the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) flowchart. The protocol has been registered in the Journal of Medical Internet Research. RESULTS: This protocol outlines the methodologies used in conducting the scoping review. A search of the specified electronic databases and after removing duplicates resulted in 1183 unique records. As of March 2024, the review process has moved to the full-text evaluation phase. At this stage, data extraction will be conducted using a pretested chart template. CONCLUSIONS: The scoping review protocol, centered on these main concepts, aims to systematically map the available evidence on patient similarity among patients with cancer. By defining the types of data sources, approaches, and methods used in the field, and aligning these with the research questions, the review will provide a foundation for future research and clinical application in personalized cancer care. This protocol will guide the literature search, data extraction, and synthesis of findings to achieve the review's objectives. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): DERR1-10.2196/58705.

Physicians' Attitudes Towards Artificial Intelligence: Results of the PEAK Project.

Artificial intelligence (AI) is rapidly reshaping the medical field. This study aimed to investigate the attitudes of physicians towards AI in medical care using focus groups. Most participants seemed to be open to the use of AI in medicine if transparency in AI systems is ensured, regulatory barriers are addressed and personal contact to the patient is maintained. This qualitative study allows insights into how German physicians perceive the use of AI in medical care. Gaining input from physicians is important when designing future applications of AI for the practical use in medical care.

A Governance Framework for the Implementation and Operation of AI Applications in a University Hospital.

BACKGROUND: Artificial intelligence (AI) is becoming increasingly important in everyday life and medical care with a notable gap between AI development in medicine there and its practical implementation in university hospitals. OBJECTIVE: The aim was to develop a governance framework to guide the procurement and implementation of AI applications in university hospitals. METHODS: Seven role-play expert interviews were conducted with interviewees from two university hospitals, simulating realistic AI system implementation scenarios. RESULTS: A detailed governance framework was developed, divided into the sections general considerations, system selection criteria, and implementation. Further, a condensed version of the checklist was also derived. CONCLUSION: Many aspects of AI applications in medical care, such as the establishment of an AI board, remain, along with numerous concerns about the technology. These findings provide valuable insights into the topic.

Dreaming of Electrical Waves: Generative Modeling of Cardiac Excitation Waves using Diffusion Models.

Electrical waves in the heart form rotating spiral or scroll waves during life-threatening arrhythmias such as atrial or ventricular fibrillation. The wave dynamics are typically modeled using coupled partial differential equations, which describe reaction-diffusion dynamics in excitable media. More recently, data-driven generative modeling has emerged as an alternative to generate spatio-temporal patterns in physical and biological systems. Here, we explore denoising diffusion probabilistic models for the generative modeling of electrical wave patterns in cardiac tissue. We trained diffusion models with simulated electrical wave patterns to be able to generate such wave patterns in unconditional and conditional generation tasks. For instance, we explored the diffusion-based i) parameter-specific generation, ii) evolution and iii) inpainting of spiral wave dynamics, including reconstructing three-dimensional scroll wave dynamics from superficial two-dimensional measurements. Further, we generated arbitrarily shaped bi-ventricular geometries and simultaneously initiated scroll wave patterns inside these geometries using diffusion. We characterized and compared the diffusion-generated solutions to solutions obtained with corresponding biophysical models and found that diffusion models learn to replicate spiral and scroll waves dynamics so well that they could be used for data-driven modeling of excitation waves in cardiac tissue. For instance, an ensemble of diffusion-generated spiral wave dynamics exhibits similar self-termination statistics as the corresponding ensemble simulated with a biophysical model. However, we also found that diffusion models {produce artifacts if training data is lacking, e.g. during self-termination,} and `hallucinate' wave patterns when insufficiently constrained.

Individualized neoantigen peptide immunization of a metastatic pancreatic cancer patient: a case report of combined tumor and liquid biopsy.

This report details a case of pancreatic cancer with liver metastasis that exhibited a positive immune response to personalized immunization therapy. Our study involved the identification of neoantigens and their corresponding immunogenic peptides using an in-house bioinformatic pipeline. This process included the identification of somatic mutations through DNA/RNA sequencing of solid tumor tissue and blood liquid biopsy. Computational prediction techniques were then employed to identify novel epitopes, followed by the design and manufacture of patient-specific immunization peptides. In combination with standard-of-care chemotherapy, the patient received a sequence of 5 biweekly prime injections followed by 2 boost injections 2 and 5 months later. The peptides were emulsified in Montanide and the injection-site was conditioned with nivolumab and imiquimod. The combined regimen of peptide immunization and chemotherapy resulted in a notable decline in CA19-9 tumor marker levels following both prime and boost applications. Subsequent MRI assessments revealed a reduction in the size of liver metastases several months post-immunization initiation. Importantly, the patient showed and improved overall survival and reported an improved quality of life without experiencing significant treatment-related adverse effects. This case underscores the potential benefits of personalized peptide-based immunization as an adjunctive therapy in the treatment of advanced pancreatic cancer, showcasing promising outcomes in tumor marker reduction, tumor shrinkage, and enhanced patient well-being.

