Time for action

Following the 2024 predictions and the JPMorgan healthcare investor conference, where AI's potential in healthcare faced both enthusiasm and skepticism, it is imperative that the sector pivots from forecasting to real-world implementation. This period is crucial for translating AI's theoretical benefits into practical, demonstrable outcomes in healthcare. It is a time for action, - emphasizing the necessity to actualize the optimistic projections of AI, and ensuring they materialize into meaningful improvements in healthcare services and outcomes. This phase is marked by a proactive approach to embracing technological advancements while addressing "Responsible AI", real-world challenges and proving the efficacy of AI in healthcare.

Having recently watched a movie about Oppenheimer and subsequently engaging in discussions with my friends about the parallel between the dual-edged nature of nuclear energy and AI, I found myself deeply intrigued by Dr. Giovanni Piedimonte's perspective. As the Vice President for Research at Tulane University, he expresses ambivalence regarding the advent of AI in healthcare, drawing his own parallels to Oppenheimer. This analogy is particularly striking when considering how AI, much like nuclear energy, can be harnessed for beneficial purposes like creating healing proteins, yet also holds the potential for harm, such as in the development of new chemical weapons. The coming debate with Yann LeCun "Should we slow down research on AI" at World AI Cannes Festival 2024 will be of utmost interest, as well as the presentation from Nayur Khan "AI in the C-Suite".

2024 starts with Responsible AI on the top of the agenda

The Davos elite embraced AI in 2023. Now they fear it. At the World Economic Forum 2024 in Davos, EU Health Commissioner Stella Kyriakides emphasized the need for safeguards in the use of AI-driven tools in healthcare, including maintaining human oversight. Still at the WEF, the state of health and healthcare was discussed and the increasing use of AI in clinical trials and administrative tasks was a key topic, with panelists from Mayo Clinic and Takeda Pharmaceuticals underscoring its significant role in enhancing healthcare efficiency and productivity. They stressed the need for reliable outcomes, emphasizing the criticality of self-regulation, thorough validation, and robust regulatory frameworks, such as the EU's AI regulation act, to ensure the efficacy and trustworthiness of AI in healthcare. Still on the WEF website, one can read on Narrow AI, which unlike GenAI, focuses on specific tasks and doesn't aim to replicate human intelligence in its entirety. NarrowAI is increasingly being integrated into healthcare to improve surgical practices, diagnostics, and patient care. Building trust in narrow AI is crucial, hinging on three pillars: reliable and unbiased data, transparent regulation and independent certification, and strict enforcement of data privacy and protection.

The World Health Organization (WHO) has released comprehensive guidelines for the ethical use and governance of large multi-modal models (LMMs) in healthcare, highlighting the rapid growth of generative AI technologies like chatgpt, Bard and Bert. The WHO also highlights concerns that AI developed in wealthier nations may not be effective or appropriate in developing countries, emphasizing the need for different considerations in performance, data training, and care standards.

"The very last thing that we want to see happen as part of this leap forward with technology is the propagation or amplification of inequities and biases in the social fabric of countries around the world,” Alain Labrique, the WHO’s director for digital health and innovation, said at a media briefing today.

Jeremy Petch, PhD , director of digital health innovation at Hamilton Health Sciences, discusses the black box models and strategies for ensuring transparent and explainable AI. To overcome lack of trust in AI due to an inability to explain the results of an AI algorithm and the "black box" nature of current models, several startups are are trying to build new AI architectures that fix a lot of the big problems around trust and reliability of today’s latest models.

1. Umnai: Based in London, led by CEO Ken Cassar, Umnai is working on a new AI architecture designed to improve accuracy and reliability in AI systems.
2. Aligned AI: Located in Oxford, this startup, co-founded by Rebecca Gorman and Stuart Armstrong, is not creating new AI models from scratch, but enhancing existing ones to improve their reliability and quality of output.
3. Conjecture: Founded by Connor Leahy in London, Conjecture is developing an AI approach called "boundedness" to provide a safer and more controllable alternative to general-purpose AI systems like GPT-4.

