Failed clinical trials costs $5–25M per study, with current AI prediction accuracy capped at 67%. In Sii we have ellaborated an AI-powered early warning system that combines four specialized BERT models with novel text analysis methods to flag high-risk trials before launch. The system analyzes design-phase documents, detecting subtle indicators such as enrollment feasibility, regulatory complexity, and sponsor confidence. Tested across diverse disease areas, it outperforms previous systems, offering potential savings of $125M in evaluated portfolios. Predictions are interpretable, enabling proactive adjustments and shifting failure management from reactive to preventive. During the talk you will learn more about the challange, the solution and potential benefits.
2025 - Piotr Ćwikliński - Fail Fast, Save Millions: AI for Smarter Clinical Trials
Failed clinical trials costs $5–25M per study, with current AI prediction accuracy capped at 67%. In Sii we have ellaborated an AI-powered early warning system that combines four specialized BERT models with novel text analysis methods to flag high-risk trials before launch. The system analyzes design-phase documents, detecting subtle indicators such as enrollment feasibility, regulatory complexity, and sponsor confidence. Tested across diverse disease areas, it outperforms previous systems, offering potential savings of $125M in evaluated portfolios. Predictions are interpretable, enabling proactive adjustments and shifting failure management from reactive to preventive. During the talk you will learn more about the challange, the solution and potential benefits.
Powiązane
2025 - Ros Apostol - Czy AI może być kreatywna?
15 godzin temu