Data Visualization Support for Tumor Boards and Clinical Oncology: Protocol for a Scoping Review.

BACKGROUND: Complex and expanding data sets in clinical oncology applications require flexible and interactive visualization of patient data to provide the maximum amount of information to physicians and other medical practitioners. Interdisciplinary tumor conferences in particular profit from customized tools to integrate, link, and visualize relevant data from all professions involved. OBJECTIVE: The scoping review proposed in this protocol aims to identify and present currently available data visualization tools for tumor boards and related areas. The objective of the review will be to provide not only an overview of digital tools currently used in tumor board settings, but also the data included, the respective visualization solutions, and their integration into hospital processes. METHODS: The planned scoping review process is based on the Arksey and O'Malley scoping study framework. The following electronic databases will be searched for articles published in English: PubMed, Web of Knowledge, and SCOPUS. Eligible articles will first undergo a deduplication step, followed by the screening of titles and abstracts. Second, a full-text screening will be used to reach the final decision about article selection. At least 2 reviewers will independently screen titles, abstracts, and full-text reports. Conflicting inclusion decisions will be resolved by a third reviewer. The remaining literature will be analyzed using a data extraction template proposed in this protocol. The template includes a variety of meta information as well as specific questions aiming to answer the research question: "What are the key features of data visualization solutions used in molecular and organ tumor boards, and how are these elements integrated and used within the clinical setting?" The findings will be compiled, charted, and presented as specified in the scoping study framework. Data for included tools may be supplemented with additional manual literature searches. The entire review process will be documented in alignment with the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) flowchart. RESULTS: The results of this scoping review will be reported per the expanded PRISMA-ScR guidelines. A preliminary search using PubMed, Web of Knowledge, and Scopus resulted in 1320 articles after deduplication that will be included in the further review process. We expect the results to be published during the second quarter of 2024. CONCLUSIONS: Visualization is a key process in leveraging a data set's potentially available information and enabling its use in an interdisciplinary setting. The scoping review described in this protocol aims to present the status quo of visualization solutions for tumor board and clinical oncology applications and their integration into hospital processes. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): DERR1-10.2196/53627.

German primary care data collection projects: a scoping review.

BACKGROUND: The widespread use of electronic health records (EHRs) has led to a growing number of large routine primary care data collection projects globally, making these records a valuable resource for health services and epidemiological and clinical research. This scoping review aims to comprehensively assess and compare strengths and limitations of all German primary care data collection projects and relevant research publications that extract data directly from practice management systems (PMS). METHODS: A literature search was conducted in the electronic databases in May 2021 and in June 2022. The search string included terms related to general practice, routine data, and Germany. The retrieved studies were classified as applied studies and methodological studies, and categorised by type of research, subject area, sample of publications, disease category, or main medication analysed. RESULTS: A total of 962 references were identified, with 241 studies included from six German projects in which databases are populated by EHRs from PMS. The projects exhibited significant heterogeneity in terms of size, data collection methods, and variables collected. The majority of the applied studies (n = 205, 85%) originated from one database with a primary focus on pharmacoepidemiological topics (n = 127, 52%) including prescription patterns (n = 68, 28%) and studies about treatment outcomes, compliance, and treatment effectiveness (n = 34, 14%). Epidemiological studies (n = 77, 32%) mainly focused on incidence and prevalence studies (n = 41, 17%) and risk and comorbidity analysis studies (n = 31, 12%). Only 10% (n = 23) of studies were in the field of health services research, such as hospitalisation. CONCLUSION: The development and durability of primary care data collection projects in Germany is hindered by insufficient public funding, technical issues of data extraction, and strict data protection regulations. There is a need for further research and collaboration to improve the usability of EHRs for health services and research.

Extending cBioPortal for Therapy Recommendation Documentation in Molecular Tumor Boards: Development and Usability Study.