AI in life sciences R&D

Accelerating Drug Discovery: AI Innovations from Insilico Medicine to Cambridge University

While Insilico Medicine, an AI-driven biotech company based in Hong Kong and in New York City, recently announced its fifth AI-designed drug — ISM5411 — has entered Phase I clinical trials, the typical lag between raw scientific discovery and patient-ready clinical indication is still around 17 years. To address this, Cleveland Clinic and IBM joined forces in 2021 to try to shorten the waits. They called their collaboration the “Discovery Accelerator", a partnership initiated in 2021 which focuses on enhancing biomedical research through high-performance computing, AI, and possibly quantum computing. Three years later, Cleveland Clinic and IBM Researchers have just published Findings on Artificial Intelligence and Immunity in "Briefings in Bioinformatics" : "Unsupervised and supervised AI on molecular dynamics simulations reveals complex characteristics of HLA-A2-peptide immunogenicity"

Meanwhile, Researchers at the University of Cambridge have developed an AI-driven 'reactome' platform, revolutionizing chemical reaction prediction and accelerating drug discovery. This approach, validated by over 39,000 reactions, shifts from traditional methods to machine learning, enhancing understanding of chemical reactivity. The team's breakthrough in machine learning for late-stage functionalization in drug design overcomes data scarcity challenges, offering precise molecular modification capabilities and paving the way for advanced pharmaceutical innovation.

Revolutionizing Pharma R&D: AI in Clinical Trials

The McKinsey report titled "Accelerating clinical trials to improve biopharma R&D productivity" highlights the issue of low R&D productivity in the biopharmaceutical industry, primarily due to extended clinical trial durations and escalating costs. The report indicates that the biggest opportunity for sponsors to accelerate clinical trials is to increase the speed and improve the efficiency of clinical trial enrollment.

McKinsey has also released a report on "Generative AI in the pharmaceutical industry: Moving from hype to reality" stressing the need for the industry to learn to scale this technology and address the industry’s unique challenges. Seeing the value potentially generated, it is imperative to leverage this one-time opportunity!

If you do not know where to start with Generative AI, you can check this article which provides a comprehensive guide for CIOs on effectively integrating generative AI into business operations, focusing on balancing innovation with risk management, aligning with business goals, and addressing ethical and technical challenges to harness AI's transformative potential.

Lastly, keep the limitations of AI in mind: The study by Chekroud et al. emphasizes a crucial caution in clinical trials: AI models show high accuracy within their development datasets but their performance dramatically drops when applied to new, independent clinical trials. This points to a significant gap in the predictive capability of AI models in clinical settings, highlighting the need for more robust testing and validation methods to ensure their reliability in diverse clinical trials.

Creating the right data foundation for your clinical trial platform to better leverage AI

All this is nice, but you need to first create the right data foundations. Raj Indupuri , CEO of eClinical Solutions, discusses the transformative impact of AI and Generative AI on various industries, particularly in Life Sciences, emphasizing the critical role of data in harnessing AI's full potential. In his last LinkedIn post, he acknowledges the challenges companies face with traditional data management systems and introduces eClinical's Clinical Data Lakehouse architecture as a modern, flexible solution for overcoming these challenges, thereby enabling more effective use of AI in clinical development. One of its competitors, Saama, has released new features in its AI-driven platform, including generative AI chat and Interactive Review Listings, evolving data analysis and collaboration in clinical development by enabling natural language queries and comprehensive data reviews in a unified system.

Nevertheless, Saama and eClinical are well-positioned to benefit from the expanding Clinical Data Management System Market, which is projected to grow from USD 2.1 billion in 2022 to USD 2.7 billion by 2030, achieving a Compound Annual Growth Rate (CAGR) of 12.9% from 2023 to 2030.

Integrating a robust clinical data lakehouse with an open data model and conversational AI could be a breakthrough in healthcare in the coming years. Adding the ability to link clinical trial data with real-world data (RWD) using data tokenization, as discussed by Kevin Pho, M.D. and Adam Mariano, Esq, MSPH , in their podcast, could significantly advance clinical data management. This combination promises to enhance both the efficiency and the scope of medical research and in patient care.

AI in Healthcare

Balancing Risks and Breakthroughs in Empathy, Accuracy, and Radiology

According to Kodiak Solutions' annual report, AI and new technologies are among the top five management risks for healthcare providers in 2024, alongside financial performance, competition, workforce, and cybersecurity challenges, with implications for legal, reputational, and operational aspects in healthcare organizations. On a more positive side, Google Claims Healthcare AI More Empathetic, Accurate Than Real-Life Doctors. Its new AI system, AMIE, designed for medical interviews and diagnostics demonstrated higher accuracy than human doctors in certain diagnoses and scored better on empathy. However, it was tested on actors, not real patients, and requires further research before real-world application. ChatGPT-4, has also shown promising results in identifying incidental findings on CT scans, according to research published in the American Journal of Roentgenology. Trained using a method called "single-shot learning," ChatGPT-4 effectively differentiated between key radiological elements, matching the performance of trained radiologists, the latest version of ChatGPT -ChatGPT-4, has shown promising results in identifying incidental findings on CT scans, according to research published in the American Journal of Roentgenology.

