AI as a Mirror: Reflecting and Reshaping Our Biases

In the reflective lens of AI, we find a poignant truth—our society's biases, long embedded in the structures of clinical research, have inadvertently shaped health outcomes. This revelation is especially stark when we consider the historical sidelining of women in clinical trials following the catastrophic thalidomide era of the 1950s. This led to the 1977 FDA recommendation to exclude women of childbearing age from early trials. Despite the fact that women want to participate in clinical trials, such biases have persisted subtly. Recent data has exposed that nearly two-thirds of contemporary trials still misconstrue sex-based eligibility, blurring essential distinctions between 'sex' as a biological attribute and 'gender' as a social identity. This ongoing confusion not only misrepresents diverse needs but also dims the promise of equitable health solutions.

However, in this challenge lies hope. AI, with its profound capacity to analyze vast datasets, offers a powerful tool to unearth and understand these biases, providing clear insights that can guide us towards more inclusive research methodologies. By integrating advanced AI and mathematical modelling, we can ensure that clinical trials reflect the genuine heterogeneity of human biology and societal roles, thereby enhancing the relevance and effectiveness of medical interventions across all segments of society.

The efforts of entities like the Diversity, Equity, and Inclusion (DEI) Committee of the Society for Mathematical Biologyor the Clinical Health Equity Collaborative Tiger Team (CHEC-TT) are heartening, as they champion the integration of DEI principles into the fabric of scientific inquiry. Their dedicated panels and discussions illuminate the path forward, advocating for a research ecosystem that embraces every individual’s unique background.

As we move forward, the fusion of AI with a robust DEI framework is not just a technical improvement but a moral imperative. It's about constructing a foundation of care that acknowledges and addresses the full spectrum of human diversity. This journey, powered by AI, inspires a vision of healthcare that is not only tailored, but also just; truly a beacon of hope for a healthier, more equitable future.

Responsible AI: Safeguarding Patient Care through Ethical Integration and Rigorous Governance in AI

In the evolving landscape of healthcare, the surge of AI technologies offers unprecedented opportunities for enhancing patient care, while also raising serious ethical and safety concerns. As an example, the Dana-Farber Cancer Institute's deployment of GPT-4 highlights the significant challenges and necessary measures, like robust governance and ethical oversight required to implement large language models in healthcare settings.

The complexity of navigating the regulatory environment is highlighted by Stanford’s review, underscoring the importance of compliance with diverse and sometimes conflicting international laws that govern patient data security and equitable treatment. Adding to this, the systematic review on AI medical applications points to the need for continuously updated guidelines that span both preclinical and clinical research to ensure AI's responsible use across various medical specialties.

Epic’s new AI validation software suite is a pivotal development, enabling healthcare providers to rigorously assess AI applications against the specific needs of local patient populations. Also, a new deep learning tool tackles 'black box' AI in medical imaging, explaining diagnostic decisions, and enhancing transparency in disease detection. Furthermore, the German Federal Office for Information Security’s report on generative AI models provides crucial insights into managing the risks associated with AI. This includes biased data and its potential misuse, advocating for heightened user awareness and stringent data management.

Together, these initiatives and guidelines form a comprehensive framework for the responsible integration of AI in healthcare, ensuring that technological advancements serve all patients with the highest standards of care and ethics.

Generative AI: Specialized Gen AI Models Enhance Medical Research and Clinical Decision-Making

Using small Generative AI models tailored for specific needs can be very effective. As an example, Stanford and Databricks researchers open-sourced BioMedLM, a specialized 2.7 billion parameter GPT-style AI model trained on PubMed texts to enhance biomedical research and healthcare applications. It also addresses challenges like cost, accessibility and data privacy. This is open source, and the team has shared that BioMedLM can be improved even further to produce insightful answers to patient inquiries about medical subjects. Another example is DrugGPT, an AI tool (type chatbot) created at Oxford University, seeks to reduce these errors by providing a safety net for both physicians and patients. Developed by researchers at Oxford’s AI for Healthcare Lab, this AI tool provides instant second opinions on prescriptions which helps clinicians make more informed decisions about medication choices. When it comes to LLMs, a study published in Nature tested ChatGPT's ability to translate English kidney transplant FAQs into Spanish, using two AI versions, GPT-3.5 and GPT-4.0. Both versions scored highly in linguistic accuracy and cultural sensitivity, indicating that ChatGPT could help provide vital transplant information to Spanish-speaking Hispanics, potentially improving health equity.