BACKGROUND: In molecular tumor boards (MTBs), patients with rare or advanced cancers are discussed by a multidisciplinary team of health care professionals. Software support for MTBs is lacking; in particular, tools for preparing and documenting MTB therapy recommendations need to be developed. OBJECTIVE: We aimed to implement an extension to cBioPortal to provide a tool for the documentation of therapy recommendations from MTB sessions in a secure and standardized manner. The developed extension should be embedded in the patient view of cBioPortal to enable easy documentation during MTB sessions. The resulting architecture for storing therapy recommendations should be integrable into various hospital information systems. METHODS: On the basis of a requirements analysis and technology analysis for authentication techniques, a prototype was developed and iteratively refined through a user-centered development process. In conclusion, the tool was evaluated via a usability evaluation, including interviews, structured questionnaires, and the System Usability Scale. RESULTS: The patient view of cBioPortal was extended with a new tab that enables users to document MTB sessions and therapy recommendations. The role-based access control was expanded to allow for a finer distinction among the rights to view, edit, and delete data. The usability evaluation showed overall good usability and a System Usability Scale score of 83.57. CONCLUSIONS: This study demonstrates how cBioPortal can be extended to not only visualize MTB patient data but also be used as a documentation platform for therapy recommendations.

Demand Analysis of a German Emergency Medical Service Feedback System.

BACKGROUND: The number of emergency medical service (EMS) calls in Germany is continuously increasing. The initial assessment, the pre-hospital care and the choice of hospital for further care by the EMS influences the patient's outcome and are the basis for further care in hospital. However, the EMS does not receive any official feedback on its decisions. OBJECTIVES: This study evaluates the demand for a feedback system from the emergency department (ED) to the EMS, what it should contain, and how it could be integrated in the electronic clinical systems. METHODS: A semi-structured interview guideline for expert interviews with members of EMS staff (n = 6) and ED staff (n = 17) was developed. A mockup to visualise a possible implementation was designed and included in the interview. RESULTS: There is a significant demand for feedback on pre-diagnosis, pre-hospital care and handover of patients from the EMS to the ED. The EDs are very interested in improving the collaboration with the paramedic services through feedback. CONCLUSION: A feedback system is strongly desired by various EMS stakeholders and, according to them, could improve both EMS and ED collaboration and overall patient care.

Usability Analysis of a Medication Visualization Tool for Decision Support.

BACKGROUND: In Germany, patients are entitled to a medication plan. While the overview is useful, it does not contain explicit information on various potential adverse drug events (ADEs). Therefore, physicians must continue to seek information from various sources to ensure medication safety. OBJECTIVE: In this project a first functional prototype of a medication therapy tool was developed that focuses on visualizing and highlighting potential ADEs. A usability analysis about the tool's functionality, design and usability was conducted. METHODS: A web application tool was developed using the MMI Pharmindex as database. ADEs are color coded and can be displayed in three different ways - as a list, a table, or a graph. To test the tool, an online survey was conducted amongst healthcare professionals (n = 9). The test included two real medication plans to check ADEs through the tool. RESULTS: The survey results indicated that the web tool was clear and self-explanatory. It scored overall "good" (score: 76.5) on the System Usability Scale questionnaire. Due to the free-text information of the database used, there were some inconsistencies in the visualized ADEs. CONCLUSION: There is a demand for a visualization tool for medications. The high quality of the database is crucial in order to correctly visualize all necessary information, such as drug-drug interactions and inclusion of patient data. This is essential to provide a trustworthy tool for medical professionals.

MIRACUM-Pipe: An Adaptable Pipeline for Next-Generation Sequencing Analysis, Reporting, and Visualization for Clinical Decision Making.

(1) Background: Next-generation sequencing (NGS) of patients with advanced tumors is becoming an established method in Molecular Tumor Boards. However, somatic variant detection, interpretation, and report generation, require in-depth knowledge of both bioinformatics and oncology. (2) Methods: MIRACUM-Pipe combines many individual tools into a seamless workflow for comprehensive analyses and annotation of NGS data including quality control, alignment, variant calling, copy number variation estimation, evaluation of complex biomarkers, and RNA fusion detection. (3) Results: MIRACUM-Pipe offers an easy-to-use, one-prompt standardized solution to analyze NGS data, including quality control, variant calling, copy number estimation, annotation, visualization, and report generation. (4) Conclusions: MIRACUM-Pipe, a versatile pipeline for NGS, can be customized according to bioinformatics and clinical needs and to support clinical decision-making with visual processing and interactive reporting.

Improved Survival of a HER2-Positive Metastatic Breast Cancer Patient Following a Personalized Peptide Immunization.