Partnerships and AI transforming healthcare

??Nuance AI's integration of its AI copilot into Epic EHR marks a significant advancement in healthcare IT, streamlining clinical documentation and workflows, thereby enhancing both clinician efficiency and patient care

??Innovaccer, a company focused on value-based care and patient experience, acquires customer engagement platform Cured to enhance patient experience.The strategic move will add more than 20 health systems and first-time digital health clients to Innovaccer’s current portfolio of customers. In addition, Innovaccer also collaborates with Wolters Kluwer to harness the Health Language Platform, aiming to optimize healthcare data utilization and improve patient care by transforming disparate data into meaningful, standardized insights.

??SOPHiA GENETICS and Karkinos Healthcare have formed a strategic partnership to advance cancer research and care in India, utilizing the SOPHiA DDM? Platform for genomic analysis of various cancers, focusing on improving diagnostics and treatments, particularly for underserved populations.

??Servier and Aitia Announce a New Collaboration Focused on Parkinson’s Disease: Aitia‘s Gemini Digital Twins will help identify subpopulations of patients who could respond favorably to Servier’s LRRK2 in development treatment

??In a pivotal stride toward advancing cancer care, Qure.ai, a leader in artificial intelligence (AI) solutions for medical imaging, and Project Data Sphere?, a nonprofit initiative of the CEO Roundtable on Cancer, have announced their partnership to augment tumor assessments using AI-enabled solutions for clinical trials and cancer care. The collaboration aims to increase efficiency and consistency in evaluating the effectiveness of cancer treatments, ultimately improving the quality of care for patients.

??Insitro, a privately held startup that has raised more than $643 million, described its early progress at using machine learning to identify new drug targets in liver disease, ALS, and cancer.

Revolutionizing Informed Consent in Healthcare: The Journey from Pharmaledger to Novartis-Astrazenca Blockchain collaboration

I believe it’s safe to say that technology typically takes longer to adopt than one anticipates: In 2018, I advocated for the creation of an Industry group focused on blockchain in healthcare. Through the collaboration of several colleagues from a variety of organizations, an industry consortium Pharmaledger was born. Fast forward to January, 2024 – I was thrilled to come across a publication by Novartis’ Xavier Briand describing the collaboration between Novartis and Astrazenca focused on blockchain based information consent. This alliance not only illustrates the transformative potential of blockchain and decentralized technologies focused on improving the informed consent process in clinical trials, but it also enhances data security, and empowers participants with control over their digital identities. While promising, the integration of these technologies face challenges that require collaboration and standardization across the healthcare industry.

Still on informed consent, LMMs can be used to improve the readability of informed consent documents as presented in this case study "Using ChatGPT to Facilitate Truly Informed Medical Consent".

Engineering

??3 Advanced Document Retrieval Techniques To Improve RAG Systems: query expansion, cross-encoder re-ranking, and embedding adaptors, aiming to align retrieved documents more closely with user queries and improve the overall effectiveness of RAG systems.

??Transforming text into conceptual knowledge graphs describes the process of using Mistral 7B, a sophisticated natural language processing system. This process transforms text into conceptual knowledge graphs and highlights steps from text corpus processing to graph creation and analysis. The result is a practical example in medical research which demonstrates how this technology can efficiently organize and visualize complex information.

??Self-Learning Knowledge Graph RAG: Demonstration of 3 key powerful applications of knowledge graph in RAG, and how it can improve RAG accuracy, dramatically reduce time to production, and is one of the first examples of a self-learning RAG.

??Top 10 Data & AI Trends for 2024 from Barr Moses :From LLMs transforming the modern data stack to data observability for vector databases; See her predictions for the top data engineering trends in 2024.

??I Tried Multiple AI Coding Assistants. These Are The Best!: Best AI coding assistants for beginners and experienced programmers.

Upcoming 2024 Conferences

• World Ai Cannes Festival, Cannes, FRANCE, February 8th-10th
• Scope summit, driving innovation in clinical trials, Orlando, USA, February 11-14, 2024
• Ai-Med Global Summit 2024: Florida, USA, May 29 - 31, 2024
• BiotechX USA 2024, Philadelphia, September 10, 2024 to September 11, 2024
• Intelligent Health AI, Basel, September 11th, 2024 to September 12, 2024
• BiotechX Europe 2024, Basel, Switzerland, October 9 to October 10, 2024
• Med Tech World, MALTA,November 6th, 8th 2024

Other Great Newsletter on healthcare and AI

Stay up to date with the latest insights and industry trends by subscribing to these informative newsletters from my esteemed colleagues and professionals in the field.

• Healthcare in the Metaverse?by?Philippe GERWILL
• DigitalhealthReadingList?by?Wolfgang Schleifer
• Future Tech Trends?by?Bernard Marr
• All about Blockchain Adoption?by?Kamlesh Nagware
• AI Frontier?by?Steve Nouri
• The Atlas by Generative AI