| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Biological Materials Science
|
| Presentation Title |
Retrieval-Augmented Generation Platform to Enable Machine Learning Predictive Models of Biomedical Alloys |
| Author(s) |
Victor Manuel Villapun Puzas, Hasan Muhammad Sayeed, Sophie C. Cox, Taylor D. Sparks |
| On-Site Speaker (Planned) |
Victor Manuel Villapun Puzas |
| Abstract Scope |
2.5 million total knee and hip replacements were performed between 2012-2021 while 500k dental implants are placed annually, nevertheless, 2-23% and 10% of such devices fail, costing $1.62 billion to the USA. Tackling implant failure requires modulation of the wound healing process, critically calling for active biomedical alloys to supersede conventional titanium materials. Recent advances in Machine Learning (ML) can provide predicting models for semiconductors or high-entropy-alloys aided by large DFT databases. Nevertheless, polycrystallinity and relevance of post-manufacturing mechanical/heat treatments, limit their application to medical alloys, requiring the creation of relevant databases. Herein, we exploited Retrieval-Augmented Generation and textual data extraction using large language models to offer a platform for the rapid collation of biomaterial properties from readily available literature. Through two exemplary databases for Ti alloys, we will present a tool enabling large biomaterials database creation, providing the cornerstone required for ML biomaterial discovery hindering a $11.4 billion industry. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Biomaterials, Machine Learning, Titanium |