Search

EU-US data transfers for health research remain a particularly thorny issue in view of the stringent rules of the EU General Data Protection Regulation (GDPR) and the challenges related to US mass surveillance programs, particularly the manner in which US law enforcement and national security agencies can access personal data originating from the EU. Since the entry into force of the GDPR, evidence of impeded collaborations is increasing, particularly in the case of sharing data with US public institutions. The adoption of a new EU-US adequacy decision in July 2023 does not hold the promise for a long-lasting solution due to the risks of being challenged and invalidated – yet again – at the Court of Justice of the EU. As the research community is calling for answers, the new proposal for a European Health Data Space regulation may hold a key to solving some of the existing issues. In this paper, we critically discuss the current rules and outline a possible way forward for transfers between public bodies. [ABSTRACT FROM AUTHOR]

The rapid advancements in artificial intelligence (AI) have led to the development of sophisticated large language models (LLMs) such as GPT-4 and Bard. The potential implementation of LLMs in healthcare settings has already garnered considerable attention because of their diverse applications that include facilitating clinical documentation, obtaining insurance pre-authorization, summarizing research papers, or working as a chatbot to answer questions for patients about their specific data and concerns. While offering transformative potential, LLMs warrant a very cautious approach since these models are trained differently from AI-based medical technologies that are regulated already, especially within the critical context of caring for patients. The newest version, GPT-4, that was released in March, 2023, brings the potentials of this technology to support multiple medical tasks; and risks from mishandling results it provides to varying reliability to a new level. Besides being an advanced LLM, it will be able to read texts on images and analyze the context of those images. The regulation of GPT-4 and generative AI in medicine and healthcare without damaging their exciting and transformative potential is a timely and critical challenge to ensure safety, maintain ethical standards, and protect patient privacy. We argue that regulatory oversight should assure medical professionals and patients can use LLMs without causing harm or compromising their data or privacy. This paper summarizes our practical recommendations for what we can expect from regulators to bring this vision to reality. [ABSTRACT FROM AUTHOR]

Suturing skill scores have demonstrated strong predictive capabilities for patient functional recovery. The suturing can be broken down into several substep components, including needle repositioning, needle entry angle, etc. Artificial intelligence (AI) systems have been explored to automate suturing skill scoring. Traditional approaches to skill assessment typically focus on evaluating individual sub-skills required for particular substeps in isolation. However, surgical procedures require the integration and coordination of multiple sub-skills to achieve successful outcomes. Significant associations among the technical sub-skill have been established by existing studies. In this paper, we propose a framework for joint skill assessment that takes into account the interconnected nature of sub-skills required in surgery. The prior known relationships among sub-skills are firstly identified. Our proposed AI system is then empowered by the prior known relationships to perform the suturing skill scoring for each sub-skill domain simultaneously. Our approach can effectively improve skill assessment performance through the prior known relationships among sub-skills. Through the proposed approach to joint skill assessment, we aspire to enhance the evaluation of surgical proficiency and ultimately improve patient outcomes in surgery. [ABSTRACT FROM AUTHOR]

This paper examines the ethical and legal challenges encountered during the GATEKEEPER Project and how these challenges informed the development of a comprehensive framework for future Large-Scale Pilot (LSP) projects. GATEKEEPER is a LSP Project with 48 partners conducting 30 implementation studies across Europe with 50,000 target participants grouped into 9 Reference Use Cases. The project underscored the complexity of obtaining ethical approval across various jurisdictions with divergent regulations and procedures. Through a detailed analysis of the issues faced and the strategies employed to navigate these challenges, this study proposes an ethical and legal framework. This framework, derived from a comparative analysis of ethical application forms and regulations, aims to streamline the ethical approval process for future LSP research projects. By addressing the hurdles encountered in GATEKEEPER, the proposed framework offers a roadmap for more efficient and effective project management, ensuring smoother implementation of similar projects in the future. [ABSTRACT FROM AUTHOR]

Recent developments in deep learning have enabled data-driven algorithms that can reach human-level performance and beyond. The development and deployment of medical image analysis methods have several challenges, including data heterogeneity due to population diversity and different device manufacturers. In addition, more input from experts is required for a reliable method development process. While the exponential growth in clinical imaging data has enabled deep learning to flourish, data heterogeneity, multi-modality, and rare or inconspicuous disease cases still need to be explored. Endoscopy being highly operator-dependent with grim clinical outcomes in some disease cases, reliable and accurate automated system guidance can improve patient care. Most designed methods must be more generalisable to the unseen target data, patient population variability, and variable disease appearances. The paper reviews recent works on endoscopic image analysis with artificial intelligence (AI) and emphasises the current unmatched needs in this field. Finally, it outlines the future directions for clinically relevant complex AI solutions to improve patient outcomes. [ABSTRACT FROM AUTHOR]

Machine learning and deep learning are two subsets of artificial intelligence that involve teaching computers to learn and make decisions from any sort of data. Most recent developments in artificial intelligence are coming from deep learning, which has proven revolutionary in almost all fields, from computer vision to health sciences. The effects of deep learning in medicine have changed the conventional ways of clinical application significantly. Although some sub-fields of medicine, such as pediatrics, have been relatively slow in receiving the critical benefits of deep learning, related research in pediatrics has started to accumulate to a significant level, too. Hence, in this paper, we review recently developed machine learning and deep learning-based solutions for neonatology applications. We systematically evaluate the roles of both classical machine learning and deep learning in neonatology applications, define the methodologies, including algorithmic developments, and describe the remaining challenges in the assessment of neonatal diseases by using PRISMA 2020 guidelines. To date, the primary areas of focus in neonatology regarding AI applications have included survival analysis, neuroimaging, analysis of vital parameters and biosignals, and retinopathy of prematurity diagnosis. We have categorically summarized 106 research articles from 1996 to 2022 and discussed their pros and cons, respectively. In this systematic review, we aimed to further enhance the comprehensiveness of the study. We also discuss possible directions for new AI models and the future of neonatology with the rising power of AI, suggesting roadmaps for the integration of AI into neonatal intensive care units. [ABSTRACT FROM AUTHOR]

