Speakers and Panel participants, AI in Healthcare

Revolutionizing Healthcare: AI-Driven Breakthroughs in Medicine and Healthcare Delivery
This presentation explores the transformative impacts of machine learning and AI on personalized medicine, clinical trials and healthcare delivery.

I will introduce several breakthrough machine learning methods developed in our lab aimed to address some of the hardest and most complex challenges in medicine and healthcare. By integrating these methods in clinical practice together with clinicians, we are not only enhancing the precision of medical care for each patient, but also revolutionizing the efficiency of healthcare systems and of clinical trials.

The talk will showcase real-world examples of how machine learning and AI can redefine the entire healthcare landscape, providing insights into the current applications, challenges, and future directions of AI in medicine. Come discuss how AI can make these transformations in healthcare and join forces with us in this endeavour!

Artificial Intelligence: A Catalyst for Quality

AI – A Beacon of Hope for Patients’ Well-being and Sustainable Healthcare?

Bio
Member of the Board of Management at Rigshospitalet and Centre Director, Centre of Head and Orthopaedics.

Martin’s areas of responsibility as a Board member include innovation and external partnerships.

Read more

Complex Clinical Cases diagnosed by GPT-4
Public interest in large language models boomed in 2023 as a result of the public release of ChatGPT. It was a global phenomenon and potential uses were immediately explored in all areas of society. This included diagnostic medicine. In our study, we assessed the performance of the newly released AI GPT-4 in diagnosing complex medical case challenges and compared the success rate to that of medical-journal readers.

GPT-4 correctly diagnosed 57% of cases, outperforming 99.98% of simulated human readers generated from online answers. We highlight the potential for AI to be a powerful supportive tool for diagnosis; however, further improvements, validation, and addressing of ethical considerations are needed before clinical implementation. Since our study, other studies have served to further highlight the potential of large language models in diagnostics.

Ambolt AI
Ambolt AI leverages cutting-edge AI technologies to revolutionize healthcare. They specialize in developing AI-driven solutions for the healthcare sector, utilizing their knowledge AI - Machine Learning - Decision support - Intelligent control systems - Computer Vision - Deep learning - NLP (Natural Language Processing). With a focus on innovation and precision, Ambolt AI empowers healthcare providers with advanced tools for better decision-making and patient care. Ambolt, in collaboration with Norwegian doctors Marius Christensen and Anders Stormo, developed Helseboka, a pioneering e-health solution. Launched in 2019, Helseboka provides users with easy access to personal health data, appointment scheduling, and secure communication channels. With over 18,000 healthcare professionals and 3+ million individuals utilizing its services, Helseboka stands as Norway's foremost e-health platform. Notably, amidst the Covid-19 pandemic, it facilitated crucial testing and vaccination appointments, with 4.4 million vaccinations recorded as of 2022, underscoring Ambolt's pivotal role in advancing public health initiatives.

AI in the primary sector, use-cases for large language models
The deployment of artificial intelligence (AI), specifically through large language models (LLMs), in healthcare and general practice represents a significant evolution in healthcare delivery.

This paper outlines multiple use-cases of LLMs in these settings, emphasizing their role in enhancing diagnostic accuracy, patient engagement, personalized care, and notably, how these tools can significantly reduce time spent on administrative tasks. LLMs are employed to analyze patient history, symptoms, and clinical data to assist in early diagnosis and treatment planning.
They also enhance patient-doctor communication by providing healthcare professionals with tools to better understand patient concerns and potentially enhance empathy. Furthermore, LLMs personalize patient education materials and treatment plans by integrating medical history with current best practices.

Additionally, LLMs help manage and synthesize the vast amounts of research and clinical guidelines that inform practice, helping clinicians stay current with advancements in medical science. The paper argues that incorporating LLMs into family medicine can lead to more informed clinical decisions, improved patient outcomes, and more efficient healthcare delivery, while also addressing challenges such as data privacy and the need for personalized patient interactions.

AI in cognitive assessments - The future of screening tools
The diagnostic pathway for a dementia diagnosis is rapidly growing with the elder burden. Currently, some places in Denmark experience waiting times of upwards to 1 year from first GP visit to final diagnosis. With new drugs entering the market, which can postpone cognitive decline, it is paramount to get an early diagnosis.

The talk will be presenting our study, which implemented a machine learning model to classify between healthy control and Alzheimer's Disease patients based on 1 minute speech samples. Furthermore, it will explain where this technology could take cognitive screenings in terms of precision, timing, scalability, availability and cost.

Demens AI
DemensAI is a pioneering startup aiming to revolutionize dementia diagnosis through cutting-edge artificial intelligence. Founded by Anton Boldrup Birn, Laurine Dargaud, and Abhista Partal, DemensAI's innovative AI technology utilizes speech recognition and language analysis to assess cognitive function swiftly and accurately. By providing a non-invasive, digital solution, DemensAI aims to facilitate early detection of Alzheimer's and other forms of dementia, ultimately paving the way for timely intervention and improved patient outcomes. Their vision is to empower healthcare professionals with a tool that can efficiently screen individuals for dementia, leading to earlier diagnoses and potentially enhancing the efficacy of emerging treatments like Lecanemab and Donanemab.

Integrating CLL-TIM into EHR System and Clinical Trial
Research algorithms often face challenges in validation and integration into clinical practice. Addressing issues like data harmonization, missing data handling, automation, and legal compliance is crucial.

We present the deployment of the Chronic Lymphocytic Leukemia (CLL) Treatment Infection Model (CLL-TIM), a data-driven decision support model, into an an EPIC-based Danish Electronic Health Record (EHR)system, offering personalized predictions for (CLL) patients.
Despite unexpected missing data challenges, our approach emphasizes handling high dimensionality and predictive confidence for reliable outcomes. Our deployment process, including automation and monitoring, adheres to Medical Device Regulation, offering a roadmap for deploying research algorithms in clinical settings

Bringing AI to general practice: Challenges and opportunities
The health care sector is under pressure due to shortages of clinicians and changing demography. AI solutions have the potential to improve clinical practice, efficiency and improve equality for patients. But a variety of challenges have hindered the deployment of such solutions, notably at the legal, technical, and political levels. Fortunately, recent and ongoing developments have led to opportunities to address some of these challenges.

Career History
Charles Vesteghem has more than 17 years of experience in innovation, with 10 years in startups, including 6 years as the founder and CEO of a data driven and digitally focused company. In the last 7 years, Charles has been conducting and leading research within clinical data science with a focus on implementation of AI in the health care sector. He is currently group leader in digital health at the Center of Clinical Data Science and is driving an AI infrastructure project at the Center for General Practice at Aalborg University.

WARD - When Public-Private Innovation Succeeds
WARD is a public-private project that has been developing a software solution since 2016. This solution uses artificial intelligence (AI) and wireless sensors to monitor patients’ conditions in hospitals and at home.

WARD reduces complications and frees up personnel resources. Along the way, the project has established a spin-out company called WARD24/7, which pursues commercial opportunities and has CE certification for the solution.

WARD’s founders, Prof. Christian S. Meyhoff and Prof. Eske K. Aasvang, will share their experiences on the journey from public research to a commercial product, including documentation of the solution’s effectiveness and the need for process optimization related to development, ownership, scaling, and potential.

Thoughts of a Global AI Deployment Lead
In this talk, I will share some hands-on experience from my work as AI Deployment Lead working within a Global AI product team at one of the world’s biggest healthcare companies, including a sneak peek into how AI-enabled eye-disease diagnosis and treatment may look in a few years.

As a patient, imagine doing your own eye disease screening from home, and as a physician, imagine getting in your office the new digital assistant, ROSA, that enables you to see more patients while giving each patient a better experience and a more precise diagnosis.

Can large language models reason about medical questions?
Foundation models have changed the way machine learning is practiced. Foundation models applied to text, so-called large language models (LLMs), have proven to be a disruptive technology. They might radically change the way we interact with computers.
In early 2022, it was clear that generalist LLMs can outperform domain-specific approaches in many domains. Benchmarks that reflect real-world scenarios were still needed, and today, it remains unclear how to best use and evaluate these models. Our results support that future LLMs might be applicable to critical real-world applications such as supporting healthcare professionals.
Link til artikel 

AI and Headache
AI and machine-learning (ML) is moving into the headache research field at full speed. This presentation will give an overview on what has been done so far, how ML is currently used with different data sources at the NorHead research centre, and how we envision it will be of value to researchers, clinicians and patients.

WARD - When Public-Private Innovation Succeeds
WARD is a public-private project that has been developing a software solution since 2016. This solution uses artificial intelligence (AI) and wireless sensors to monitor patients’ conditions in hospitals and at home.

WARD reduces complications and frees up personnel resources. Along the way, the project has established a spin-out company called WARD24/7, which pursues commercial opportunities and has CE certification for the solution.

WARD’s founders, Prof. Christian S. Meyhoff and Prof. Eske K. Aasvang, will share their experiences on the journey from public research to a commercial product, including documentation of the solution’s effectiveness and the need for process optimization related to development, ownership, scaling, and potential.

Machine learning and physician prescribing: a path to reduced antibiotic use
Artificial Intelligence has the potential to improve human decisions in complex environments but its effectiveness can remain limited if humans hold context-specific private information. Using the empirical example of antibiotic prescribing for urinary tract infections, we show that automated prescribing would fail to improve on physician decisions. Instead, optimally delegating a share of decisions to physicians, where they possess private diagnostic information, effectively utilizes the complementarity between algorithmic and human decisions.