Other recent announcements:

• ?? Google launched new AI tools including Vertex AI Search for Healthcare and updates to its MedLM model, aimed at improving clinical data usage and workflow efficiency for healthcare professionals.
• ?? NYU Langone Health developed a generative AI tool that simplifies doctors' notes from hospital visits into patient-friendly language, making them more understandable by reducing the reading level from the 11th-grade to the 6th-grade level.
• ?? Oracle introduced generative AI tool and related upgrades, aiming to boost care management efficiency, enhance system performance and improve clinical analytics for healthcare stakeholders.

Drug Discovery: AI-Driven Innovations Propel Advances in Protein Engineering and Small Molecule Development

MIT researchers have devised a new computational strategy using convolutional neural networks to efficiently engineer proteins, predicting effective mutations with enhanced accuracy. This breakthrough could transform fields like neuroscience and medicine. Meanwhile, Ginkgo Bioworks' acquisition of AgBiome bolsters its R&D capabilities, integrating a vast library of microbial strains and gene sequences to advance biological discoveries through AI. In parallel, Johnson & Johnson's collaboration with Odyssey Therapeutics harnesses artificial intelligence and machine learning to innovate in the development of small molecule medicines, promising faster progress in clinical trials.

Clinical Development: Novo Nordisk Leads AI Integration in Pharma, Enhancing Drug Development and Patient Safety Standards

Novo Nordisk is pioneering the integration of AI in pharmaceutical drug development through collaborations with Valo Health, Microsoft and MIT, driving innovative high-tech solutions to transform healthcare and improve patient care.

While AI may not completely replace traditional drug development, those who leverage AI will certainly replace those who do not embrace this critical shift in the biopharma playbook. Faisal Khan, PhD, corporate vice president of AI and analytics at Novo Nordisk

Both the FDA and EMA are not seeking to legislate AI technology directly but rather initiate discussions within the research community on optimizing AI use in drug development, ensuring patient safety isn't compromised. The key difference is that the EMA emphasizes a risk-based regulatory scrutiny, while the FDA's guidance focuses more on the monitoring of AI systems and human oversight across the AI model lifecycle, from development to discontinuation. In addition, the rapid expansion of remote patient monitoring (RPM) technologies is significantly advancing the potential for decentralized clinical trials (DCTs) enabled by a number of DCT players. This allows pharmaceutical companies to monitor patient health data across various treatments more efficiently, which could lead to broader implementation and enhanced patient accessibility.

Lastly, the FDA's positive response to the use of real-world data in rare disease drug approvals emphasizes the growing importance of integrating diverse data sources to strengthen drug development processes, despite some concerns about data quality and comparability.

More on this topic:

• The scientists reviewed 20 applications for rare disease treatments that were submitted to the FDA between January, 2017 and October, 2022. All of them included real-world data.
• Wang and Schneeweiss advocate for conducting thorough data checks before preregistering real-world evidence study protocols to ensure the validity of study designs and address challenges unique to secondary health care data analysis, thus improving the integrity and quality of research findings.
• COTA, a leader in real-world data (RWD) and analytics for oncology, announces strategic collaboration with Sanofi to Leverage Real-World Data and AI to Accelerate Cancer Trials
• A study by the University of Exeter advocates for establishing clear guidelines for synthetic data to ensure transparency, accountability and fairness. This highlights the need for proper regulation to address risks like re-identification and potential misuse in AI applications. In conclusion, "synthetic data should be clearly labeled as such, and information about its generation should be provided to users."