Potential benefits of precision medicine in cardiovascular disease (CVD) include more accurate phenotyping of individual patients with the same condition or presentation, using multiple clinical, imaging, molecular and other variables to guide diagnosis and treatment. An approach to realising this potential is the digital twin concept, whereby a virtual representation of a patient is constructed and receives real-time updates of a range of data variables in order to predict disease and optimise treatment selection for the real-life patient. We explored the term digital twin, its defining concepts, the challenges as an emerging field, and potentially important applications in CVD. A mapping review was undertaken using a systematic search of peer-reviewed literature. Industry-based participants and patent applications were identified through web-based sources. Searches of Compendex, EMBASE, Medline, ProQuest and Scopus databases yielded 88 papers related to cardiovascular conditions (28%, n = 25), non-cardiovascular conditions (41%, n = 36), and general aspects of the health digital twin (31%, n = 27). Fifteen companies with a commercial interest in health digital twin or simulation modelling had products focused on CVD. The patent search identified 18 applications from 11 applicants, of which 73% were companies and 27% were universities. Three applicants had cardiac-related inventions. For CVD, digital twin research within industry and academia is recent, interdisciplinary, and established globally. Overall, the applications were numerical simulation models, although precursor models exist for the real-time cyber-physical system characteristic of a true digital twin. Implementation challenges include ethical constraints and clinical barriers to the adoption of decision tools derived from artificial intelligence systems. [ABSTRACT FROM AUTHOR]

While the literature on putting a "human in the loop" in artificial intelligence (AI) and machine learning (ML) has grown significantly, limited attention has been paid to how human expertise ought to be combined with AI/ML judgments. This design question arises because of the ubiquity and quantity of algorithmic decisions being made today in the face of widespread public reluctance to forgo human expert judgment. To resolve this conflict, we propose that human expert judges be included via appeals processes for review of algorithmic decisions. Thus, the human intervenes only in a limited number of cases and only after an initial AI/ML judgment has been made. Based on an analogy with appellate processes in judiciary decision-making, we argue that this is, in many respects, a more efficient way to divide the labor between a human and a machine. Human reviewers can add more nuanced clinical, moral, or legal reasoning, and they can consider case-specific information that is not easily quantified and, as such, not available to the AI/ML at an initial stage. In doing so, the human can serve as a crucial error correction check on the AI/ML, while retaining much of the efficiency of AI/ML's use in the decision-making process. In this paper, we develop these widely applicable arguments while focusing primarily on examples from the use of AI/ML in medicine, including organ allocation, fertility care, and hospital readmission. [ABSTRACT FROM AUTHOR]

This paper reviews the current state of patient safety and the application of artificial intelligence (AI) techniques to patient safety. This paper defines patient safety broadly, not just inpatient care but across the continuum of care, including diagnostic errors, misdiagnosis, adverse events, injuries, and measurement issues. It outlines the major current uses of AI in patient safety and the relative adoption of these techniques in hospitals and health systems. It also outlines some of the limitations of these AI systems and the challenges with evaluation of these systems. Finally, it outlines the importance of developing a proactive agenda for AI in healthcare that includes marked increased funding of research and evaluation in this area. [ABSTRACT FROM AUTHOR]

Digital Health Technologies (DHTs) are being applied in a widening range of scenarios in medicine. We describe the emerging phenomenon of the grouping of individual DHTs, with a clinical use case and regulatory approval in their own right, into packages to perform specific clinical tasks in defined settings. Example groupings include suites of devices for remote monitoring, or for smart clinics. In this first article of a two-article series, we describe challenges in implementation and limitations in frameworks for the regulation, health technology assessment, and reimbursement of these device suites and linked novel care pathways. [ABSTRACT FROM AUTHOR]

In machine learning (ML), association patterns in the data, paths in decision trees, and weights between layers of the neural network are often entangled due to multiple underlying causes, thus masking the pattern-to-source relation, weakening prediction, and defying explanation. This paper presents a revolutionary ML paradigm: pattern discovery and disentanglement (PDD) that disentangles associations and provides an all-in-one knowledge system capable of (a) disentangling patterns to associate with distinct primary sources; (b) discovering rare/imbalanced groups, detecting anomalies and rectifying discrepancies to improve class association, pattern and entity clustering; and (c) organizing knowledge for statistically supported interpretability for causal exploration. Results from case studies have validated such capabilities. The explainable knowledge reveals pattern-source relations on entities, and underlying factors for causal inference, and clinical study and practice; thus, addressing the major concern of interpretability, trust, and reliability when applying ML to healthcare, which is a step towards closing the AI chasm. [ABSTRACT FROM AUTHOR]

Explainable artificial intelligence (XAI) has emerged as a promising solution for addressing the implementation challenges of AI/ML in healthcare. However, little is known about how developers and clinicians interpret XAI and what conflicting goals and requirements they may have. This paper presents the findings of a longitudinal multi-method study involving 112 developers and clinicians co-designing an XAI solution for a clinical decision support system. Our study identifies three key differences between developer and clinician mental models of XAI, including opposing goals (model interpretability vs. clinical plausibility), different sources of truth (data vs. patient), and the role of exploring new vs. exploiting old knowledge. Based on our findings, we propose design solutions that can help address the XAI conundrum in healthcare, including the use of causal inference models, personalized explanations, and ambidexterity between exploration and exploitation mindsets. Our study highlights the importance of considering the perspectives of both developers and clinicians in the design of XAI systems and provides practical recommendations for improving the effectiveness and usability of XAI in healthcare. [ABSTRACT FROM AUTHOR]

Pain is a complex and personal experience that presents diverse measurement challenges. Different sensing technologies can be used as a surrogate measure of pain to overcome these challenges. The objective of this review is to summarise and synthesise the published literature to: (a) identify relevant non-invasive physiological sensing technologies that can be used for the assessment of human pain, (b) describe the analytical tools used in artificial intelligence (AI) to decode pain data collected from sensing technologies, and (c) describe the main implications in the application of these technologies. A literature search was conducted in July 2022 to query PubMed, Web of Sciences, and Scopus. Papers published between January 2013 and July 2022 are considered. Forty-eight studies are included in this literature review. Two main sensing technologies (neurological and physiological) are identified in the literature. The sensing technologies and their modality (unimodal or multimodal) are presented. The literature provided numerous examples of how different analytical tools in AI have been applied to decode pain. This review identifies different non-invasive sensing technologies, their analytical tools, and the implications for their use. There are significant opportunities to leverage multimodal sensing and deep learning to improve accuracy of pain monitoring systems. This review also identifies the need for analyses and datasets that explore the inclusion of neural and physiological information together. Finally, challenges and opportunities for designing better systems for pain assessment are also presented. [ABSTRACT FROM AUTHOR]

The increasing need for mental health support and a shortage of therapists have led to the development of the eXtended-reality Artificial Intelligence Assistant (XAIA). This platform combines spatial computing, virtual reality (VR), and artificial intelligence (AI) to provide immersive mental health support. Utilizing GPT-4 for AI-driven therapy, XAIA engaged participants with mild-to-moderate anxiety or depression in biophilic VR environments. Speaking with an AI therapy avatar in VR was considered acceptable, helpful, and safe, with participants observed to engage genuinely with the program. However, some still favored human interaction and identified shortcomings with using a digital VR therapist. The study provides initial evidence of the acceptability and safety of AI psychotherapy via spatial computing, warranting further research on technical enhancements and clinical impact. [ABSTRACT FROM AUTHOR]

