🚨 Pulse of Innovation 🚨

Breaking news in the healthcare AI

NVIDIA’s New Processor supporting HINDI large language model (LLM) to boost Health AI

Indian life sciences and healthcare organizations are at the forefront of leveraging generative AI to revolutionize healthcare delivery and research. Powered by NVIDIA NIM microservices, these organizations are making significant strides in various areas:

Neuroscience Research:

• IIT Madras' Brain Centre is pioneering AI-driven analysis of massive brain datasets, enabling a deeper understanding of brain structure and function. This research has the potential to unlock groundbreaking discoveries in neuroscience and neurodegenerative diseases. Additionally, the center is developing an AI chatbot using the Nemotron-4 Hindi NIM microservice to make complex neuroscience information accessible to a wider audience, including STEM students and researchers.

Combatting Antimicrobial Resistance:

• IIIT-Delhi is leading the charge in addressing the global health threat of antimicrobial resistance. By utilizing AI-powered data integration and predictive analytics, the institute's AMRSense tool aims to improve accuracy and speed in identifying and mitigating antimicrobial resistance. This innovative solution is designed to be implemented in both hospital and community settings, contributing to a more effective and proactive approach to combating this critical issue.

AI-Powered Medical Imaging:

• 5C Network's Bionic suite of medical imaging tools is transforming radiology reporting by automating the analysis of medical scans and generating comprehensive reports. This AI-driven technology empowers clinicians with actionable insights, leading to more accurate diagnoses and timely treatment decisions.

Drug Discovery and Analysis:

• Innoplexus, an AI-powered life sciences platform, is accelerating drug discovery and analysis through the use of NVIDIA NIM microservices. By leveraging AI, Innoplexus can analyze vast amounts of data, predict protein-protein interactions, and perform virtual drug screenings at unprecedented speeds. Additionally, the platform is being used to analyze Ayurvedic medicine, a traditional Indian healing system, and generate easy-to-understand explanations of medical prescriptions and reports.

These advancements are fueled by the Indian government's significant investments in foundational AI models and the rapid growth of the Indian healthcare market. By harnessing the power of AI in regional language NVIDIA poised impact lives millions of people in lower socio-economy countries where English is not the primary language.

🧑🏼‍🔬 Bench to Bedside👨🏽‍🔬

Developments in healthcare AI research and innovations

Development and Validation of a Quantitative Coronary CT Angiography Model for Diagnosis of Vessel-Specific Coronary Ischemia

• Artificial intelligence (AI) is poised to bring a new level of precision to coronary artery disease (CAD) diagnosis and management.

• Recent studies validated an AI-driven model, AI-QCTISCHEMIA, which uses next-generation quantitative coronary CTA to assess stenosis, atherosclerosis, and vessel morphology, showing potential to enhance diagnostic accuracy and predict cardiovascular risk more effectively than conventional imaging.

Study Design and Statistical Approach

• The AI-QCTISCHEMIA model was developed using data from PACIFIC-1 and CREDENCE, with advanced AI algorithms analyzing a broad range of parameters, including plaque volume, lesion length, and stenosis severity.

• Through random forest modeling, researchers identified patients at high risk of ischemia by integrating plaque diffuseness and other coronary characteristics.

• Performance analysis showed AI-QCTISCHEMIA had a high diagnostic accuracy (AUC ~0.90), surpassing standard techniques such as SPECT and coronary CTA-based fractional flow reserve (FFRCT).

• Sensitivity, specificity, and AUC metrics confirmed AI-QCTISCHEMIA’s superior ability to diagnose ischemia compared to conventional methods, except for the PET scan, where it performed similarly.

Key Results

• The AI-QCTISCHEMIA model demonstrated remarkable efficacy in predicting per-patient ischemia, with an AUC close to 0.90.

• In PACIFIC-1, patients flagged as ischemia-positive had a sevenfold higher risk of major adverse cardiovascular events (MACE) over eight years. This long-term prognostic utility extends beyond traditional CTA assessments, where stenosis alone often overlooks the impact of diffuse or nonobstructive plaques.

• Additionally, the AI-QCTISCHEMIA model provides incremental prognostic value by combining plaque diffuseness and lesion length into its analysis.