Data Linkage and Data Sharing

The European Union approved the European Health Data Space to facilitate easier access and exchange of health data across member states, aiming to enhance healthcare delivery, research and policy making. In response, Datavant and Promptly Health have partnered to use this initiative to transform isolated medical data into valuable assets, initially focusing on the Iberian region with plans to expand into the UK and Sweden. This approach could mirror the success seen in the US by breaking down data silos using data linkage, thus enabling a more seamless flow of information that can be pivotal for advancing healthcare standards and treatments across Europe.

Datavant’s tokenization capabilities were crucial for real-world data companies scaling up in the US market, as it provided access to a broader data reach - General Manager, Europe and UK at Datavant, Devin Gilliam .

Sharing data within the scientific community boosts collaboration, ignites new research, and hastens medical breakthroughs by breaking down communication barriers, promoting transparency, and leveraging open-source programming and infrastructure. The Novartis-backed Multiple Sclerosis Implementation Network (MSIN) has partnered with Seqster?to enhance multiple sclerosis management by leveraging?Seqster’s data sharing platform to ensure timely and accurate treatment delivery to patients. In hospital systems, data sharing has significantly shortened the time required for patients to access their imaging records, with 80% of patients now viewing their results, thereby improving transparency and patient engagement. Ueli Rutishauser, a neuroscientist at Cedars-Sinai, advocates for widespread data sharing and collaborative research under the NIH BRAIN Initiative, aiming to deepen the understanding and educational impact of human brain studies. Lastly, Loren Frank's HHMI lab at UCSF has launched "Spyglass," a groundbreaking data analysis framework to facilitate reproducible and shareable neuroscience research. Frank‘s aim is to foster collaboration over competition in the scientific community by standardizing and sharing vast datasets and analysis methods. This initiative aligns with broader movements towards open science, supported by policy shifts like the OSTP's 2023 Year of Open Science, promising more effective and collaborative approaches to complex scientific challenges.

AI in Healthcare: Revolutionizing Diagnosis and Treatment with Cutting-Edge Artificial Intelligence Tools”

Artificial intelligence is reshaping healthcare across multiple fronts. A systematic review highlights AI's transformative role in nursing, enhancing patient outcomes and efficiency. The AI system Tyche is innovating in medical imaging and cancer detection by integrating diverse data sources for early diagnosis. Meanwhile, the ADA model excels in diagnosing post-traumatic stress disorder following childbirth with high accuracy. Additionally, the AI tool "Foresight," using data from NHS records, predicts future health conditions with impressive accuracy, aligning closely with clinician predictions. This advancement promises to significantly enhance decision-making in healthcare.

More news on Diagnosis and Digital Health from the past few weeks:

• ?? Swedish researchers have developed an AI model, LARS, that detects lymphatic cancer with 90% accuracy using image analysis from over 17,000 patient scans. The goal is to aid radiologists and improve diagnostic consistency.
• ?? A new AI tool, VA-ResNet-50, predicts fatal heart rhythms with 80% accuracy by analyzing Holter ECGs. It successfully identified patients at risk of ventricular arrhythmia, potentially improving early detection and treatment for this deadly condition.
• ??An AI tool can accurately detect pulmonary hypertension in newborns, a serious condition requiring swift treatment. Trained on ultrasound video data, the algorithm correctly diagnoses the condition with high accuracy and identifies its severity.
• ??An NHS pilot study shows the AI tool MIA, short for Mammography Intelligent Assessment, effectively detects early-stage breast cancers missed by specialists, promising improved early diagnosis and treatment outcomes.
• ??A new AI model called AsymMirai predicts cancer risk based on breast asymmetry, highlighting how differences between left and right breasts can indicate future health.
• [AI makes retinal imaging 100 times faster compared to manual method](https://www.nih.gov/news-events/news-releases/ai-makes-retinal-imaging-100-times-faster-compared-manual-method#:~:text=Researchers at the National Institutes,improves image contrast 3.5-fold.)s. NIH scientists use artificial intelligence called ‘P-GAN’ to improve next-generation imaging of cells in the back of the eye.

