Epistemic Uncertainty-Aware Defect Detection for Medical Devices
What it is
This paper proposes a machine learning framework that uses Knowledge Graphs to represent heterogeneous manufacturing and device-report data. It introduces a principled rejection strategy that abstains from predictions when estimated epistemic uncertainty exceeds a threshold, allowing explicit control over the reliability of automated defect detection.
The problem it solves
Standard machine learning models in high-stakes manufacturing often issue confident but incorrect predictions on uncertain data, leading to potential safety risks and reduced trust in automated quality control systems.
What you could use it for
- Automated quality control in medical device manufacturing
- Filtering high-risk cases for manual inspection in FDA MAUDE report analysis
- Improving the safety of AI-based decision support in high-stakes industrial settings
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.