• Compared to FFRCT, the AI-QCTISCHEMIA model had a lower vessel rejection rate (3.6% vs. 15.5%), significantly reducing the need for further diagnostic tests.

Limitations

• Performance metrics in the CREDENCE study were slightly lower than in PACIFIC-1, potentially due to a higher burden of CAD in CREDENCE participants and differences in FFR measurement techniques.

• In PACIFIC-1, FFR was measured distally, while CREDENCE recorded measurements proximal to the site of maximal stenosis.

• PACIFIC-1’s small sample size and limited low-density plaque volume constrained its outcome analysis.

Conclusion

• AI-QCTISCHEMIA’s comprehensive approach could soon redefine CAD diagnostics by integrating anatomical assessment, ischemia detection, and prognostication into a single model and may guide management decisions.

 JACC Article 

Model

• A simple representation of an aspect of the real world. It is a programme that has been trained on a set of data to recognise certain patterns or make certain decisions without further human intervention.

Deep learning model

• A form of machine learning that employs artificial neural networks, inspired by the human brain, to learn from vast amounts of data (including labelled and unlabelled, structured and unstructured data).

• These networks enable the digital systems to learn and make decisions automatically and independently without human intervention. 

• These models are increasingly used in areas such as pathology, radiology, and genomics.

Foundation model

• A machine learning model trained on a vast amount of data so that it can be easily adapted for a wide range of applications.

• A common type of foundation model is large language models, which power chatbots such as ChatGPT.

Human-in-the-loop

• A system comprising a human and an AI component, in which the human can intervene in some significant way, such as by training, tuning or testing the system’s algorithm, so that it produces more useful results.

• It is a way of combining human and machine intelligence, helping to make up for the shortcomings of both.

Large language model (LLM)

• A machine learning model capable of performing various natural language processing tasks. These tasks include generating and classifying texts and images, answering questions conversationally, translating between languages, predicting, and summarising content. It uses deep learning algorithms and a vast dataset to achieve these capabilities. 

• In healthcare, these models assist in clinical decision support and patient interaction.

Multimodal model

• It processes and combines different types of data, such as images, videos and text, to make more accurate determinations, draw insightful conclusions, or make precise predictions about real-world problems.

• Multimodal models can include data from an electronic health record, an image captured by an X-ray and a radiologists written description of an X-ray to derive conclusions around diagnoses

FDA Perspective on Regulating AI in Healthcare and Biomedicine

The FDA has a long history of regulating AI-enabled medical devices, with the first approval dating back to 1995. While the FDA has made strides in adapting to the rapid evolution of AI, several key challenges and opportunities remain.

Key Challenges and Opportunities:

• Keeping up with the pace of change: AI is evolving rapidly, and the FDA must adapt its regulatory framework to keep pace. This includes developing flexible, risk-based approaches that can accommodate the diverse range of AI applications.

• Flexible approaches across the spectrum of AI models: AI models vary in complexity, from simple algorithms to sophisticated machine learning models. The FDA needs to adopt a risk-based approach, tailoring regulations to the specific risks associated with each model.

• AI in medical product development: AI has the potential to revolutionize drug discovery, clinical trial design, and post-market surveillance. The FDA must ensure that AI-powered innovations are safe and effective while fostering innovation.

• Large language models and generative AI: These emerging technologies present unique challenges due to their potential for unforeseen consequences. The FDA needs to develop strategies to assess and regulate these technologies while mitigating risks.

• AI life cycle management: AI models are dynamic and can evolve over time. The FDA must establish robust mechanisms for monitoring and evaluating AI performance throughout its life cycle.

• Supply chain considerations: AI can play a crucial role in managing supply chains, but it is also vulnerable to disruptions. The FDA must consider the impact of AI on supply chain resilience and security.

• Balancing financial returns and health outcomes: AI has the potential to improve patient outcomes, but it can also be used to optimize financial returns. The FDA must ensure that AI is used to prioritize patient well-being.

Conclusion:

Strong oversight by the FDA protects the long-term success of industries by focusing on evaluation to advance regulated technologies that improve health. The FDA will continue to play a central role in ensuring safe, effective, and trustworthy AI tools to improve the lives of patients and clinicians alike. However, all involved entities will need to attend to AI with the rigor this transformative technology merits.

JAMA Article