Combining physician and algorithmic decisions can achieve a reduction in inefficient overprescribing of antibiotics by 20.3 percent. We explore potential drivers of these improvements, including diagnostic information and physicians’ objectives when prescribing antibiotics.

Career History
Hannes Ullrich is an applied microeconomist with research interests in empirical industrial organization, health economics, and digitization. He is an Associate Professor of Economics at the University of Copenhagen and Deputy Head of the Department Firms and Markets at the DIW Berlin. His work has been published in outlets such as the Economic Journal, Journal of Health Economics, Journal of Human Resources, and Management Science. He holds a Starting Grant from the European Research Council for research on antibiotic prescribing and resistance.

A randomized clinical trial of a decision support tool: The Ischemic Heart Disease Algorithm PM Heart^IHD

Using AI based decision support for improving outcomes in cancer surgery
Despite improvements in the surgical and perioperative care of patients undergoing cancer surgery, there are considerable challenges in in both short and long term outcomes.

In the recent years it has been shown that introducing bundles of care with prehabilitation, closer observation in the in-hospital phase of patients undergoing cancer surgery, may improve outcomes considerably.

These initiatives are resource demanding and it is crucial to identify the right patients for the right treatment at the right time of their surgical trajectory. In this presentation the early clinical results of implementing an AI based decision support in patients undergoing cancer surgery will be described. A special emphasis will be made concerning the importance of developing scalable solutions.

Bio
A Professor of Surgery at Copenhagen University heading Center for Surgical sciences, a translational research unit focusing on individualizing care in cancer surgery.

AI in Clinical Diagnosis
Encouraged by high AUC values in detecting cancer among GP office patients by neural network algorithms based on 23 biomarkers used in general practice , we created new algorithms based on clinical observational data and 62 biomarkers produced within one hour after the admission of acute patients in the emergency departments (EDs) of Kolding and Vejle Hospital ( Clin Chem Lab Med 2022;60(12):2005-2016 )

We found predictive values for early death, sepsis and 16 different outcomes of between 85 and 94%

The results have been validated and we are now ready to do a randomized controlled trial (RCT) on 16.000 patients to see if the use of AI will increase patient survival, save time,manpower and costs in the EDs ,burdened by an increased patient load

A reorganization of IT structure, installation of an AI platform, and new data servers for data storage and processing has been necessary as the Danish health care system is not prepared to the need of data access to different registers and 24/7 data flow from these to enable AI use in acute care

In a routine application we expect at least a 10-20% reduction in mortality, process time and cost reduction in the RCT

The costs of investment and RCT to document the utility are considerable and the lack of funding to this area may delay its clinical application

10+ years of AI in mental health - from correlations to disease prediction
In 2010, we started to investigate how the collection and analysis of sensor data collected from smartphones could be used in the diagnosis and treatment of mental health problems, with a particular focus on affective disorders like depression and bipolar disorder.

Over the last 10+ years, we have investigated this in great detail and, together with other researchers, shown how this kind of everyday sensing data combined with advanced data science methods can shed light on mental health problems. In this talk, I will survey this line of research and show how mobile sensing enables a unique insight into mental health patients' everyday lives and behavior, thereby revealing novel "digital biomarkers" for diagnosing, treating, and caring for mental health problems.

Safeguarding Accuracy: Clinical Examples Highlighting the Role of Quality Assurance in AI-Driven Radiology
In this Talk, I will explore the critical importance of quality assurance in the use of AI in radiology.

As AI technologies revolutionize medical imaging, ensuring their accuracy and reliability becomes paramount to patient care. Through clinical examples, I will highlight instances where AI has both succeeded and failed in diagnosing conditions. These examples underscore the potential risks of AI misdiagnosis, which can lead to severe patient harm.

To address these challenges, I will propose a comprehensive solution that includes validation protocols, continuous monitoring, and highlight the need for collaboration between clinicians and AI developers. By implementing robust quality assurance measures, we can harness the full potential of AI in radiology, enhancing diagnostic accuracy and ultimately improving patient outcomes.

PrivateGPT
PrivateGPT specializes in providing secure and compliant Generative AI solutions tailored for Danish businesses. By leveraging Microsoft Azure's platform within Europe, they ensure confidentiality and protection against unauthorized access to user data and interactions. Their services include custom GPT model development to meet specific business needs, seamless integration with existing Microsoft environments, AI-driven process optimization for improved performance, and expert support from a team of experienced AI professionals. Whether it's data analysis or customer service enhancement, PrivateGPT offers bespoke AI solutions and empowers businesses with both technology and training to maximize their benefits.

