Top GenAI Applications for Healthcare Innovation

OpenAI's work on synthetic voice generation could be immensely beneficial for patients. With this technology, my brilliant friend, Dr. Rohaid Ali, colleagues, and the Brown Neurosurgery Department have utilized OpenAI's voice engine to help a patient recover her voice after she lost her speech due to a vascular brain tumor. She was able to use a 15 second clip from a video for a class project to be the reference audio source for her synthetically reconstructed voice. For patients with neurodegenerative diseases or other chronic or as well as acute conditions that impact speech such as traumatic brain injury or stroke, this technology could help recover their voices using a simple voicemail they've left on a friend or family member's phone. A tremendous example of an application of AI in medicine that delivers real patient value and impact. This project also addresses what I've described as The Faster Horse Problem in AI (https://lnkd.in/eVv9WXqY). Instead of building a co-pilot, Dr. Ali and team are unlocking an entirely new class of tools for patients. With each new AI model and capability, we must work tirelessly on behalf of our patients to identify applications with such profound impact. Brown Neurosurgery Lifespan The Warren Alpert Medical School of Brown University #ai #artificialintelligence #generativeai #healthcare #medicine #neurosurgery https://lnkd.in/eB7hwwgd

🚀 Congratulations Louis Blankemeier Akshay Chaudhari Curtis Langlotz on their groundbreaking work with “Merlin,” a state-of-the-art vision language foundation model for 3D CT. Stanford University 🩻✨ Harnessing the power of: 🔍 1.8+ million structured EHR codes 📜 6+ million natural language radiology report tokens 📝 Paper: https://lnkd.in/gMVsa9ym With over 85 million CT scans annually in the US and a growing shortage of radiologists, AI-driven vision language models (VLMs) like Merlin are pivotal. Unlike current medical VLMs, which are limited to 2D images and short reports, Merlin leverages comprehensive EHR data to deliver exceptional results across various tasks: 🔬 Top 3 Takeaways: 1. Unprecedented Performance in Zero-shot Classification: • Outperforms OpenCLIP and BioMedCLIP in classifying 30 findings using both internal (5,137 CTs) and external (7,000 CTs) datasets. • Superior fracture detection on the public VerSe dataset. 2. Comprehensive Patient Phenotype Classification: • Evaluates 692 patient phenotypes with best performance from inflated 3D ResNet architecture. • Introduces 3D counterfactual generation for visual model prediction explanations. 3. Enhanced Disease Prediction and Radiology Reporting: • Predicts 5-year disease outcomes across 6 chronic diseases, demonstrating significant benefits with less training data. • Outperforms RadFM in generating radiology reports from 13 anatomical sections. Merlin’s advancements are set to revolutionize medical imaging and healthcare. HOPPR #GenAI #MedicalImaging #AI #HealthcareInnovation

Healthcare systems are slowly developing generative AI use cases. This great STAT resource tracks publicly disclosed activity. Not surprisingly, ambient clinical note documentation is the most common application. Several systems (including mine) are harnessing generative AI to draft responses to patient messages. A few are using generative AI to identify missed diagnoses and care gaps and predict inpatient length of stays and mortality risk. I see considerable opportunities in two areas. 1. Summarizing patient records and referrals. Did you know physicians spend one-third of their EHR time on chart review? (doi:10.7326/M18-3684). 2. Helping clinicians better manage their In Boxes, which is a job in itself. (e.g., doi:10.1007/s11606-022-07766-0). As Benedict Evans recently wrote, "When we get a new tool, we start by forcing it to fit our existing ways of working, and then over time, we change the work to fit the new tool." So, how we use generative AI (and how we work) will undoubtedly shift over time. #healthcareai #healthcareinnovation #healthcareonlinkedin https://lnkd.in/gQkPkq9E

Revolutionizing Medical Diagnostics with the Power of Generative AI The future of healthcare is here, and it's powered by cutting-edge technology like generative AI. In the realm of medical imaging, this revolutionary tool is enabling advancements we could only dream of a few years ago. Imagine, instead of relying on limited datasets of real patient scans, we could synthesize high-quality medical images on demand. This is exactly what generative AI models are accomplishing, providing: Vastly Expanded Training Data: By generating realistic synthetic images of various diseases, anatomical structures, and pathologies, we can train image-recognition algorithms on a much richer and diverse dataset. This leads to: * More accurate diagnoses: Improved algorithms can detect subtle abnormalities and patterns with greater precision, potentially saving lives by catching diseases earlier. * Reduced reliance on real patient data: Synthetic images can help address privacy concerns and ethical considerations surrounding sensitive medical data. * Faster development of new diagnostic tools: The ability to quickly generate specific types of images allows for rapid prototyping and testing of new algorithms, accelerating the path to innovative diagnostic solutions. This isn't just science fiction - generative AI is already making waves in healthcare. Researchers are using it to: * Develop AI models that outperform human radiologists in detecting certain cancers. * Create realistic simulations of medical procedures for training purposes, improving surgeon skills and patient outcomes. * Generate personalized 3D models of organs and tissues to guide surgical planning and treatment. Of course, challenges remain. Ensuring the accuracy and realism of synthetic images, mitigating potential biases in AI models, and navigating ethical considerations are all crucial areas of ongoing research. But the potential of generative AI in healthcare is undeniable. It holds the promise of a future where: * Everyone has access to accurate and timely diagnoses, regardless of location or socioeconomic status. * Medical errors are significantly reduced, thanks to advanced AI-powered diagnostics. * Personalized medicine becomes a reality, with treatments tailored to each individual's unique needs. The future of healthcare is bright, and generative AI is a powerful light guiding the way. Let's embrace this technology responsibly and ethically, working together to ensure it leads to a healthier and more equitable world for all. Join the conversation: 1) What are your thoughts on the potential of generative AI in healthcare? 2) What are some ethical considerations we should be mindful of? 3) How can we ensure equitable access to these groundbreaking technologies? Share your insights and let's continue shaping the future of healthcare together! #GenerativeAI #Healthcare #MedicalImaging #Diagnostics #Innovation #FutureofMedicine

Here are my Top 5 developments in Healthcare Generative AI last month, showcasing a rather rich diversity of use cases. Gen AI has>> 🤖Helped pinpoint brain damage from strokes by analyzing health histories & neuro exams, potentially improving care where access to scans or specialists is limited 🤖Been used by NYU Langone Health to simplify doctors' notes from hospital visits into patient-friendly language, reducing the reading level from the 11th-grade to the 6th-grade level 🤖Been shown to enhance healthcare-associated infection surveillance. Two AI-driven tools can accurately detect cases of central line-associated bloodstream infection and catheter-associated urinary tract infection 🤖Fuelled a NVIDIA and Hippocratic AI collaboration to develop AI nurses that demonstrate empathy and high performance in medical tasks, potentially addressing global healthcare worker shortages 🤖Powered a Chinese trial of a virtual surgical consultant designed to aid neurosurgeons during critical procedures. The chatbot should be capable of providing real time advice by processing various text, audio and image medical inputs 💌Look out for the full AI newsletter coming soon! 👇Links to articles in comments #DigitalHealth #AI