Regular readers of these weekly blogs will know that in 2023 I have covered numerous stories on artificial intelligence (AI). These have included the WHO and AI regulation, the AuntMinnie awards, the FDA and AI applied to medical images, a well-executed study on AI and mammography, a new EU law on AI, and point-of-care ultrasound joining forces with AI. The American Association for the Advancement of Science has just published a 20-page booklet entitled Frontiers of Medical Research: Artificial Intelligence that provides an excellent summary of the field (click here to download a free copy).
The first article introduces common AI terminology, then describes where we are going, speaks about regulating AI, and concludes with a plea that, despite the existential threat of AI, there is a moral imperative to advance the application of AI in healthcare. The second article covers artificial neural networks and deep learning, describing supervised and unsupervised learning, segmentation of medical images, cancer detection in pathology images, large language models, and foundation models. The diagram below left (© J Gregory) illustrates how, in the future, multi-modal foundation models could be applied.
The third article emphasizes how omics – the sum of constituents within a cell – combined with AI is bringing the bench closer to the bedside, while the fourth article considers how AI can advance cancer research. This includes prevention and early detection (particularly relevant for breast cancer), immunotherapy, and targeted protein therapy, illustrating what’s possible with a specific disease example – multiple myeloma – in an era of personalized medicine. Article number five describes how AI is transforming neuroscience, providing insights into the healthy brain, neurological and psychiatric diseases, and improved therapies.
Article six presents a case study for delivering value to patients and clinicians alike by focusing on AI at the bedside, showing how an image of a patient’s retina can be used to predict cardiovascular risk factors, and to identify cognitive impairment.
Article seven, entitled “An AI doctor on your wrist: wearable devices to monitor health and disease,” is illustrated in the diagram at right (© J Gregory). As can be seen, various “smart” sensors measure a range of physiological parameters and, when combined with AI, provide information on mental health, disease management, treatment, and wellness status. Finally, in the eighth article, a health systems perspective is advanced, with a view to enabling and accelerating the implementation of AI. There are four accelerators – governance, AI platforms, advanced datasets, and an innovation-driven culture – proposed by the Icahn School of Medicine at Mount Sinai in New York City, whose academic staff produced this useful booklet.