AI and Medicine Weekly
Welcome to our blog! We believe that knowledge is the cornerstone of good health. Our mission is to provide clear, accurate, and engaging content on a wide range of medical topics, from the latest breakthroughs in healthcare to practical tips for everyday wellness. Whether you're a healthcare professional, a patient, or simply someone curious about health, you'll find resources here to empower and inform you. Together, let's navigate the fascinating world of medicine and make informed choices for a healthier future.
AI & Medicine Weekly Digest
Saturday, March 15, 2025
Researchers at Mass General Brigham have developed an AI tool capable of predicting cognitive decline years before symptoms manifest. By analyzing brain wave patterns recorded during sleep using EEG technology, the tool flagged 85% of individuals who later experienced cognitive decline, achieving an accuracy of 77%. Early detection could enable lifestyle interventions to preserve cognitive health and offers potential for altering EEG patterns during sleep to reduce dementia risk.
Pharmaceutical leaders such as Johnson & Johnson, Merck, and Eli Lilly are prioritizing AI literacy among their workforce to integrate advanced technologies into drug development and operations. Initiatives include mandatory AI training for employees, development of proprietary AI platforms, and extensive use of AI tools like ChatGPT. These efforts aim to enhance efficiency, reduce costs, and shift recruitment priorities towards AI fluency alongside traditional biotech skills.
St Vincent's Hospital in Sydney successfully performed Australia's first BiVACOR total artificial heart implant on a patient with severe heart failure. The patient, who underwent the six-hour procedure in November 2024, received a donor heart in early March 2025, marking the longest duration between receiving an artificial heart and a transplant. This breakthrough could revolutionize heart failure treatment and significantly reduce transplant wait times.
Google DeepMind introduced two advanced AI models, Gemini Robotics and Gemini Robotics-ER, to enhance robots' capabilities in performing real-world tasks. Leveraging the Gemini 2.0 model, these AI systems enable robots to understand diverse scenarios and perform precise tasks such as folding paper or unscrewing bottle caps. Safety measures are integrated, with AI assessing the safety of actions before execution, aiming to develop more intelligent, responsive, and robust robots adaptable to varied environments.
"New AI tool predicts brain decline before symptoms appear" – New York Post
"Pharmaceutical companies embrace AI in drug discovery efforts" – Business Insider
"Unmitigated success: Australian medicine celebrates major breakthrough" – News.com.au
"Google DeepMind Gemini Robotics AI models" – The Verge
Stay informed with the latest advancements at the intersection of artificial intelligence and medicine.
Our approach is "Patient first" unfortunately many see healthcare as a business opportunity. Below is a white paper that analyzes the corporatization of American healthcare and its impact on marginalized communities. It points out rising costs, reduced access to quality coverage, and systemic inequalities.
Drug Discovery
Saturday, March 8, 2025
Written by Keynote Contributor Soroosh Afyouni, head of health data science at bioXcelerate AI.
Every household seems to have that drawer of miscellaneous belongings. When you need something, you know it’s in there, but it still takes hours to sift through the clutter you should’ve tossed years ago. The massive amount of data being stored in ever-expanding biobanks is a bit like ‘that drawer,’ and it makes finding consistently good quality data increasingly challenging.
Living in the digital age, there’s no shortage of information available, and organizations of all sorts are making it their business to use it. This is especially true now, with advanced technology allowing businesses to analyze vast swathes of data in a way that, up until recently, simply wasn’t possible.
Over the past few decades, healthcare authorities worldwide have established large biobanks, including patient medical records, clinical trial data, and genetic information. While these biobanks hold immense value, the data quality is often inconsistent and comes in a variety of shapes and formats, making it difficult to draw reliable comparisons. Trying to analyze them side by side is like comparing apples to oranges – it just doesn’t work.
This inconsistency is becoming a significant issue. It not only hampers collaboration but also limits the potential of powerful tools like AI and machine learning to deliver meaningful insights. That’s why it’s important to develop a strategy for managing health data globally to streamline the drug development process and, in turn, create more effective outcomes for patients.
There’s no denying that data offers immense opportunities to transform a range of industries, especially healthcare. However, if data quality is poor, it can hinder how effectively the data can be used. According to a study by Sun et al. (2022)1, up to 50% of clinical trials fail due to a lack of clinical efficacy. Put simply, if the initial insights aren’t based on strong data-driven evidence, then the treatment is unlikely to be optimally effective.
As advances in technology, such as AI, continue to enhance our ability to extract insights from data, the drive to collect it at an unprecedented pace – and the potential to revolutionize drug development – has never been greater. However, the technology is only as powerful as the data it processes; without high-quality, well-curated data, even the most advanced AI models can falter.
