AI has now the attention of CEO!

I had the opportunity to explore the canyons in the American West over the past two weeks and reflect on the concept of time, the grandeur of nature, and the steps that humanity has taken and has yet to take. Against this backdrop, the news from the last two weeks may seem insignificant in the grand scheme of things, but every contribution plays a part in shaping the future advancements of healthcare technology. The last two weeks have been quite intense for the Tech Healthcare industry with many conferences ( in Paris, Global Annual meeting in Boston, in Vegas) and many articles in the press. In an effort to summarize all the recent developments, I have done my best to capture the key points below.

Artificial Intelligence

AI has now the attention of CEO

While Sanofi Goes All-In On AI, The CEO of pharma giant Eli Lilly shares 3 ways AI could transform his industry and the recent traction of ChatGPT, Bard and other generative AI technologies has caught the CEO's attention. Many articles detail the use cases for AI in healthcare, and as an example the WEF Report Addresses the Role Of AI In Healthcare and Pharmaforum publishes on AI in healthcare. As mentioned by John Nebergall in Health data management, “We’re at the beginning of the hype cycle for AI in healthcare, and the push to adopt the latest offering — namely, ChatGPT — is strong. But determining the right use cases for large-language models in healthcare is, at best, a work in progress.” And, for those interested to go deeper into Generative AI, Andrew Ng has just released a new course on created with : Generative AI with Large Language Models. This course goes deep into the technical foundations of LLMs and how to use them. If you are interetsed, you can sign up here. For the one looking for a book for this summer, you can follow Vas Narasimhan recommendation " The AI Revolution in Medicine: GPT-4 and Beyond" by Peter Lee, Carey Goldberg, and Isaac Kohane.?

Causal AI

Causal AI, also known as Causal Inference or Causal Machine Learning, is a field of artificial intelligence that aims to understand cause-and-effect relationships within data. It goes beyond correlation and attempts to identify the underlying causal factors driving observed outcomes. In the context of healthcare, causal AI has significant implications. It can help researchers and healthcare professionals gain insights into the effectiveness of treatments, interventions, and policies by determining the causal impact of specific factors on patient outcomes. It goes beyond identifying associations and aims to establish causal relationships between variables. One of the leader, Aitia ( was rebranded as beginning 2023), will present at the upcoming Alzheimer’s Association International Conference (AAIC) being held from July 16-20 in Amsterdam, Netherlands "Alzheimer’s Disease Mechanistic Pathways Discovered through Aitia’s Digital Twins and in silico Experiments in Causal AI to be Presented at AAIC 2023". The global causal AI market is projected to grow from USD 8.01 million in 2023 to USD 119.5 million by 2030, with a CAGR of 47.1% during the forecast period. The Global Causal AI Market Report 2023 has been released and the healthcare and life sciences sector is expected to be the largest market, driven by the adoption of causal AI in drug discovery, patient diagnosis, personalized medicine, and more, while on-premises deployment and training, support, and maintenance services are anticipated to contribute significantly to market growth.

AI assisting healthcare personnel

AI-assisted healthcare offers numerous benefits such as improved diagnosis accuracy, personalized treatment plans, and increased efficiency. However, it also comes with risks including privacy concerns, algorithmic bias, and overreliance on technology. The use cases for AI in healthcare are expanding rapidly, but cautious implementation is necessary to ensure ethical considerations and human oversight, maximizing the benefits while mitigating potential risks. A number of announcements have been done in the last two weeks:

• Pathology AI: In certain areas, AI algorithms have demonstrated comparable or even superior performance to human pathologists in specific tasks. For example, AI models have shown high accuracy in detecting certain types of cancers from histopathology images. These AI algorithms can analyze large volumes of data quickly, potentially improving efficiency and reducing the burden on pathologists. In this context, Paige, a digital diagnostic company, is expanding its suite of AI-enabled solutions for breast cancer diagnosis. The suite, known as Paige Breast Suite, includes tools such as Paige Breast Detect, Neoplasm, Mitosis, Lymph Node, and HER2Complete, aimed at improving the efficiency and accuracy of pathologists' workflows. The AI-powered tools help detect breast cancer metastasis, prioritize slide reviews, and assist in mitotic counting, enhancing pathologists' confidence and reducing subjectivity in diagnosis.
• Google's new ai describes x-rays and answers patient questions - ICT&health: Google has unveiled PaLM 2, an AI platform for analyzing medical data. It aims to assist doctors with routine tasks and provide more reliable answers to patient questions than “Dr. Google.”
• New machine-learning model can detect dementia in speech: Scientists have conceived of a ML model capable of detecting speech patterns that are linked to a diagnosis of Alzheimer’s or other forms of dementia. The new tool will be used for early evaluation of the conditions.
• Predicting Dementia in Parkinson’s: EEG Might Hold the Key - Neuroscience News: A few minutes of data recorded from a single electrode placed on top of the head may be sufficient to predict thinking problems, including dementia, in patients with Parkinson’s disease (PD). The non-invasive and inexpensive nature of EEG makes it a potential tool for diagnosing cognitive impairment in Parkinson’s patients.
• 'AI doctor' better at predicting patient outcomes, including death - Japan Today: A new AI tool has demonstrated the ability to read physicians' notes and accurately anticipate patients' risk of death, readmission to hospital, and other outcomes important to their care. Designed by a team at NYU Grossman School of Medicine, the software is currently in use at the university's affiliated hospitals throughout New York, with the hope that it will become a standard part of health care.

Responsible AI

All this is good, but Zac Amos writes about Who Is Responsible If Healthcare AI Fails? Depending on the situation, it could be the AI developer, a healthcare professional or even the patient. Liability is an increasingly complex and serious concern as AI becomes more common in healthcare. Who is responsible for AI gone wrong and how can accidents be prevented?

On the same length, Diagnostic Robotics, a provider of AI solutions for healthcare, has filed a public comment in response to the U.S. Department of Commerce's request for expert feedback on developing an AI accountability ecosystem. The company highlights concerns about bias in AI systems and the need to differentiate between fake and real data, emphasizing the importance of ethics guidelines and accountability measures to promote trustworthy AI in healthcare and other critical domains

Data

Tokenization and data linkage

Data linkage with tokens, as done by companies like DataVant and HealthVerity, is a method used to connect and correlate data from multiple sources while preserving privacy. It involves the use of tokenization techniques generated using attributes such as first name, last name and date of birth. The tokenization process ensures that the same individual or entity is consistently represented by the same token across different datasets.This is a useful technique for Pharma to enrich trial analysis with get long-term follow-up on efficacy and safety data, or perform retrospective cohort analysis. Datavant made several announcement: a partnership with Atropos Health, the leader in generating personalized real-world evidence at the point and pace of care, that will extend access to the Atropos Evidence Platform to members of the ecosystem. Another collaboration with OMNY Health has been announced?to Accelerate Delivery of Real-World Data between Healthcare Stakeholders to Enhance Clinical Research and Drug Development

Zero trust and confidential computing

Zero-trust and confidential computing capabilities are crucial in healthcare to protect patient data and ensure privacy. emphasizes continuous authentication and access controls, treating all users and devices as untrusted. Confidential computing secures data while it is being processed, allowing for secure collaborations and data sharing. By implementing these approaches, healthcare organizations can strengthen their security posture, enable secure data processing, and protect patient privacy in an increasingly connected healthcare landscape.

#BeekeeperAI has raised $12.1 million in Series A funding to accelerate the development of AI on privacy-protected healthcare data. The company's EscrowAI platform, which integrates Azure confidential computing capabilities, enables secure collaboration workflows between AI developers and privacy stewards, addressing challenges related to data sovereignty, security, and intellectual property protection. The funding will be used to expand the features of EscrowAI and support the growing demand for zero-trust computing environments in healthcare and regulated data industries.

Data Sharing

Databricks Announces New Partners to Accelerate Data Sharing: has announced new partnerships with Cloudflare, Dell, Oracle, and Twilio to expand its Delta Sharing ecosystem, allowing for secure and open data sharing across platforms. Delta Sharing enables organizations to share and consume live data sets without dependencies on specific services, facilitating collaboration and accelerating data-driven insights.