The increasing need for mental health support and a shortage of therapists have led to the development of the eXtended-reality Artificial Intelligence Assistant (XAIA). This platform combines spatial computing, virtual reality (VR), and artificial intelligence (AI) to provide immersive mental health support. Utilizing GPT-4 for AI-driven therapy, XAIA engaged participants with mild-to-moderate anxiety or depression in biophilic VR environments. Speaking with an AI therapy avatar in VR was considered acceptable, helpful, and safe, with participants observed to engage genuinely with the program. However, some still favored human interaction and identified shortcomings with using a digital VR therapist. The study provides initial evidence of the acceptability and safety of AI psychotherapy via spatial computing, warranting further research on technical enhancements and clinical impact. [ABSTRACT FROM AUTHOR]

In low- and middle-income countries (LMICs), AI has been promoted as a potential means of strengthening healthcare systems by a growing number of publications. We aimed to evaluate the scope and nature of AI technologies in the specific context of LMICs. In this systematic scoping review, we used a broad variety of AI and healthcare search terms. Our literature search included records published between 1st January 2009 and 30th September 2021 from the Scopus, EMBASE, MEDLINE, Global Health and APA PsycInfo databases, and grey literature from a Google Scholar search. We included studies that reported a quantitative and/or qualitative evaluation of a real-world application of AI in an LMIC health context. A total of 10 references evaluating the application of AI in an LMIC were included. Applications varied widely, including: clinical decision support systems, treatment planning and triage assistants and health chatbots. Only half of the papers reported which algorithms and datasets were used in order to train the AI. A number of challenges of using AI tools were reported, including issues with reliability, mixed impacts on workflows, poor user friendliness and lack of adeptness with local contexts. Many barriers exists that prevent the successful development and adoption of well-performing, context-specific AI tools, such as limited data availability, trust and evidence of cost-effectiveness in LMICs. Additional evaluations of the use of AI in healthcare in LMICs are needed in order to identify their effectiveness and reliability in real-world settings and to generate understanding for best practices for future implementations. [ABSTRACT FROM AUTHOR]

Even as innovation occurs within digital medicine, challenges around equity and racial health disparities remain. Golden et al. evaluate structural racism in their recent paper focused on reproductive health. They recommend a framework to Remove, Repair, Restructure, and Remediate. We propose applying the framework to three areas within digital medicine: artificial intelligence (AI) applications, wearable devices, and telehealth. With this approach, we can continue to work towards an equitable future for digital medicine. [ABSTRACT FROM AUTHOR]

Delivery of serious illness communication (SIC) is necessary to ensure that all seriously ill patients receive goal-concordant care. However, the current SIC delivery process contains barriers that prevent the delivery of timely and effective SIC. In this paper, we describe the current bottlenecks of the traditional SIC workflow and explore how a hybrid artificial intelligence-human workflow may improve the efficiency and effectiveness of SIC delivery in busy practice settings. [ABSTRACT FROM AUTHOR]

Digital health data are multimodal and high-dimensional. A patient's health state can be characterized by a multitude of signals including medical imaging, clinical variables, genome sequencing, conversations between clinicians and patients, and continuous signals from wearables, among others. This high volume, personalized data stream aggregated over patients' lives has spurred interest in developing new artificial intelligence (AI) models for higher-precision diagnosis, prognosis, and tracking. While the promise of these algorithms is undeniable, their dissemination and adoption have been slow, owing partially to unpredictable AI model performance once deployed in the real world. We posit that one of the rate-limiting factors in developing algorithms that generalize to real-world scenarios is the very attribute that makes the data exciting—their high-dimensional nature. This paper considers how the large number of features in vast digital health data can challenge the development of robust AI models—a phenomenon known as "the curse of dimensionality" in statistical learning theory. We provide an overview of the curse of dimensionality in the context of digital health, demonstrate how it can negatively impact out-of-sample performance, and highlight important considerations for researchers and algorithm designers. [ABSTRACT FROM AUTHOR]

Autonomous AI systems in medicine promise improved outcomes but raise concerns about liability, regulation, and costs. With the advent of large-language models, which can understand and generate medical text, the urgency for addressing these concerns increases as they create opportunities for more sophisticated autonomous AI systems. This perspective explores the liability implications for physicians, hospitals, and creators of AI technology, as well as the evolving regulatory landscape and payment models. Physicians may be favored in malpractice cases if they follow rigorously validated AI recommendations. However, AI developers may face liability for failing to adhere to industry-standard best practices during development and implementation. The evolving regulatory landscape, led by the FDA, seeks to ensure transparency, evaluation, and real-world monitoring of AI systems, while payment models such as MPFS, NTAP, and commercial payers adapt to accommodate them. The widespread adoption of autonomous AI systems can potentially streamline workflows and allow doctors to concentrate on the human aspects of healthcare. [ABSTRACT FROM AUTHOR]

Large language models such as ChatGPT can produce increasingly realistic text, with unknown information on the accuracy and integrity of using these models in scientific writing. We gathered fifth research abstracts from five high-impact factor medical journals and asked ChatGPT to generate research abstracts based on their titles and journals. Most generated abstracts were detected using an AI output detector, 'GPT-2 Output Detector', with % 'fake' scores (higher meaning more likely to be generated) of median [interquartile range] of 99.98% 'fake' [12.73%, 99.98%] compared with median 0.02% [IQR 0.02%, 0.09%] for the original abstracts. The AUROC of the AI output detector was 0.94. Generated abstracts scored lower than original abstracts when run through a plagiarism detector website and iThenticate (higher scores meaning more matching text found). When given a mixture of original and general abstracts, blinded human reviewers correctly identified 68% of generated abstracts as being generated by ChatGPT, but incorrectly identified 14% of original abstracts as being generated. Reviewers indicated that it was surprisingly difficult to differentiate between the two, though abstracts they suspected were generated were vaguer and more formulaic. ChatGPT writes believable scientific abstracts, though with completely generated data. Depending on publisher-specific guidelines, AI output detectors may serve as an editorial tool to help maintain scientific standards. The boundaries of ethical and acceptable use of large language models to help scientific writing are still being discussed, and different journals and conferences are adopting varying policies. [ABSTRACT FROM AUTHOR]

