AI has been used in medicine for many years, especially in diagnosing mild diseases and improving access to medicine.
Now Google has announced some areas that will improve even more with new developments.
Research Med-PaLM 2
Last year, Med-PaLM, a doctor-focused version of PaLM, was introduced. This tool was the first to pass similar questions on the US Medical Exam, solving not only complex questions, but also open questions, and providing justifications and evaluations of individual answers.
Recently, Med-PaLM 2 was introduced, which significantly improved its performance with a rating of 85%. While this is an exciting development, it is important to note that there is still much work to be done to ensure that this technology is safe, fair and scalable. These models were evaluated under 14 criteria, and some limitations were found in answering medical questions and meeting product excellence standards.
New partners for AI-powered ultrasound
Ultrasound is an essential tool for identifying important information such as gestational age in pregnant women and early detection of breast cancer. However, these devices often require experts with years of experience to perform the tests and interpret the images, which can be a problem in areas with a lack of specialized ultrasounds.
To help bridge the gap, AI models are being built that can simplify the acquisition and interpretation of ultrasound images. For this purpose, two new strategic partnerships have been established.
– The first partnership was established with Jacaranda Health, a Kenya-based non-profit organization that is engaged in improving health outcomes for mothers and babies in public hospitals. Through this partnership, digital solutions will be researched to facilitate the solution of ultrasound scans in Kenya.
– A second partnership was established with Chang Gung Memorial Hospital (CGMH) in Taiwan to explore the use of ultrasound in breast cancer screening. He is investigating whether AI models can help in the early detection of breast cancer using ultrasound.
AI for cancer treatment planning with the Mayo Clinic
For the past three years, they have collaborated with the Mayo Clinic to explore how AI can help in the planning process for radiation therapy, a common treatment for more than half of all cancers in the US. Radiation therapy planning is laborious and time-consuming. especially delineation, where professionals draw lines on CT images to separate parts of the cancer from nearby healthy areas that may be damaged during treatment. This process can take up to 7 hours for one patient.
The findings and the radiation treatment model developed in collaboration with the Mayo Clinic will soon be published. In addition, an agreement is being formed to explore future research, model development, and joint commercialization of this technology.
Detection of tuberculosis with AI
Tuberculosis is one of the leading causes of death worldwide, and while it is being treated, cost-effective screening solutions are needed to detect the disease early and disseminate it in communities. To address this question, work was undertaken to implement artificial intelligence TB screening solutions in sub-Saharan Africa.
In collaboration with the non-profit organization Right to Care, 100,000 AI-assisted TB screenings will be performed in the country. This initiative will help to improve the detection and treatment of the disease, reducing its spread.
As you can see, AI has great potential to improve healthcare around the world, although it is important to note that AI in medicine must be performed in a safe and fair way to ensure patient well-being and accurate diagnostics. Collaboration between researchers, non-profit organizations and hospitals is critical to the development and effective implementation of these technologies.