The National Institutes of Health (NIH) has released a new data-management and data-sharing policy to address challenges with public health data interoperability. The policy aims to increase accessibility and improve data interoperability across the diverse IT systems within NIH and its network of partners in the public health community. Implementation of the policy will depend on effective collaboration, community engagement, and the ability to adapt as the policy evolves. Note this interesting NIH publishing on understanding the value of secondary research data

Data Privacy and security

• Protecting Privacy and security of DNA data: Protecting DNA data presents numerous challenges due to its highly personal and complex nature. Managing consent and withdrawal of consent requires careful attention, while ownership concerns emerge from ethical and legal considerations. Balancing data sharing and interoperability with privacy poses additional challenges, and measures must be taken to safeguard against potential genetic discrimination. Addressing these challenges requires robust security measures, but as well clear regulations and promoting public awareness to ensure responsible handling of DNA data. A Genetic-Test Firm is Accused of Exposing User Data in FTC First: The Federal Trade Commission (FTC) has taken action against genetic testing firm 1health.io for failing to secure sensitive data and deceiving consumers about privacy. According to the complaint “Vitagene did not encrypt that data, restrict access to it, log or monitor access to it, or inventory information to help ensure its security" In a proposed settlement, 1health.io must enhance data protection measures and instruct third-party laboratories to destroy stored DNA samples, along with facing a $75,000 fine and providing consumer refunds.
• A new health data breach: At least 100,000 could have had data exposed after US health department was hit by global cyberattack: The US Department of Health and Human Services (HHS) is the latest government agency to be targeted in a cyberattack linked to Russian cybercriminals, potentially compromising data of over 100,000 individuals. The attack exploited a vulnerability in third-party software, highlighting the ongoing threat of cybercrime and the need for enhanced cybersecurity measures.

Platform

Clinical Trial Platform

Integrating more AI into clinical trial platforms holds significant potential benefits, particularly in automating clinical data review and management processes. By leveraging AI technologies, such as natural language processing, generative AI and machine learning, clinical trial platforms can streamline and expedite the review of large volumes of clinical data, enhance data quality and accuracy, and improve overall efficiency in managing clinical trial information. This can ultimately lead to faster and more effective decision-making, reduced administrative burden, and enhanced patient safety in the drug development process. Several companies are making progress on this front:

• Saama Launches Industry’s First AI-driven Data Platform to Accelerate Clinical Development | Business Wire: Saama, a provider of AI- and ML-based solutions for clinical development, has launched a unified platform of SaaS-based products that automate clinical trial processes and offer real-time insights into trial operations and patient behaviors. The platform reduces query generation times, streamlines data transformations, and accelerates the analysis of clinical data, enabling greater efficiencies and visibility for life science organizations.
• At , Syneos Health and uMotif Partner announced that they have formed a partnership to collaborate on a platform that aims to improve patient-centricity and data collection in clinical trials. The platform will leverage uMotif's patient-centric data capture technology and Syneos Health's clinical trial expertise to enhance patient engagement, provide real-time insights, and streamline data collection processes, ultimately improving the efficiency and effectiveness of clinical trials.

Engineering

Technology and engineering play a pivotal role in healthcare and AI, which is why I have decided to introduce a dedicated section in this newsletter to highlight some the latest news and advancements in this field. `

Calibo

Calibo offers an exceptional self-service platform that enables seamless orchestration and automation of any tech stack, resulting in over 50% reduction in time-to-market for data and digital applications. Having personally tested this solution last year, I can confidently say that it is an excellent platform for managing the lifecycle of digital products, automating deployments, and monitoring solutions. Moreover, the highly skilled and customer-focused Calibo team ensures a smooth experience. If you are seeking to streamline your end to end digital product lifecycle, to deliver faster, to manage the complexity of the technology stacks while deploying, and to have everything in one place (From design to deployment), is definitely worth exploring. Following snowflake Vegas 2023, more on Calibo here

OpenAI Wants to Be an App-Dev Platform

New York — Attendees of QCon in New York got a preview of where the exciting world of Large Language Model-based AI may be going.

OpenAI showcased a new feature called Functions for ChatGPT, allowing developers to connect the language model with external services and perform actions on behalf of users. By expanding the capabilities of ChatGPT through Functions, OpenAI is addressing a previous limitation of the model's knowledge, extending its usefulness beyond its pre-2022 knowledge base. Indeed, by integrating with third-party services like Yelp, ChatGPT can retrieve real-time data and provide formatted results, indicating OpenAI's intention to position itself as an app development platform for developers. More on using ChatGPT functions here and a code snippet below: