Spotlight #16: 🧐 Sora in medicine, 🚑 Emergency medicine AI, 🩻 Chatbots for orthopaedic information, 🧩 AI-powered protein mapping tool
🧐 OpenAI’s Sora challenges the art of medicine
News
OpenAI has recently unveiled its latest innovation, Sora, a text-to-video model that is able to generate quality videos based on simple text prompts. Sora's adaptable AI framework enables it to extend its intelligence across diverse sectors, including healthcare, finance, education, and entertainment. This versatility empowers Sora to deliver solutions that are informed, accurate, and highly relevant to each field. In healthcare, Sora could assist in diagnosing diseases and suggesting treatments, potentially saving lives with its speed and accuracy.
This article points out some challenges that Sora might bring to healthcare. The author concerned about the essence of creativity in medicine as algorithms produce new content such as suggested diagnoses from previously learned data. “Our diagnoses and treatments are fueled by our experiences, perspectives and creativity—not just bare clinical data.” Clinicians are also concerned about the potential loss of individuality and emotion in patient care.
🚑 Keynote: AI announcements, patient engagement, and the future of emergency medicine with Kevin Maloy
Podcast
Dr. Kevin Maloy, an assistant professor of innovatioin at Georgetown University School of Medicine, talks about the potential of conversational technologies to redefine patient engagement and the traditional call center model. They discuss various applications of AI, such as using GPT for synthesizing patient data, automating tasks, and summarizing medical transcripts. They also explore the potential impact of AI on healthcare documentation, patient engagement, and medical research. The conversation also touches the challenges of implementing AI applications in the field, such as data privacy, transparency, and the need for context-aware AI systems.
Under the context of emergency medicine, they highlighted two unique challenges. One is to improve documentation efficiency to assist physicians in quickly documenting patient encounters. This includes features like auto-completion that would streamline the documentation process, particularly in fast-paced environment like the emergency room. Another challenge is to improve computer hearing and listening. There is a desire for advanced noise cancellation and speech recognition. Such function will allow clearer communication between healthcare professionals and patients.
🩻 Studies show AI chatbots provide inconsistent accuracy for musculoskeletal health information
News
Three studies presented at the 2024 AAOS meeting assessed the accuracy of orthopaedic information provided by three large language model (LLM) chatbots OpenAI ChatGPT 4.0, Google Bard, and BingAI. These studies found that while the chatbots offer concise summaries of orthopaedic conditions, their accuracy varies depending on the category.
One study asked these three chatbots to explain basic orthopaedic concepts, integrate clinical information and address patient queries. It was found that chatbots generally struggled with clinical management suggestions, often deviating from standard care practices. Another study found that ChatGPT's accuracy significantly improved when prompted to answer questions “as an orthopaedic surgeon”. The third study evaluated ChatGPT’s ability to provide information on the Latarjet procedure, finding that it consistently derived information from academic sources but also included non-academic resources.
Overall, researchers concluded that orthopaedic surgeons remain the most reliable source of information. Moreover, ChatGPT is not yet an adequate resource to answer patient questions and further work is needed to improve the accuracy in orthopaedic field.
🧩 New AI-powered protein mapping tool optimizes cancer therapy
News
It is difficult to identify the right drug treatment regime for each patient in cancers such as clear cell renal cell carcinoma (ccRCC) as responses to existing treatments are different for them. For instance, Hypoxia-Induced Factor Alpha (HIF2α), a key target of ccRCC that is blocked by Belzutifan, has shown different responses to the aggressiveness of the tumor. HIF2α was less active when there were greater levels of the protein present. Scientists have recently developed an AI tool called FuncOmap that maps the function of proteins in tumors. This tool helps clinicians determine precise treatment strategies by visualizing directly how proteins interact within the tumor. It also enables personalized medicine by allowing clinicians to predict individual responses to drugs and tailor treatments accordingly. The interdisciplinary team behind the tool includes biophysicists, biologists, and computational scientists. They are collaborating with Stanford University School of Medicine to further develop and optimize FuncOmap for clinical use, aiming to improve cancer treatment outcomes through personalized medicine.
🥳 I hope you enjoyed the findings!
Do you think the release of Rosa will influence the art of medicine? How do you think of AI implementations in precision medicine?
Feel free to leave your comments and suggestions here.