The digital revolution in healthcare, amplified by the COVID-19 pandemic and artificial intelligence (AI) advances, has led to a surge in the development of digital technologies. However, integrating digital health solutions, especially AI-based ones, in rare diseases like Waldenström macroglobulinemia (WM) remains challenging due to limited data, among other factors. CURATE.AI, a clinical decision support system, offers an alternative to big data approaches by calibrating individual treatment profiles based on that individual's data alone. We present a case study from the PRECISE CURATE.AI trial with a WM patient, where, over two years, CURATE.AI provided dynamic Ibrutinib dose recommendations to clinicians (users) aimed at achieving optimal IgM levels. An 80-year-old male with newly diagnosed WM requiring treatment due to anemia was recruited to the trial for CURATE.AI-based dosing of the Bruton tyrosine kinase inhibitor Ibrutinib. The primary and secondary outcome measures were focused on scientific and logistical feasibility. Preliminary results underscore the platform's potential in enhancing user and patient engagement, in addition to clinical efficacy. Based on a two-year-long patient enrollment into the CURATE.AI-augmented treatment, this study showcases how AI-enabled tools can support the management of rare diseases, emphasizing the integration of AI to enhance personalized therapy. [ABSTRACT FROM AUTHOR]

The importance of a trust-based relationship between patients and medical professionals has been recognized as one of the most important predictors of treatment success and patients' satisfaction. We have developed a novel legal, social and regulatory envelopment of medical AI that is explicitly based on the preservation of trust between patients and medical professionals. We require that the envelopment fosters reliance on the medical AI by both patients and medical professionals. Focusing on this triangle of desirable attitudes allows us to develop eight envelopment components that will support, strengthen and preserve these attitudes. We then demonstrate how each envelopment component can be enacted during different stages of the systems development life cycle and demonstrate that this requires the involvement of medical professionals and patients at the earliest stages of the life cycle. Therefore, this framework requires medical AI start-ups to cooperate with medical professionals and patients throughout. [ABSTRACT FROM AUTHOR]

Digital twins, innovative virtual models synthesizing real-time biological, environmental, and lifestyle data, herald a new era in personalized medicine, particularly dermatology. These models, integrating medical-purpose Internet of Things (IoT) devices, deep and digital phenotyping, and advanced artificial intelligence (AI), offer unprecedented precision in simulating real-world physical conditions and health outcomes. Originating in aerospace and manufacturing for system behavior prediction, their application in healthcare signifies a paradigm shift towards patient-specific care pathways. In dermatology, digital twins promise enhanced diagnostic accuracy, optimized treatment plans, and improved patient monitoring by accommodating the unique complexities of skin conditions. However, a comprehensive review across PubMed, Embase, Web of Science, Cochrane, and Scopus until February 5th, 2024, underscores a significant research gap; no direct studies on digital twins' application in dermatology is identified. This gap signals challenges, including the intricate nature of skin diseases, ethical and privacy concerns, and the necessity for specialized algorithms. Overcoming these barriers through interdisciplinary efforts and focused research is essential for realizing digital twins' potential in dermatology. This study advocates for a proactive exploration of digital twins, emphasizing the need for a tailored approach to dermatological care that is as personalized as the patients themselves. [ABSTRACT FROM AUTHOR]

The European Union's recently adopted Artificial Intelligence (AI) Act is the first comprehensive legal framework specifically on AI. This is particularly important for the healthcare domain, as other existing harmonisation legislation, such as the Medical Device Regulation, do not explicitly cover medical AI applications. Given the far-reaching impact of this regulation on the medical AI sector, this commentary provides an overview of the key elements of the AI Act, with easy-to-follow references to the relevant chapters. [ABSTRACT FROM AUTHOR]

This perspective article explores the challenges and potential of using speech as a biomarker in clinical settings, particularly when constrained by the small clinical datasets typically available in such contexts. We contend that by integrating insights from speech science and clinical research, we can reduce sample complexity in clinical speech AI models with the potential to decrease timelines to translation. Most existing models are based on high-dimensional feature representations trained with limited sample sizes and often do not leverage insights from speech science and clinical research. This approach can lead to overfitting, where the models perform exceptionally well on training data but fail to generalize to new, unseen data. Additionally, without incorporating theoretical knowledge, these models may lack interpretability and robustness, making them challenging to troubleshoot or improve post-deployment. We propose a framework for organizing health conditions based on their impact on speech and promote the use of speech analytics in diverse clinical contexts beyond cross-sectional classification. For high-stakes clinical use cases, we advocate for a focus on explainable and individually-validated measures and stress the importance of rigorous validation frameworks and ethical considerations for responsible deployment. Bridging the gap between AI research and clinical speech research presents new opportunities for more efficient translation of speech-based AI tools and advancement of scientific discoveries in this interdisciplinary space, particularly if limited to small or retrospective datasets. [ABSTRACT FROM AUTHOR]

The adoption of machine learning (ML) and, more specifically, deep learning (DL) applications into all major areas of our lives is underway. The development of trustworthy AI is especially important in medicine due to the large implications for patients' lives. While trustworthiness concerns various aspects including ethical, transparency and safety requirements, we focus on the importance of data quality (training/test) in DL. Since data quality dictates the behaviour of ML products, evaluating data quality will play a key part in the regulatory approval of medical ML products. We perform a systematic review following PRISMA guidelines using the databases Web of Science, PubMed and ACM Digital Library. We identify 5408 studies, out of which 120 records fulfil our eligibility criteria. From this literature, we synthesise the existing knowledge on data quality frameworks and combine it with the perspective of ML applications in medicine. As a result, we propose the METRIC-framework, a specialised data quality framework for medical training data comprising 15 awareness dimensions, along which developers of medical ML applications should investigate the content of a dataset. This knowledge helps to reduce biases as a major source of unfairness, increase robustness, facilitate interpretability and thus lays the foundation for trustworthy AI in medicine. The METRIC-framework may serve as a base for systematically assessing training datasets, establishing reference datasets, and designing test datasets which has the potential to accelerate the approval of medical ML products. [ABSTRACT FROM AUTHOR]

We studied clinical AI-supported decision-making as an example of a high-stakes setting in which explainable AI (XAI) has been proposed as useful (by theoretically providing physicians with context for the AI suggestion and thereby helping them to reject unsafe AI recommendations). Here, we used objective neurobehavioural measures (eye-tracking) to see how physicians respond to XAI with N = 19 ICU physicians in a hospital's clinical simulation suite. Prescription decisions were made both pre- and post-reveal of either a safe or unsafe AI recommendation and four different types of simultaneously presented XAI. We used overt visual attention as a marker for where physician mental attention was directed during the simulations. Unsafe AI recommendations attracted significantly greater attention than safe AI recommendations. However, there was no appreciably higher level of attention placed onto any of the four types of explanation during unsafe AI scenarios (i.e. XAI did not appear to 'rescue' decision-makers). Furthermore, self-reported usefulness of explanations by physicians did not correlate with the level of attention they devoted to the explanations reinforcing the notion that using self-reports alone to evaluate XAI tools misses key aspects of the interaction behaviour between human and machine. [ABSTRACT FROM AUTHOR]

