Max Anfilofyev’s Post

Excellent work on medical ⚕ data retrieval in particular situations. Use a cutting-edge method (RAG) combined with sources of dependable, interlinked, peer-reviewed studies. This enables domain-specific semantic search, chunking, and context capture to improve traditional approaches 📅 In this article, Jund Wu et al. explore the intriguing world of medicine. A massive potential to explore. At the Clinic of AI, we have used the same methodology for agriculture and wrapped it with a simple UX chatbot to allow farmers 👩🔧 of different plants ☘ to get the latest access to studies and information in real-time. ℹ This project was made with one of the latest additions to the team: Mert Oğul, and was an impressive feat as one of his first dives into the world of RAG-based/Arxvix interconnected bots. - we might be publishing a workshop on its conceptions, stay tuned. Daniel Küng #agriculture #RAGbot #semanticsearch #Information #chunking #artificialintelligence #medical

MedGraphRAG: Towards Safe Medical LLMs Large Language Models (LLMs) have shown impressive capabilities in various domains, including healthcare. However, when it comes to handling sensitive medical data, generating reliable and evidence-based responses is crucial. Traditional methods often fall short in capturing the full context of medical information, leading to potential inaccuracies and safety concerns. MedGraphRAG introduces a novel graph-based Retrieval-Augmented Generation (RAG) framework tailored for the medical domain. It starts by employing a hybrid static-semantic approach for document chunking, which significantly improves context capture compared to traditional methods. Extracted entities are then used to construct a three-tier hierarchical graph structure, connecting entities to foundational medical knowledge from papers and dictionaries. These entities are further linked to form meta-graphs, which are merged based on semantic similarities to create a comprehensive global graph. To learn more about MedGraphRAG and other AI highlights, check out this week's LLM Watch: https://lnkd.in/dwm3iKc5

OpenAI’s o1-preview large language model (LLM) has demonstrated superior performance in medical reasoning tasks compared to both human physicians and previous AI models. In differential diagnosis generation, o1-preview correctly included the diagnosis in 78.3% of New England Journal of Medicine cases, surpassing GPT-4’s 72.9% accuracy. Additionally, in management reasoning vignettes, o1-preview outperformed both GPT-4 and human physicians by over 40%, highlighting its potential to enhance clinical decision-making. #ArtificialIntelligence #Healthcare

🔊 Researchers from the University of Oxford have introduced MedGraphRAG, designed to improve Large Language Models (LLMs), like GPT-4 in the medical field It addresses one of the biggest challenges LLMs face—contextual accuracy in medicine. MedGraphRAG's approach connects important medical entities, such as diseases and treatments, to reliable knowledge bases. MedGraphRAG outperforms state-of-the-art models across multiple medical Q&A benchmarks. Read more: https://lnkd.in/grFKP2ep ♻️ If you found this content useful, please share it with your friends and connections. 🔔 Follow MindfulEngineer for more AI-related content. https://lnkd.in/gHdY4uiK #AIinHealthcare #MedicalAI #LLM #MachineLearning #ArtificialIntelligence

On 'Hustling the physician-patient relationship', the regulatory cues and emphasis on patient-centred communication were great. I liked the principle that 'Human Supervision of AI is good practise' - it reflects on where we are now. Bias is a golden thread - to add in another paper on racial bias and LLMs, I found this a helpful read too https://lnkd.in/g-wgX85t

Qualitative metrics from the biomedical literature for evaluating large language models in clinical decision-making: a narrative review - Most frequently used LLMs: GPT-3.5, GPT-4, Bard, LLaMa/Alpaca-based models, Bing Chat. - Top criteria for scoring LLM outputs: accuracy, completeness, appropriateness, insight, consistency. - Need for standardized reporting of qualitative evaluation metrics to enhance research on LLMs in healthcare. DOI: https://lnkd.in/e3iikhzJ Link: https://lnkd.in/e2v5VSwp

Qualitative metrics from the biomedical literature for evaluating large language models in clinical decision-making: a narrative review - Most frequently used LLMs include GPT-3.5, GPT-4, Bard, LLaMa/Alpaca-based models, and Bing Chat. - Key evaluation criteria for LLM outputs are accuracy, completeness, appropriateness, insight, and consistency. - There is significant variation in how studies report findings and assess LLM performance; standardized reporting metrics are needed. DOI: https://lnkd.in/e3iikhzJ DOI: doi:10.1186/s12911-024-02757-z

Large Language Models (LLMs) can revolutionize how patients find the right clinical trials and healthcare professionals find the right candidates. This video by NIH National Library of Medicine shows how they applied OpenAI GPT models to find the right clinical trials quickly and efficiently.

Volume 2, No. 2 is now available! Here are the latest articles available in the February issue of NEJM AI:   Save this post to revisit later (click the 💬 button at top right of post).    𝗘𝗱𝗶𝘁𝗼𝗿𝗶𝗮𝗹𝘀  ⚖️  It’s Time to Bench the Medical Exam Benchmark https://nejm.ai/3PK8v1D    𝗣𝗲𝗿𝘀𝗽𝗲𝗰𝘁𝗶𝘃𝗲𝘀  🌍 Using Large Language Models to Promote Health Equity https://nejm.ai/40gbwvn     📝 The Burden of Reviewing LLM-Generated Content https://nejm.ai/3WnJD3r     🌟 AI Grand Rounds: Cultivating Health Care’s AI Future https://nejm.ai/4aurOW2    𝗢𝗿𝗶𝗴𝗶𝗻𝗮𝗹 𝗔𝗿𝘁𝗶𝗰𝗹𝗲𝘀  🗣️ A Cross-Sectional Study of GPT-4–Based Plain Language Translation of Clinical Notes to Improve Patient Comprehension of Disease Course and Management https://nejm.ai/4avSdTw     👁️ Developing ICU Clinical Behavioral Atlas Using Ambient Intelligence and Computer Vision https://nejm.ai/4g8D23N     🔬 Multicenter Double-Blind Study Evaluating AI-Driven Detection of Proximal Deep Vein Thrombosis https://nejm.ai/4awOqFj    𝗣𝗼𝗹𝗶𝗰𝘆 𝗖𝗼𝗿𝗻𝗲𝗿  🤖 Disclosure, Humanizing, and Contextual Vulnerability of Generative AI Chatbots https://nejm.ai/40rPlSY    𝗥𝗲𝘃𝗶𝗲𝘄 𝗔𝗿𝘁𝗶𝗰𝗹𝗲  ⚠️ Not All Clinical AI Monitoring Systems Are Created Equal: Review and Recommendations https://nejm.ai/40J53KJ    Visit http://ai.nejm.org to read all the latest articles on AI and machine learning in clinical medicine.    #ArtificialIntelligence #AIinMedicine 

This is quite intriguing. Following emerging evidence in recent months that some large language models (LLMs) are beating experienced physicians in the quality of diagnoses, a paper authored by medical faculties claims to have detected cognitive impairment in LLMs as they age - a very human-like degradation of AI 'thinking' ability. Would be very interesting to see if this finding can be replicated.