AI is playing an increasingly important role in medicine, though quite often this does not happen as intended. The tech is just as likely to lead to breakthroughs and discoveries in unintended or serendipitous ways, points out AI and ML expert Moein Shariatnia.

New AI tools and models are being developed every day, though many of them feature a great deal of uncertainty. Quantifying and detecting this uncertainty goes a long way toward ensuring that AI is more reliable and trustworthy, argues AI and ML expert Moein Shariatnia.

Success stories abound about the adoption of AI in healthcare and other domains, yet failures may prove even more instructive as governments, companies, and persons seek to navigate the age of AI. AI and ML expert Moein Shariatnia shines the spotlight on one such cautionary tale.

In this episode, we discuss AI models reshaping diabetic retinopathy diagnosis and management, addressing challenges with enhanced accuracy. Real-world studies validate their effectiveness, often outperforming manual grading. Tune in to listen why the convergence of medical and machine learning expertise becomes increasingly crucial for future healthcare professionals.

The dynamic landscape of AI in healthcare has witnessed a remarkable shift from single-task models to the emergence of powerful multimodal deep learning models. Tune in as we navigate the intersection of technology and healthcare, highlighting both the promises and pitfalls in the journey toward a more efficient and patient-centric medical future.

‘Hallucination,’ which refers to LLMs and other AIs presenting falsehoods as seemingly plausible facts, is a grave problem and one yet to be resolved. This becomes especially urgent when patients’ lives are at stake, although hope is at hand, writes ML developer Moein Shariatnia.

Deep learning models have made great strides since the time when all data samples had to be labeled with ground truths and input manually for such models to execute tasks. They now offer ever greater capabilities and reliability of prognoses and thus much instruction for experts.

The ever swifter, wider use of AI in healthcare research makes for many mistakes along the way and often yields inherently flawed research, explains ML developer and medical student Moein Shariatnia, who devises computer vision and NLP applications in healthcare with DL models.

Unlike most AI, which focuses on and is trained for handling just one specific modality, multimodal AI models can handle different modalities such as image, text, and speech at the same time, in addition to narrow and specific tasks. This means that, if and when multi-modality AI progresses to the point where it can deploy on a vast scale, it will truly revolutionize medicine and many other fields.

Generative models are hot items in tech news today. Stable Diffusion and Midjourney - which create amazing artwork - and ChatGPT, which writes excellent essays for an input prompt - have astounded many. Similar models are also in the works for healthcare - especially for radiology - to address their real-world challenges. ML developer and medical student Moein Shariatnia sketches a brief history of deep generative models and explains how they work and their use in medical imaging.

Google has some of the most formidable AI capabilities in the world, though among their applications diabetic retinopathy is not the first that springs to mind. As diabetics are likelier to develop eye-related disorders and their numbers continue to grow worldwide, this is extremely valuable.