The 12-lead electrocardiogram (ECG) is an integral component to the diagnosis of a multitude of cardiovascular conditions. It is performed using a complex set of skin surface electrodes, limiting its use outside traditional clinical settings. We developed an artificial intelligence algorithm, trained over 600,000 clinically acquired ECGs, to explore whether fewer leads as input are sufficient to reconstruct a 12-lead ECG. Two limb leads (I and II) and one precordial lead (V3) were required to generate a reconstructed 12-lead ECG highly correlated with the original ECG. An automatic algorithm for detection of ECG features consistent with acute myocardial infarction (MI) performed similarly for original and reconstructed ECGs (AUC = 0.95). When interpreted by cardiologists, reconstructed ECGs achieved an accuracy of 81.4 ± 5.0% in identifying ECG features of ST-segment elevation MI, comparable with the original 12-lead ECGs (accuracy 84.6 ± 4.6%). These results will impact development efforts to innovate ECG acquisition methods with simplified tools in non-specialized settings. [ABSTRACT FROM AUTHOR]

Given the current state of medical artificial intelligence (AI) and perceptions towards it, collaborative systems are becoming the preferred choice for clinical workflows. This work aims to address expert interaction with medical AI support systems to gain insight towards how these systems can be better designed with the user in mind. As eye tracking metrics have been shown to be robust indicators of usability, we employ them for evaluating the usability and user interaction with medical AI support systems. We use expert gaze to assess experts' interaction with an AI software for caries detection in bitewing x-ray images. We compared standard viewing of bitewing images without AI support versus viewing where AI support could be freely toggled on and off. We found that experts turned the AI on for roughly 25% of the total inspection task, and generally turned it on halfway through the course of the inspection. Gaze behavior showed that when supported by AI, more attention was dedicated to user interface elements related to the AI support, with more frequent transitions from the image itself to these elements. When considering that expert visual strategy is already optimized for fast and effective image inspection, such interruptions in attention can lead to increased time needed for the overall assessment. Gaze analysis provided valuable insights into an AI's usability for medical image inspection. Further analyses of these tools and how to delineate metrical measures of usability should be developed. [ABSTRACT FROM AUTHOR]

Increasing evidence supports reduced accuracy of noninvasive assessment tools, such as pulse oximetry, temperature probes, and AI skin diagnosis benchmarks, in patients with darker skin tones. The FDA is exploring potential strategies for device regulation to improve performance across diverse skin tones by including skin tone criteria. However, there is no consensus about how prospective studies should perform skin tone assessment in order to take this bias into account. There are several tools available to conduct skin tone assessments including administered visual scales (e.g., Fitzpatrick Skin Type, Pantone, Monk Skin Tone) and color measurement tools (e.g., reflectance colorimeters, reflectance spectrophotometers, cameras), although none are consistently used or validated across multiple medical domains. Accurate and consistent skin tone measurement depends on many factors including standardized environments, lighting, body parts assessed, patient conditions, and choice of skin tone assessment tool(s). As race and ethnicity are inadequate proxies for skin tone, these considerations can be helpful in standardizing the effect of skin tone on studies such as AI dermatology diagnoses, pulse oximetry, and temporal thermometers. Skin tone bias in medical devices is likely due to systemic factors that lead to inadequate validation across diverse skin tones. There is an opportunity for researchers to use skin tone assessment methods with standardized considerations in prospective studies of noninvasive tools that may be affected by skin tone. We propose considerations that researchers must take in order to improve device robustness to skin tone bias. [ABSTRACT FROM AUTHOR]

With the introduction of ChatGPT, Large Language Models (LLMs) have received enormous attention in healthcare. Despite potential benefits, researchers have underscored various ethical implications. While individual instances have garnered attention, a systematic and comprehensive overview of practical applications currently researched and ethical issues connected to them is lacking. Against this background, this work maps the ethical landscape surrounding the current deployment of LLMs in medicine and healthcare through a systematic review. Electronic databases and preprint servers were queried using a comprehensive search strategy which generated 796 records. Studies were screened and extracted following a modified rapid review approach. Methodological quality was assessed using a hybrid approach. For 53 records, a meta-aggregative synthesis was performed. Four general fields of applications emerged showcasing a dynamic exploration phase. Advantages of using LLMs are attributed to their capacity in data analysis, information provisioning, support in decision-making or mitigating information loss and enhancing information accessibility. However, our study also identifies recurrent ethical concerns connected to fairness, bias, non-maleficence, transparency, and privacy. A distinctive concern is the tendency to produce harmful or convincing but inaccurate content. Calls for ethical guidance and human oversight are recurrent. We suggest that the ethical guidance debate should be reframed to focus on defining what constitutes acceptable human oversight across the spectrum of applications. This involves considering the diversity of settings, varying potentials for harm, and different acceptable thresholds for performance and certainty in healthcare. Additionally, critical inquiry is needed to evaluate the necessity and justification of LLMs' current experimental use. [ABSTRACT FROM AUTHOR]

Automatic assessment of impairment and disease severity is a key challenge in data-driven medicine. We propose a framework to address this challenge, which leverages AI models trained exclusively on healthy individuals. The COnfidence-Based chaRacterization of Anomalies (COBRA) score exploits the decrease in confidence of these models when presented with impaired or diseased patients to quantify their deviation from the healthy population. We applied the COBRA score to address a key limitation of current clinical evaluation of upper-body impairment in stroke patients. The gold-standard Fugl-Meyer Assessment (FMA) requires in-person administration by a trained assessor for 30-45 minutes, which restricts monitoring frequency and precludes physicians from adapting rehabilitation protocols to the progress of each patient. The COBRA score, computed automatically in under one minute, is shown to be strongly correlated with the FMA on an independent test cohort for two different data modalities: wearable sensors (ρ = 0.814, 95% CI [0.700,0.888]) and video (ρ = 0.736, 95% C.I [0.584, 0.838]). To demonstrate the generalizability of the approach to other conditions, the COBRA score was also applied to quantify severity of knee osteoarthritis from magnetic-resonance imaging scans, again achieving significant correlation with an independent clinical assessment (ρ = 0.644, 95% C.I [0.585,0.696]). [ABSTRACT FROM AUTHOR]

AI-enabled ECGs have previously been shown to accurately predict patient sex in adults and correlate with sex hormone levels. We aimed to test the ability of AI-enabled ECGs to predict sex in the pediatric population and study the influence of pubertal development. AI-enabled ECG models were created using a convolutional neural network trained on pediatric 10-second, 12-lead ECGs. The first model was trained de novo using pediatric data. The second model used transfer learning from a previously validated adult data-derived algorithm. We analyzed the first ECG from 90,133 unique pediatric patients (aged ≤18 years) recorded between 1987–2022, and divided the cohort into training, validation, and testing datasets. Subgroup analysis was performed on prepubertal (0–7 years), peripubertal (8–14 years), and postpubertal (15–18 years) patients. The cohort was 46.7% male, with 21,678 prepubertal, 26,740 peripubertal, and 41,715 postpubertal children. The de novo pediatric model demonstrated 81% accuracy and an area under the curve (AUC) of 0.91. Model sensitivity was 0.79, specificity was 0.83, positive predicted value was 0.84, and the negative predicted value was 0.78, for the entire test cohort. The model's discriminatory ability was highest in postpubertal (AUC = 0.98), lower in the peripubertal age group (AUC = 0.91), and poor in the prepubertal age group (AUC = 0.67). There was no significant performance difference observed between the transfer learning and de novo models. AI-enabled interpretation of ECG can estimate sex in peripubertal and postpubertal children with high accuracy. [ABSTRACT FROM AUTHOR]

Computational drug repurposing methods adapt Artificial intelligence (AI) algorithms for the discovery of new applications of approved or investigational drugs. Among the heterogeneous datasets, electronic health records (EHRs) datasets provide rich longitudinal and pathophysiological data that facilitate the generation and validation of drug repurposing. Here, we present an appraisal of recently published research on computational drug repurposing utilizing the EHR. Thirty-three research articles, retrieved from Embase, Medline, Scopus, and Web of Science between January 2000 and January 2022, were included in the final review. Four themes, (1) publication venue, (2) data types and sources, (3) method for data processing and prediction, and (4) targeted disease, validation, and released tools were presented. The review summarized the contribution of EHR used in drug repurposing as well as revealed that the utilization is hindered by the validation, accessibility, and understanding of EHRs. These findings can support researchers in the utilization of medical data resources and the development of computational methods for drug repurposing. [ABSTRACT FROM AUTHOR]

The electrocardiogram (ECG) can capture obesity-related cardiac changes. Artificial intelligence-enhanced ECG (AI-ECG) can identify subclinical disease. We trained an AI-ECG model to predict body mass index (BMI) from the ECG alone. Developed from 512,950 12-lead ECGs from the Beth Israel Deaconess Medical Center (BIDMC), a secondary care cohort, and validated on UK Biobank (UKB) (n = 42,386), the model achieved a Pearson correlation coefficient (r) of 0.65 and 0.62, and an R2 of 0.43 and 0.39 in the BIDMC cohort and UK Biobank, respectively for AI-ECG BMI vs. measured BMI. We found delta-BMI, the difference between measured BMI and AI-ECG-predicted BMI (AI-ECG-BMI), to be a biomarker of cardiometabolic health. The top tertile of delta-BMI showed increased risk of future cardiometabolic disease (BIDMC: HR 1.15, p < 0.001; UKB: HR 1.58, p < 0.001) and diabetes mellitus (BIDMC: HR 1.25, p < 0.001; UKB: HR 2.28, p < 0.001) after adjusting for covariates including measured BMI. Significant enhancements in model fit, reclassification and improvements in discriminatory power were observed with the inclusion of delta-BMI in both cohorts. Phenotypic profiling highlighted associations between delta-BMI and cardiometabolic diseases, anthropometric measures of truncal obesity, and pericardial fat mass. Metabolic and proteomic profiling associates delta-BMI positively with valine, lipids in small HDL, syntaxin-3, and carnosine dipeptidase 1, and inversely with glutamine, glycine, colipase, and adiponectin. A genome-wide association study revealed associations with regulators of cardiovascular/metabolic traits, including SCN10A, SCN5A, EXOG and RXRG. In summary, our AI-ECG-BMI model accurately predicts BMI and introduces delta-BMI as a non-invasive biomarker for cardiometabolic risk stratification. [ABSTRACT FROM AUTHOR]

The coronary angiogram is the gold standard for evaluating the severity of coronary artery disease stenoses. Presently, the assessment is conducted visually by cardiologists, a method that lacks standardization. This study introduces DeepCoro, a ground-breaking AI-driven pipeline that integrates advanced vessel tracking and a video-based Swin3D model that was trained and validated on a dataset comprised of 182,418 coronary angiography videos spanning 5 years. DeepCoro achieved a notable precision of 71.89% in identifying coronary artery segments and demonstrated a mean absolute error of 20.15% (95% CI: 19.88–20.40) and a classification AUROC of 0.8294 (95% CI: 0.8215–0.8373) in stenosis percentage prediction compared to traditional cardiologist assessments. When compared to two expert interventional cardiologists, DeepCoro achieved lower variability than the clinical reports (19.09%; 95% CI: 18.55–19.58 vs 21.00%; 95% CI: 20.20–21.76, respectively). In addition, DeepCoro can be fine-tuned to a different modality type. When fine-tuned on quantitative coronary angiography assessments, DeepCoro attained an even lower mean absolute error of 7.75% (95% CI: 7.37–8.07), underscoring the reduced variability inherent to this method. This study establishes DeepCoro as an innovative video-based, adaptable tool in coronary artery disease analysis, significantly enhancing the precision and reliability of stenosis assessment. [ABSTRACT FROM AUTHOR]

Generalization – the ability of AI systems to apply and/or extrapolate their knowledge to new data which might differ from the original training data – is a major challenge for the effective and responsible implementation of human-centric AI applications. Current debate in bioethics proposes selective prediction as a solution. Here we explore data-based reasons for generalization challenges and look at how selective predictions might be implemented technically, focusing on clinical AI applications in real-world healthcare settings. [ABSTRACT FROM AUTHOR]

Determining acute ischemic stroke (AIS) etiology is fundamental to secondary stroke prevention efforts but can be diagnostically challenging. We trained and validated an automated classification tool, StrokeClassifier, using electronic health record (EHR) text from 2039 non-cryptogenic AIS patients at 2 academic hospitals to predict the 4-level outcome of stroke etiology adjudicated by agreement of at least 2 board-certified vascular neurologists' review of the EHR. StrokeClassifier is an ensemble consensus meta-model of 9 machine learning classifiers applied to features extracted from discharge summary texts by natural language processing. StrokeClassifier was externally validated in 406 discharge summaries from the MIMIC-III dataset reviewed by a vascular neurologist to ascertain stroke etiology. Compared with vascular neurologists' diagnoses, StrokeClassifier achieved the mean cross-validated accuracy of 0.74 and weighted F1 of 0.74 for multi-class classification. In MIMIC-III, its accuracy and weighted F1 were 0.70 and 0.71, respectively. In binary classification, the two metrics ranged from 0.77 to 0.96. The top 5 features contributing to stroke etiology prediction were atrial fibrillation, age, middle cerebral artery occlusion, internal carotid artery occlusion, and frontal stroke location. We designed a certainty heuristic to grade the confidence of StrokeClassifier's diagnosis as non-cryptogenic by the degree of consensus among the 9 classifiers and applied it to 788 cryptogenic patients, reducing cryptogenic diagnoses from 25.2% to 7.2%. StrokeClassifier is a validated artificial intelligence tool that rivals the performance of vascular neurologists in classifying ischemic stroke etiology. With further training, StrokeClassifier may have downstream applications including its use as a clinical decision support system. [ABSTRACT FROM AUTHOR]

