Advancing Direct Tablet Compression with AI: New Research Published in the European Journal of Pharmaceutical Sciences
We are pleased to share that our latest research article, “Advancing Direct Tablet Compression with AI: A Multi-task Framework for Quality Control, Batch Acceptance, and Causal Analysis”, has been published in the European Journal of Pharmaceutical Sciences (June 11, 2025).
This study introduces a novel AI-driven framework that integrates regression, classification, and text generation into a unified model for quality control in pharmaceutical tablet manufacturing. Using data from the Harvard Dataverse, our multi-task system predicts four critical quality attributes—friability, hardness, disintegration time, and water absorption ratio—and determines whether tablets or batches should be accepted or rejected. It also provides textual explanations to support transparent decision-making.
Key findings:
- Achieved 91.8% R² for regression tasks and 95.5% accuracy in classification.
- Supports real-time monitoring and process optimization in alignment with Pharma 4.0 goals.
- Offers an interpretable, end-to-end AI solution that minimizes waste and enhances regulatory compliance.
The framework stands out by combining data augmentation, neural networks, and generative AI to improve prediction accuracy and causal reasoning. We believe it marks a significant step forward in intelligent pharmaceutical manufacturing.
This work was conducted as part of the LifeFactFuture (LFF) project and funded by Business Finland. We thank all project partners for their collaboration and support.
Yazid Bounab, Osmo Antikainen, Mia Sivén & Anne Juppo
University of Helsinki, Faculty of Pharmacy