Inconsistent formats, varying degrees of accuracy, and incomplete datasets are common challenges that, without proper attention, can undermine the effectiveness of data-driven insights2. Disparities like these can result in spurious conclusions, ineffective treatments, and missed opportunities to address patients’ unmet needs.
Establishing a repository of health data that is accurate, comparable, and reusable is a vital step for maximizing the potential benefits of AI and machine learning in drug development, and reproducibility is crucial to achieving this4. When scientists can revisit previous work and obtain the same results, findings are verified, and trust in the insights drawn from the data is built. Additionally, reproducibility helps minimize errors by revealing inconsistencies in results. Finally, adherence to FAIR data principles facilitates reproducible outcomes, making collaboration across teams and industries smoother and more effective.
By working together, industry and academia can take a unified approach to ensure data is carefully curated, standardized, and relevant. Ultimately, this will drive faster and more effective patient outcomes while guaranteeing that the data remains adaptable and, as far as possible, future-proof. In doing so, we can enable advanced analytical methodologies to reach their full potential and drive progression in healthcare and beyond.
To fully maximize the potential of advanced techniques in healthcare – such as AI, machine learning, and other analytical methods – meticulous data curation and management are essential3. Clean, comprehensive, and well-structured data provides the solid foundation upon which these advanced techniques can build transformative innovations in the healthcare industry.
Adherence to the FAIR data principles—making data findable, accessible, interoperable, and reusable—is a key component in achieving this. By ensuring data meets these standards, it becomes easier to organize, integrate, and reuse it across various AI-driven applications. Importantly, these principles foster a culture of transparency, innovation, and collaboration, further strengthening the integrity of insights derived from advanced analytical techniques, including AI.
The importance of creating a robust repository of coherent, standardized data extends beyond just this sector. Effective data housekeeping is essential for enabling meaningful collaboration between industry and academia. Cutting-edge research facilities, coupled with the practical clinical development used in industry, leverage the strengths of both sectors. By working together, they can establish standardized data practices, share valuable insights – from pre-clinical research through to development – and construct robust frameworks for data management that align with FAIR principles.
This collaboration ensures that data is not only well-organized and of high quality but also readily accessible and reusable, enhancing its relevance and applicability to real-world challenges.
Sun, D., Gao, W., Hu, H. and Zhou, S. (2022). Why 90% of clinical drug development fails and how to improve it? Acta Pharmaceutica Sinica B, [online] 12(7). Available at: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9293739/.
Tala Talaei Khoei and Singh, A. (2024). Data reduction in big data: a survey of methods, challenges and future directions. International journal of data science and analytics. doi:https://doi.org/10.1007/s41060-024-00603-z.
Wise, J., de Barron, A.G., Splendiani, A., Balali-Mood, B., Vasant, D., Little, E., Mellino, G., Harrow, I., Smith, I., Taubert, J., van Bochove, K., Romacker, M., Walgemoed, P., Jimenez, R.C., Winnenburg, R., Plasterer, T., Gupta, V. and Hedley, V. (2019). Implementation and relevance of FAIR data principles in biopharmaceutical R&D. Drug Discovery Today, [online] 24(4), pp.933–938. doi:https://doi.org/10.1016/j.drudis.2019.01.008.
Lake, F. (2019). Artificial intelligence in drug discovery: what is new, and what is next? Future Drug Discovery, 1(2), p.FDD19. doi:https://doi.org/10.4155/fdd-2019-0025.
Dr Soroosh Afyouni is the Head of Health Data Sciences at bioXcelerate AI. Prior to completing his PhD in Statistical Neuroimaging at the University of Warwick in 2017, where he specialized in statistical network and time series analysis, Soroosh received a Master of Engineering from the University of Birmingham in 2012.
From 2017 to 2020, Soroosh continued his research as a (junior and senior) postdoctoral researcher by joining the Big Data Institute at the University of Oxford, where he developed time series models for accurate estimates of human brain activities in large-scale datasets such as the UK Biobank. During his time at Oxford, Soroosh received a Merit Award from the Internal Organization for Human Brain Mapping. In 2021, Soroosh joined the University of Cambridge’s Department of Psychology and Faculty of Mathematics to focus on developing ML methods for early diagnosis of Alzheimer’s disease.
In addition to his academic background, Soroosh has spent close to two years at a US-based management consulting firm, where he worked with the biggest pharmaceutical companies to address their R&D and commercial strategic challenges ranging from utilizing electronic health records in clinical trials to designing and pressure testing newly designed R&D operating models. Soroosh joined bioXcelerate in 2023, where he will be working on the application of statistical and machine learning methods in precision medicine.
Disclaimer: This article has not been subjected to peer review and is presented as the personal views of a qualified expert in the subject in accordance with the general terms and conditions of use of the News-Medical.Net website.