At Google’s annual health event, The Check Up, Google shared health AI updates regarding their progress on their medical large language model (LLM) research, the ways that AI can assist with disease detection, and the partnerships that are bringing solutions into real-world settings.
The focus continues to grow around LLMs, which are tools designed to perform a variety of natural language processing tasks, including summarising, predicating and generating text.
Med-PaLM (Google’s medical LLM) obtained a passing score on U.S. medical licensing-style questions, and showed the ability to provide rationale and evaluate its own responses.
Google’s next iteration, Med-PaLM 2, surpassed its predecessor, having performed at a level of an expert doctor on medical questions, scoring 85%.
Although the model reportedly tested against 14 criteria — including scientific factuality, precision, medical consensus, reasoning, bias and harm — and was evaluated by clinicians and non-clinicians from a range of backgrounds and countries, there are still improvements to be made to Med-PaLM 2 before it can apply in real-world settings.
Google has announced new partnerships for enhancing the ways AI can help detect diseases. They are partnering with Jacaranda health, a Kenya-based nonprofit, to help them with their approaches to ultrasound delivery. They will explore the role of AI-assisted ultrasound to improve image interpretation and improve health outcomes for mothers and babies.
They will be partnering with Chang Gung Memorial Hospital in Taiwan to explore whether their AI models can help with early detection of breast cancer using ultrasound.
Additionally, they will be finalising their agreement with Mayo Clinic to explore how AI can support the time-consuming process of planning for radiotherapy.
Finally, Google will be working with an AI-based organization headed by Right to Care to ensure AI-powered, chest X-ray screenings are more widely available across Sub-Saharan Africa.
Overall, through the support of AI tools and new partnerships, Google is helping to bring high-quality care to communities that require it.
Source: Google Blog