Next generation digital platform for accountable and confidential AI
Regulation poses a lot of requirements for the use of AI especially within healthcare where data is not just publicly available on the internet. This talk will discuss a digital platform designed to address the ideal properties for any digital infrastructure - confidentiality, integrity and availability - and how this addresses current and upcoming regulatory requirements. Concrete services and use cases will be used to showcase how to activate more sensitive data in AI and healthcare in a compliant way and for the greater good.

New opportunities and challenges using AI for mental health
The talk highlights new possibilities for treatment and monitoring made possible through use of sensors, digital tools, and artificial intelligence.

We illustrate cases like interventions for children with OCD, interventions based on suicidal risks, as well as diagnostics and preventive measures made possible through the new technologies. We will also discuss some of the challenges we face when using these tools.

Clinical decision support - A bumpy story about access to both retrospective and prospective clinical data in real time.

Treat Systems
Treat Systems is focused on developing and market software for antimicrobial stewardship. Our mission is to improve the clinical outcome of patients with infections while preserving the effectiveness of antimicrobials for future human generations. Our expertise lies in the understanding of both the clinical- and technical aspects of antimicrobial stewardship, and we combine this knowledge into innovative product development for the benefit of patients, clinicians, hospitals, and society. TREAT Steward™ is our current product which supports all aspects of antimicrobial therapy in clinical practice. Treat Steward™ is build upon our extensive knowledge of- and experience with health informatics, decision support and data mining technologies. We believe it is the best and most comprehensive software solution for Antimicrobial Stewardship on the market.

Treat Steward™ is an excellent tool for:
• Fighting bacterial resistance
• Improving guideline compliance
• Lowering the use of broad-spectrum agents
• Saving lives and reducing the number of bed days

Crash-testing AI in practice
How do we find the balance between deploying AI as fast as possible to reap the potential benefits, while at the same time doing so in the safe, secure, inclusive and sustainable manner that we expect. However, managing expectations is easier said than done, especially because Denmark is connected to Europe and to the rest of the planet, its people and markets. Martin Brynskov will give a global status and outlook, focusing on EU actions in a global context.

Bio
Orcid

Machine Learning and AI for improving Mental Health Care Delivery 
Our Precision Psychiatry work includes utilization of machine learning and AI for enhancing the precision of mental health care and for improving prevention, diagnoses and health care delivery in psychiatry.

The most promising decision support tools will be tested in clinical trials paving the way for implementation to improve the current treatments in psychiatry. Clinical testing includes deployment of the most clinically relevant decision support models directly into our EPIC-based Danish Electronic Health Record system, facilitating the clinical utility of the developed models.

As part of our Clinical Test Center for AI and data-driven treatments in psychiatry, we will furthermore investigate the needs and potential barriers among clinicians and patients to facilitate implementation and willingness to utilize decision support tools for assisting in optimizing our mental health care delivery.

AI Governance
Why everyone across industries have to start prepare for implementing this now, and the best way forward  to do so.

Career History
Mikael Munck is the CEO of 2021.AI and has 30+ years of experience from the technology sector, where he previously has been globally responsible for Technology and Operations at Saxo Bank as well as a number of CEO and General Manager positions at international technology companies.

Cerebriu
Cerebriu aims at simplifying and automating radiological processes and improving the efficiency of MRI scans using artificial clinical intelligence. In particular, they automate brain MRI workflows during scanning, thus simplifying the MRI process and improving efficiency, heightened accuracy, and time savings for radiologists.

Must Health AI be Explainable if it is Reliable?
Impressively accurate machine learning are being developed for clinical decision-support. A widespread concern is that the output of these algorithms e.g. diagnostic classifications, treatment suggestions, and risk scores cannot be explained to the relevant users e.g., doctors and patients.

In this talk I discuss why explanations should be required if the algorithm has been tested to be reliable. I relate the question to  norms of shared decision-making in medical practice and the use of drugs which are approved for use despite a lack of understanding of the underlying mechanism by which they work.

Career History
Sune Holm is associate professor in philosophy, Ph.D.(University of St. Andrews). Sune's current research focuses on questions concerning the ethics of AI, philosophy of biology, and bioethics. He participates in several national and international research projects on the use of AI in healthcare, and he is co-director of the Trustworthy AI Lab hosted by the Dept. of Datascience as well as a member of the scientific committee of the European Workshop on Algorithmic Fairness (EWAF). From 2016-2020 Sune was PI of the DFF2-project Living Machines? which examined philosophical issues relating to the machine-organism analogy.

Human Bytes
Human Bytes is an AI solution provider for the healthcare sector. Their mission is to bridge the gap between AI research and practical implementation, resulting in benefits for patients and the healthcare economy. Human Bytes deploys artificial intelligence across diverse healthcare domains and specialties in the Nordics, including radiology and mammography