Real-World and Regulatory Perspectives of Artificial Intelligence in Cardiovascular Imaging

Clearing the Fog: A Scoping Literature Review on the Ethical Issues Surrounding Artificial Intelligence-Based Medical Devices

The use of AI in healthcare has sparked much debate among philosophers, ethicists, regulators and policymakers who raised concerns about the implications of such technologies. The presented scoping review captures the progression of the ethical and legal debate and the proposed ethical frameworks available concerning the use of AI-based medical technologies, capturing key themes across a wide range of medical contexts. The ethical dimensions are synthesised in order to produce a coherent ethical framework for AI-based medical technologies, highlighting how transparency, accountability, confidentiality, autonomy, trust and fairness are the top six recurrent ethical issues. The literature also highlighted how it is essential to increase ethical awareness through interdisciplinary research, such that researchers, AI developers and regulators have the necessary education/competence or networks and tools to ensure proper consideration of ethical matters in the conception and design of new AI technologies and their norms. Interdisciplinarity throughout research, regulation and implementation will help ensure AI-based medical devices are ethical, clinically effective and safe. Achieving these goals will facilitate successful translation of AI into healthcare systems, which currently is lagging behind other sectors, to ensure timely achievement of health benefits to patients and the public.

Device innovation in cardiovascular medicine: a report from the European Society of Cardiology Cardiovascular Round Table

Research performed in Europe has driven cardiovascular device innovation. This includes, but is not limited to, percutaneous coronary intervention, cardiac imaging, transcatheter heart valve implantation, and device therapy of cardiac arrhythmias and heart failure. An important part of future medical progress involves the evolution of medical technology and the ongoing development of artificial intelligence and machine learning. There is a need to foster an environment conducive to medical technology development and validation so that Europe can continue to play a major role in device innovation while providing high standards of safety. This paper summarizes viewpoints on the topic of device innovation in cardiovascular medicine at the European Society of Cardiology Cardiovascular Round Table, a strategic forum for high-level dialogue to discuss issues related to the future of cardiovascular health in Europe. Devices are developed and improved through an iterative process throughout their lifecycle. Early feasibility studies demonstrate proof of concept and help to optimize the design of a device. If successful, this should ideally be followed by randomized clinical trials comparing novel devices vs. accepted standards of care when available and the collection of post-market real-world evidence through registries. Unfortunately, standardized procedures for feasibility studies across various device categories have not yet been implemented in Europe. Cardiovascular imaging can be used to diagnose and characterize patients for interventions to improve procedural results and to monitor devices long term after implantation. Randomized clinical trials often use cardiac imaging-based inclusion criteria, while less frequently trials randomize patients to compare the diagnostic or prognostic value of different modalities. Applications using machine learning are increasingly important, but specific regulatory standards and pathways remain in development in both Europe and the USA. Standards are also needed for smart devices and digital technologies that support device-driven biomonitoring. Changes in device regulation introduced by the European Union aim to improve clinical evidence, transparency, and safety, but they may impact the speed of innovation, access, and availability. Device development programmes including dialogue on unmet needs and advice on study designs must be driven by a community of physicians, trialists, patients, regulators, payers, and industry to ensure that patients have access to innovative care.

Multilevel comparison of deep learning models for function quantification in cardiovascular magnetic resonance: On the redundancy of architectural variations

Background

Cardiac function quantification in cardiovascular magnetic resonance requires precise contouring of the heart chambers. This time-consuming task is increasingly being addressed by a plethora of ever more complex deep learning methods. However, only a small fraction of these have made their way from academia into clinical practice. In the quality assessment and control of medical artificial intelligence, the opaque reasoning and associated distinctive errors of neural networks meet an extraordinarily low tolerance for failure.

Aim

The aim of this study is a multilevel analysis and comparison of the performance of three popular convolutional neural network (CNN) models for cardiac function quantification.

Methods

U-Net, FCN, and MultiResUNet were trained for the segmentation of the left and right ventricles on short-axis cine images of 119 patients from clinical routine. The training pipeline and hyperparameters were kept constant to isolate the influence of network architecture. CNN performance was evaluated against expert segmentations for 29 test cases on contour level and in terms of quantitative clinical parameters. Multilevel analysis included breakdown of results by slice position, as well as visualization of segmentation deviations and linkage of volume differences to segmentation metrics via correlation plots for qualitative analysis.

Results

All models showed strong correlation to the expert with respect to quantitative clinical parameters (rz ' = 0.978, 0.977, 0.978 for U-Net, FCN, MultiResUNet respectively). The MultiResUNet significantly underestimated ventricular volumes and left ventricular myocardial mass. Segmentation difficulties and failures clustered in basal and apical slices for all CNNs, with the largest volume differences in the basal slices (mean absolute error per slice: 4.2 ± 4.5 ml for basal, 0.9 ± 1.3 ml for midventricular, 0.9 ± 0.9 ml for apical slices). Results for the right ventricle had higher variance and more outliers compared to the left ventricle. Intraclass correlation for clinical parameters was excellent (≥0.91) among the CNNs.

Conclusion

Modifications to CNN architecture were not critical to the quality of error for our dataset. Despite good overall agreement with the expert, errors accumulated in basal and apical slices for all models.