Advancing Technology: Key Innovations in Data Management and AI Across Industries

Highlights from the past few weeks:

• The HIMSS 2024 conference highlighted how high-performance computing (HPC) and cybersecurity are crucial for transforming healthcare data into actionable insights, improving patient care, and simplifying infrastructure management.
• Snowflake has introduced Data Clean Rooms, integrating Samooha's technology into their Native App Framework. This integration offers data engineers a solution for secure, efficient, and collaborative data analysis across clouds without compromising privacy.
• AVEVA and Databricks have formed a strategic partnership to revolutionize industrial software by integrating their platforms for enhanced data sharing and AI-driven insights. The goal is to accelerate the digital transformation of industries.
• Tempus introduces Workspaces on its Lens platform, a coding-based cloud environment trusted by nearly 100 institutions.
• A New Study Compares RAG & Fine-Tuning For Knowledge Based Use-Cases
• The Business Guide to Tailoring Language AI: A framework for unlocking custom LLM solutions you’ll understand
• Meta has launched Llama 3, a new series of open-source large language models with options of 8 billion and 70 billion parameters, integrated into major platforms like Google Cloud and Azure AI for enhanced AI services. These models, available for both academic and commercial use, promise improved flexibility, performance, and ethical usage, with future upgrades anticipated
• GPT-4 Turbo OpenAI has released GPT-4 Turbo, the latest version of its language model, available to paid ChatGPT subscribers. This update includes support for up to 64k tokens, significantly more than previous models. Key enhancements include improved capabilities in coding, logical reasoning, writing, and mathematical computations. GPT-4 Turbo introduces new features like function calling and JSON mode in vision capabilities, which allow for structured data extraction from images. The model also provides responses that are more concise and conversational. This version updates its knowledge base to December 2023, demonstrating improved performance in benchmarks like the HumanEval and retaining a leading position on the Arena leaderboard.

Upcoming 2024 Conferences

• Swiss Biotech Day 2024, at the Congress Center Basel, Switzerland, April 22-23, 2024
• HIMSS Middle East Executive Summit, Dubai , April 25th 2024
• LSX World Congress 2024, London, UK, April 29th-30th
• HIMSS24 Europe, Roma, Italy, May29th-31st, 2024
• 5th MENA Telehealth & Virtual Care Expo, Riyadh,? May 7th and 8th
• Abu Dhabi Global Healthcare Week, Abu Dhabi May 13th-15th, 2024
• IFCC WorldLab 2024, Dubai, May 26th-30th
• Ai-Med Global Summit 2024: Florida, USA, May 29-31, 2024
• 82nd FIP Wolrd Congress for Pharmacy and Pharmaceutical Sciences, Cape TownSeptember 1st-4th
• Pharma Days 2024, Geneva, Switzerland, June 5th-6th, 2024
• HLTH Europe, Amsterdam, Netherland, June 17th-20th 2024
• Future Labs Live, Basel, Switzerland, June 25th-27th, 2024
• BiotechX USA 2024, Philadelphia, PA, September 10-11, 2024
• Intelligent Health AI, Basel, Switzerland, September 11-12, 2024
• BiotechX Europe 2024, Basel, Switzerland, October 9-10, 2024
• Med Tech World, MALTA, November 6th, 8th, 2024
• 6th GCC eHealth Workforce Development Conference, Dubai, November 6th-7th, 2024
• DigiHealthDay 2024, Pfarrkirchen, Germany, November 11th-15th
• GIANT Health 2024 , London, December 9th-10th, 2024

Other Great Newsletters 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