Healthcare datasets are becoming larger and more complex, necessitating the development of accurate and generalizable AI models for medical applications. Unstructured datasets, including medical imaging, electrocardiograms, and natural language data, are gaining attention with advancements in deep convolutional neural networks and large language models. However, estimating the generalizability of these models to new healthcare settings without extensive validation on external data remains challenging. In experiments across 13 datasets including X-rays, CTs, ECGs, clinical discharge summaries, and lung auscultation data, our results demonstrate that model performance is frequently overestimated by up to 20% on average due to shortcut learning of hidden data acquisition biases (DAB). Shortcut learning refers to a phenomenon in which an AI model learns to solve a task based on spurious correlations present in the data as opposed to features directly related to the task itself. We propose an open source, bias-corrected external accuracy estimate, PEst, that better estimates external accuracy to within 4% on average by measuring and calibrating for DAB-induced shortcut learning. [ABSTRACT FROM AUTHOR]

Scientific research of artificial intelligence (AI) in dermatology has increased exponentially. The objective of this study was to perform a systematic review and meta-analysis to evaluate the performance of AI algorithms for skin cancer classification in comparison to clinicians with different levels of expertise. Based on PRISMA guidelines, 3 electronic databases (PubMed, Embase, and Cochrane Library) were screened for relevant articles up to August 2022. The quality of the studies was assessed using QUADAS-2. A meta-analysis of sensitivity and specificity was performed for the accuracy of AI and clinicians. Fifty-three studies were included in the systematic review, and 19 met the inclusion criteria for the meta-analysis. Considering all studies and all subgroups of clinicians, we found a sensitivity (Sn) and specificity (Sp) of 87.0% and 77.1% for AI algorithms, respectively, and a Sn of 79.78% and Sp of 73.6% for all clinicians (overall); differences were statistically significant for both Sn and Sp. The difference between AI performance (Sn 92.5%, Sp 66.5%) vs. generalists (Sn 64.6%, Sp 72.8%), was greater, when compared with expert clinicians. Performance between AI algorithms (Sn 86.3%, Sp 78.4%) vs expert dermatologists (Sn 84.2%, Sp 74.4%) was clinically comparable. Limitations of AI algorithms in clinical practice should be considered, and future studies should focus on real-world settings, and towards AI-assistance. [ABSTRACT FROM AUTHOR]

Distribution shifts remain a problem for the safe application of regulated medical AI systems, and may impact their real-world performance if undetected. Postmarket shifts can occur for example if algorithms developed on data from various acquisition settings and a heterogeneous population are predominantly applied in hospitals with lower quality data acquisition or other centre-specific acquisition factors, or where some ethnicities are over-represented. Therefore, distribution shift detection could be important for monitoring AI-based medical products during postmarket surveillance. We implemented and evaluated three deep-learning based shift detection techniques (classifier-based, deep kernel, and multiple univariate kolmogorov-smirnov tests) on simulated shifts in a dataset of 130'486 retinal images. We trained a deep learning classifier for diabetic retinopathy grading. We then simulated population shifts by changing the prevalence of patients' sex, ethnicity, and co-morbidities, and example acquisition shifts by changes in image quality. We observed classification subgroup performance disparities w.r.t. image quality, patient sex, ethnicity and co-morbidity presence. The sensitivity at detecting referable diabetic retinopathy ranged from 0.50 to 0.79 for different ethnicities. This motivates the need for detecting shifts after deployment. Classifier-based tests performed best overall, with perfect detection rates for quality and co-morbidity subgroup shifts at a sample size of 1000. It was the only method to detect shifts in patient sex, but required large sample sizes ( > 3 0 ′ 000 ). All methods identified easier-to-detect out-of-distribution shifts with small (≤300) sample sizes. We conclude that effective tools exist for detecting clinically relevant distribution shifts. In particular classifier-based tests can be easily implemented components in the post-market surveillance strategy of medical device manufacturers. [ABSTRACT FROM AUTHOR]

AI holds the potential to transform healthcare, promising improvements in patient care. Yet, realizing this potential is hampered by over-reliance on limited datasets and a lack of transparency in validation processes. To overcome these obstacles, we advocate the creation of a detailed registry for AI algorithms. This registry would document the development, training, and validation of AI models, ensuring scientific integrity and transparency. Additionally, it would serve as a platform for peer review and ethical oversight. By bridging the gap between scientific validation and regulatory approval, such as by the FDA, we aim to enhance the integrity and trustworthiness of AI applications in healthcare. [ABSTRACT FROM AUTHOR]

Ensuring diagnostic performance of artificial intelligence (AI) before introduction into clinical practice is essential. Growing numbers of studies using AI for digital pathology have been reported over recent years. The aim of this work is to examine the diagnostic accuracy of AI in digital pathology images for any disease. This systematic review and meta-analysis included diagnostic accuracy studies using any type of AI applied to whole slide images (WSIs) for any disease. The reference standard was diagnosis by histopathological assessment and/or immunohistochemistry. Searches were conducted in PubMed, EMBASE and CENTRAL in June 2022. Risk of bias and concerns of applicability were assessed using the QUADAS-2 tool. Data extraction was conducted by two investigators and meta-analysis was performed using a bivariate random effects model, with additional subgroup analyses also performed. Of 2976 identified studies, 100 were included in the review and 48 in the meta-analysis. Studies were from a range of countries, including over 152,000 whole slide images (WSIs), representing many diseases. These studies reported a mean sensitivity of 96.3% (CI 94.1–97.7) and mean specificity of 93.3% (CI 90.5–95.4). There was heterogeneity in study design and 99% of studies identified for inclusion had at least one area at high or unclear risk of bias or applicability concerns. Details on selection of cases, division of model development and validation data and raw performance data were frequently ambiguous or missing. AI is reported as having high diagnostic accuracy in the reported areas but requires more rigorous evaluation of its performance. [ABSTRACT FROM AUTHOR]