Cancer neoantigens that arise from somatic mutations have emerged as important targets for personalized immunization. Here, we report an improved overall survival of a HER2-positive metastatic breast cancer patient using a bioinformatic-based personalized peptide immunization called BITAP (BioInformatic Tumor Address Peptides). The epitopes were predicted using our in-house bioinformatic pipeline, and the immunogenicity was tested by IFN-γ ELISPOT and intracellular cytokine staining assays. In total, a significant peptide-specific T-cell response was detected against 18 out of the 76 (≈24%) tested peptides. The patient's follow-up by measuring serologic markers showed a significant reduction in the tumor marker levels following BITAP immunization. Along with standard treatment, the patient treated with the BITAP showed stable disease with a remarkably improved overall survival, and no serious treatment-related adverse effects. In conclusion, our findings suggest that BITAP immunization is feasible, and safe, and may induce tumor regressions in patients with HER2-positive subsets of breast cancer.

The Contribution of the Nrf2/ARE System to Mechanotransduction in Musculoskeletal and Periodontal Tissues.

Mechanosensing plays an essential role in maintaining tissue functions. Across the human body, several tissues (i.e., striated muscles, bones, tendons, ligaments, as well as cartilage) require mechanical loading to exert their physiological functions. Contrary, mechanical unloading triggers pathological remodeling of these tissues and, consequently, human body dysfunctions. At the cellular level, both mechanical loading and unloading regulate a wide spectrum of cellular pathways. Among those, pathways regulated by oxidants such as reactive oxygen species (ROS) represent an essential node critically controlling tissue organization and function. Hence, a sensitive balance between the generation and elimination of oxidants keeps them within a physiological range. Here, the Nuclear Factor-E2-related factor 2/Antioxidant response element (Nrf2/ARE) system plays an essential role as it constitutes the major cellular regulation against exogenous and endogenous oxidative stresses. Dysregulations of this system advance, i.a., liver, neurodegenerative, and cancer diseases. Herein, we extend our comprehension of the Nrf2 system to the aforementioned mechanically sensitive tissues to explore its role in their physiology and pathology. We demonstrate the relevance of it for the tissues' functionality and highlight the imperative to further explore the Nrf2 system to understand the physiology and pathology of mechanically sensitive tissues in the context of redox biology.

[BeoNet-Halle-development of a multifunctional database for the automated extraction of healthcare data from general practitioner and specialist practices]. / BeoNet-Halle ­ Aufbau einer multifunktionalen Datenbank zur automatisierten Extraktion von Versorgungsdaten aus Haus- und Facharztpraxen.

The Beobachtungspraxennetzwerk Halle (BeoNet-Halle) is an innovative database of outpatient care that has been collecting patient data from participating primary care and specialty practices throughout Germany since 2020 and making it available for research and care. The database is set up and maintained by the Institute of Medical Epidemiology, Biometrics and Informatics and the Institute of General Practice and Family Medicine of the Martin Luther University Halle-Wittenberg. Furthermore, the Data Integration Center of the University Medical Center Halle is involved in the project. In principle, anonymized and pseudonymized patient data from all commercially available practice management systems should flow into the databases.In this article, we describe the structure and methods of the multi-purpose database BeoNet and quantify the current data stock. The workflow of collection, transfer, and storage of broad consents is described and advantages and limitations of the database are discussed.BeoNet-Halle currently contains anonymized data of approximately 73,043 patients from five physician practices. Furthermore, it includes data from more than 2,653,437 ICD-10 diagnoses, 1,403,726 prescriptions, and 1,894,074 laboratory results. Pseudonymized data were successfully exported from 481 patients.BeoNet-Halle enables an almost seamless representation of the care provided in the participating practices. In the future, the database will map patient treatment pathways across practices and provide high-quality care data to contribute to health policy decision-making and optimization of care processes.

The application of gradient dose segmented analysis of in-vivo EPID images for patients undergoing VMAT in a resource-constrained environment.

The gamma analysis metric is a commonly used metric for VMAT plan evaluation. The major drawback of this is the lack of correlation between gamma passing rates and DVH values. The novel GDSAmean metric was developed by Steers et al. to quantify changes in the PTV mean dose (Dmean ) for VMAT patients. The aim of this work is to apply the GDSA retrospectively on head-and-neck cancer patients treated on the newly acquired Varian Halcyon, to assess changes in GDSAmean , and to evaluate the cause of day-to-day changes in the time-plot series. In-vivo EPID transmission images of head-and-neck cancer patients treated between August 2019 and July 2020 were analyzed retrospectively. The GDSAmean was determined for all patients treated. The changes in patient anatomy and rotational errors were quantified using the daily CBCT images and added to a time-plot with the daily change in GDSAmean . Over 97% of the delivered treatment fractions had a GDSAmean  < 3%. Thirteen of the patients received at least one treatment fraction where the GDSAmean  > 3%. Most of these deviations occurred for the later fractions of radiotherapy treatment. Additionally, 92% of these patients were treated for malignancies involving the larynx and oropharynx. Notable deviations in the effective separation diameters were observed for 62% of the patients where the change in GDSAmean  > 3%. For the other five cases with GDSAmean  < 3%, the mean pitch, roll, and yaw rotational errors were 0.90°, 0.45°, and 0.43°, respectively. A GDSAmean  > 3% was more likely due to a change in separation, whereas a GDSAmean  < 3% was likely caused by rotational errors. Pitch errors were shown to be the most dominant. The GDSAmean is easily implementable and can aid in scheduling new CT scans for patients before significant deviations in dose delivery occur.

