LLM Identifies Stroke Innovation Patent Trends
Marcus et al. (2026) analyze patent and publication records from 1993–2023 to quantify innovation in stroke care, using a 13-billion-parameter Llama LLM to filter patents. The LLM achieved 99.2% accuracy (96.5% sensitivity, 99.6% specificity), reducing 237,035 patents to 28,225 stroke-related filings grouped into seven innovation clusters. Pharmacological treatments dominated earlier patents but plateaued, while AI methods, rehabilitation devices, and imaging show rapid patent growth.
Key Points
- 1Identifies 28,225 stroke-related patents (1993–2023) after LLM filtering in this study
- 2Demonstrates LLM classifier accuracy 99.2% with 96.5% sensitivity and 99.6% specificity
- 3Highlights rapid patent growth in AI, rehabilitation, and imaging, guiding research and investment priorities
Scoring Rationale
Strong methodology and peer-reviewed evidence supports high practical relevance, but scope limited to stroke-specific innovation.
Sources
Public references used for this report.
Practice with real Logistics & Shipping data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Logistics & Shipping problems