Digital health innovation is expected to transform healthcare, but it also generates ethical and societal concerns, such as privacy risks, and biases that can compound existing health inequalities. While such concerns are widely recognized, existing regulatory principles, oversight methods and ethical frameworks seem out of sync with digital health innovation. New governance and innovation best practices are thus needed to bring such principles to bear with the reality of business, innovation, and regulation. To grant practical insight into best practices for responsible digital health innovation, we conducted a qualitative study based on an interactive engagement methodology. We engaged key stakeholders (n = 46) operating at the translational frontier of digital health. This approach allowed us to identify three clusters of governance and innovation best practices in digital health innovation: i) inclusive co-creation, ii) responsive regulation, and iii) value-driven innovation. Our study shows that realizing responsible digital health requires diverse stakeholders' commitment to adapt innovation and regulation practices, embracing co-creation as the default modus operandi for digital health development. We describe these collaborative practices and show how they can ensure that innovation is neither slowed by overregulation, nor leads to unethical outcomes. [ABSTRACT FROM AUTHOR]

Existing natural language processing (NLP) methods to convert free-text clinical notes into structured data often require problem-specific annotations and model training. This study aims to evaluate ChatGPT's capacity to extract information from free-text medical notes efficiently and comprehensively. We developed a large language model (LLM)-based workflow, utilizing systems engineering methodology and spiral "prompt engineering" process, leveraging OpenAI's API for batch querying ChatGPT. We evaluated the effectiveness of this method using a dataset of more than 1000 lung cancer pathology reports and a dataset of 191 pediatric osteosarcoma pathology reports, comparing the ChatGPT-3.5 (gpt-3.5-turbo-16k) outputs with expert-curated structured data. ChatGPT-3.5 demonstrated the ability to extract pathological classifications with an overall accuracy of 89%, in lung cancer dataset, outperforming the performance of two traditional NLP methods. The performance is influenced by the design of the instructive prompt. Our case analysis shows that most misclassifications were due to the lack of highly specialized pathology terminology, and erroneous interpretation of TNM staging rules. Reproducibility shows the relatively stable performance of ChatGPT-3.5 over time. In pediatric osteosarcoma dataset, ChatGPT-3.5 accurately classified both grades and margin status with accuracy of 98.6% and 100% respectively. Our study shows the feasibility of using ChatGPT to process large volumes of clinical notes for structured information extraction without requiring extensive task-specific human annotation and model training. The results underscore the potential role of LLMs in transforming unstructured healthcare data into structured formats, thereby supporting research and aiding clinical decision-making. [ABSTRACT FROM AUTHOR]

Digital health technologies (DHTs) can transform neurological assessments, improving quality and continuity of care. In the United States, the Food & Drug Administration (FDA) oversees the safety and efficacy of these technologies, employing a detailed regulatory process that classifies devices based on risk and requires rigorous review and post-market surveillance. Following FDA approval, DHTs enter the Current Procedural Terminology, Relative Value Scale Update Committee, and Centers for Medicare & Medicaid Services coding and valuation processes leading to coverage and payment decisions. DHT adoption is challenged by rapid technologic advancements, an inconsistent evidence base, marketing discrepancies, ambiguous coding guidance, and variable health insurance coverage. Regulators, policymakers, and payers will need to develop better methods to evaluate these promising technologies and guide their deployment. This includes striking a balance between patient safety and clinical effectiveness versus promotion of innovation, especially as DHTs increasingly incorporate artificial intelligence. Data validity, cybersecurity, risk management, societal, and ethical responsibilities should be addressed. Regulatory advances can support adoption of these promising tools by ensuring DHTs are safe, effective, accessible, and equitable. [ABSTRACT FROM AUTHOR]

An increasing number of artificial intelligence (AI) tools are moving towards the clinical realm in histopathology and across medicine. The introduction of such tools will bring several benefits to diagnostic specialities, namely increased diagnostic accuracy and efficiency, however, as no AI tool is infallible, their use will inevitably introduce novel errors. These errors made by AI tools are, most fundamentally, misclassifications made by a computational algorithm. Understanding of how these translate into clinical impact on patients is often lacking, meaning true reporting of AI tool safety is incomplete. In this Perspective we consider AI diagnostic tools in histopathology, which are predominantly assessed in terms of technical performance metrics such as sensitivity, specificity and area under the receiver operating characteristic curve. Although these metrics are essential and allow tool comparison, they alone give an incomplete picture of how an AI tool's errors could impact a patient's diagnosis, management and prognosis. We instead suggest assessing and reporting AI tool errors from a pathological and clinical stance, demonstrating how this is done in studies on human pathologist errors, and giving examples where available from pathology and radiology. Although this seems a significant task, we discuss ways to move towards this approach in terms of study design, guidelines and regulation. This Perspective seeks to initiate broader consideration of the assessment of AI tool errors in histopathology and across diagnostic specialities, in an attempt to keep patient safety at the forefront of AI tool development and facilitate safe clinical deployment. [ABSTRACT FROM AUTHOR]

Cardiac digital twins provide a physics and physiology informed framework to deliver personalized medicine. However, high-fidelity multi-scale cardiac models remain a barrier to adoption due to their extensive computational costs. Artificial Intelligence-based methods can make the creation of fast and accurate whole-heart digital twins feasible. We use Latent Neural Ordinary Differential Equations (LNODEs) to learn the pressure-volume dynamics of a heart failure patient. Our surrogate model is trained from 400 simulations while accounting for 43 parameters describing cell-to-organ cardiac electromechanics and cardiovascular hemodynamics. LNODEs provide a compact representation of the 3D-0D model in a latent space by means of an Artificial Neural Network that retains only 3 hidden layers with 13 neurons per layer and allows for numerical simulations of cardiac function on a single processor. We employ LNODEs to perform global sensitivity analysis and parameter estimation with uncertainty quantification in 3 hours of computations, still on a single processor. [ABSTRACT FROM AUTHOR]

When integrating AI tools in healthcare settings, complex interactions between technologies and primary users are not always fully understood or visible. This deficient and ambiguous understanding hampers attempts by healthcare organizations to adopt AI/ML, and it also creates new challenges for researchers to identify opportunities for simplifying adoption and developing best practices for the use of AI-based solutions. Our study fills this gap by documenting the process of designing, building, and maintaining an AI solution called SepsisWatch at Duke University Health System. We conducted 20 interviews with the team of engineers and scientists that led the multi-year effort to build the tool, integrate it into practice, and maintain the solution. This "Algorithm Journey Map" enumerates all social and technical activities throughout the AI solution's procurement, development, integration, and full lifecycle management. In addition to mapping the "who?" and "what?" of the adoption of the AI tool, we also show several 'lessons learned' throughout the algorithm journey maps including modeling assumptions, stakeholder inclusion, and organizational structure. In doing so, we identify generalizable insights about how to recognize and navigate barriers to AI/ML adoption in healthcare settings. We expect that this effort will further the development of best practices for operationalizing and sustaining ethical principles—in algorithmic systems. [ABSTRACT FROM AUTHOR]