Changing how doctors diagnose diseases with AI | Microsoft Azure and NVIDIA

Key Concepts

• Undertreatment/Undiagnosed Patients: A significant portion of patients, particularly in women's health and conditions like COPD, remain undiagnosed or undertreated despite available data.
• PALLUX: Pangaea Data's AI platform designed to identify undiagnosed and miscoded patients by analyzing patient data against clinical guidelines.
• Clinical Guidelines: Standardized recommendations for diagnosing and treating specific conditions, used by PALLUX to assess patient data.
• AI Inference: The process by which PALLUX interprets patient data in relation to clinical guidelines to provide real-time insights to clinicians.
• Guideline Concordant Therapy: Treatment that aligns with established clinical guidelines for a specific condition.
• Sovereign AI: The approach of bringing personalized and democratized care across the globe.
• NVIDIA FLARE: A federated learning approach that helps train centralized models while preserving patient privacy without moving the data.
• Large Language Models (LLMs): Advanced AI models, powered by NVIDIA GPUs and available through Microsoft Azure, used by PALLUX for complex reasoning similar to clinicians.
• Cancer Cachexia: A wasting disease directly responsible for at least 20% of cancer deaths, targeted by PALLUX for early diagnosis.
• Data Silos: The problem of valuable patient data being locked up in health records, hindering comprehensive analysis and treatment.

The Problem: Gaps in Healthcare and the Power of Data

The healthcare system faces a critical challenge: a significant number of patients remain undiagnosed or undertreated, even when the data needed for diagnosis and treatment already exists. Sarah's experience with endometriosis, taking 20 years to get diagnosed, exemplifies this issue. The video highlights that for every woman diagnosed with a health condition, four go undiagnosed. This problem stems from the overwhelming amount of data that clinicians must process, exceeding human capacity. Doctors are limited by time constraints (e.g., 10-minute appointments) and the need to simultaneously listen to patients, review electronic health records, and recall relevant medical knowledge.

Pangaea Data's Solution: PALLUX and AI-Powered Diagnosis

Pangaea Data aims to address this problem with its AI platform, PALLUX. The goal is to mimic a physician's diagnostic process by imbibing medical knowledge and patient interaction aspects into the technology. PALLUX works by:

1. Data Capture and Retrieval: Gathering patient data from electronic health records and other sources.
2. Matching Data with Clinical Criteria: Comparing the captured data against clinical guidelines for specific conditions.
3. AI Inference: Interpreting the patient data in relation to the clinical guidelines to identify potential diagnoses or treatment gaps.
4. Real-Time Feedback: Providing clinicians with real-time insights to inform their decision-making.

PALLUX is designed to be compatible with various healthcare systems and interfaces, facilitating seamless integration into existing workflows.

Real-World Applications and Impact

• COPD: PALLUX has been shown to identify five times more COPD patients compared to traditional methods, leading to faster treatment and improved patient outcomes. It is stated that over half of patients with COPD remain undiagnosed or undertreated.
• Cancer Cachexia: In the UK, PALLUX was deployed within the NHS to discover undiagnosed and miscoded cancer cachexia patients. It found six times the patients compared to traditional approaches. Early diagnosis can cut cachexia treatment costs by as much as 50% and potentially save as much as 1 billion pounds per year.

These examples demonstrate the potential of AI to improve diagnosis rates, reduce healthcare costs, and enhance patient quality of life.

Technology and Partnerships

Pangaea Data leverages advanced AI technologies, including:

• Foundation Models/Generative Models: Powerful AI models that enable complex reasoning and data analysis.
• NVIDIA GPUs: Used to power the AI models and accelerate data processing.
• Microsoft Azure: Provides a compliant and secure cloud environment for deploying and scaling PALLUX.
• NVIDIA FLARE: A federated learning approach that helps train centralized models while preserving patient privacy without moving the data.

The collaboration with Microsoft and NVIDIA is crucial for accessing cutting-edge AI technologies and ensuring compliance with data privacy regulations.

The Human Element and Future of Healthcare

The video emphasizes that AI is not intended to replace clinicians but to augment their capabilities. By automating mundane tasks and providing data-driven insights, AI can free up clinicians to focus on patient interaction, empathy, and personalized care. The ultimate goal is to create an "AI-powered clinician" who is better equipped to understand and treat patients.

The speaker shares a personal anecdote about a dinner conversation where someone expressed gratitude for Pangaea Data's work, highlighting the profound impact of timely diagnosis and treatment on patients and their families.

Conclusion

Pangaea Data's PALLUX platform represents a significant step towards addressing the problem of undertreatment and undiagnosed patients in healthcare. By leveraging AI and partnerships with technology leaders like Microsoft and NVIDIA, PALLUX has the potential to unlock the vast amount of data stored in health records, improve diagnostic accuracy, and ultimately enhance patient outcomes. The focus remains on empowering clinicians with better tools and information, allowing them to provide more personalized and empathetic care.