Nrf2 Activation Does Not Protect from Aldosterone-Induced Kidney Damage in Mice.

Nuclear factor erythroid 2-related factor 2 (Nrf2) is downregulated in chronic kidney disease (CKD). Activation of Nrf2 might be a therapeutic option in CKD. Here we investigate the effect of Nrf2 activation on aldosterone (Aldo)-induced renal injury. Wild-type (WT) mice, transgenic Keap1 hypomorphic (Nrf2ꜛ, genotype results in upregulation of Nrf2 expression) mice and WT mice treated with the Nrf2 activator sulforaphane (Sulf) received Aldo for 4 weeks. In Aldo-treated mice, kidneys were significantly heavier and pathologically altered, reflected by increased urinary albumin levels and tissue damage. In Nrf2ꜛ-Aldo mice the tubule damage marker NGAL was significantly decreased. Increased oxidative damage markers (8-OHdG, 15-isoprostane F2t) were measured in all Aldo-treated groups. Aldo-increased Nrf2 amounts were mainly found in the late tubule system. The amount of phosphorylated and thus putatively active Nrf2 was significantly increased by Aldo only in WT mice. However, expression of Nrf2 target genes NQO1 and HO1 was decreased in all Aldo-infused mice. GSK3ß, which promotes Nrf2 degradation, was significantly increased in the kidneys of Aldo-treated WT mice. Neither genetic nor pharmacological Nrf2 activation was able to prevent oxidative injury induced by Aldo, probably due to induction of negative regulators of Nrf2.

Optical mapping of contracting hearts.

Optical mapping is a widely used tool to record and visualize the electrophysiological properties in a variety of myocardial preparations such as Langendorff-perfused isolated hearts, coronary-perfused wedge preparations, and cell culture monolayers. Motion artifact originating from the mechanical contraction of the myocardium creates a significant challenge to performing optical mapping of contracting hearts. Hence, to minimize the motion artifact, cardiac optical mapping studies are mostly performed on non-contracting hearts, where the mechanical contraction is removed using pharmacological excitation-contraction uncouplers. However, such experimental preparations eliminate the possibility of electromechanical interaction, and effects such as mechano-electric feedback cannot be studied. Recent developments in computer vision algorithms and ratiometric techniques have opened the possibility of performing optical mapping studies on isolated contracting hearts. In this review, we discuss the existing techniques and challenges of optical mapping of contracting hearts.

Reconstruction of three-dimensional scroll waves in excitable media from two-dimensional observations using deep neural networks.

Scroll wave dynamics are thought to underlie life-threatening ventricular fibrillation. However, direct observations of three-dimensional electrical scroll waves remain elusive, as there is no direct way to measure action potential wave patterns transmurally throughout the thick ventricular heart muscle. Here we study whether it is possible to reconstruct simulated scroll waves and scroll wave chaos using deep learning. We trained encoding-decoding convolutional neural networks to predict three-dimensional scroll wave dynamics inside bulk-shaped excitable media from two-dimensional observations of the wave dynamics on the bulk's surface. We tested whether observations from one or two opposing surfaces would be sufficient and whether transparency or measurements of surface deformations enhances the reconstruction. Further, we evaluated the approach's robustness against noise and tested the feasibility of predicting the bulk's thickness. We distinguished isotropic and anisotropic, as well as opaque and transparent, excitable media as models for cardiac tissue and the Belousov-Zhabotinsky chemical reaction, respectively. While we demonstrate that it is possible to reconstruct three-dimensional scroll wave dynamics, we also show that it is challenging to reconstruct complicated scroll wave chaos and that prediction outcomes depend on various factors such as transparency, anisotropy, and ultimately the thickness of the medium compared to the size of the scroll waves. In particular, we found that anisotropy provides crucial information for neural networks to decode depth, which facilitates the reconstructions. In the future, deep neural networks could be used to visualize intramural action potential wave patterns from epi- or endocardial measurements.