Agentic data science with the Data Agent Kit
cloud.google.com·6h ago
AI Research
arXiv CS.AI·Jul 27·5 min read
A Statistical Multi-Objective Framework for Assessing Sensitivity of Radiomic AI Models to Acquisition Parameters
TL;DR
A new framework quantifies how radiomic AI models for lung cancer diagnosis are affected by varying CT scan parameters. It identifies critical parameter values that enhance model robustness across different datasets.
✦ Why It Matters
Evaluate and adjust CT scan parameters to improve the performance of radiomic AI models in clinical settings.
Key Takeaways
- Implement mixed-effects modeling to assess parameter sensitivity.
- Use hierarchical Pareto strategies for performance evaluation.
- Select critical scan parameters to enhance model robustness.
- Validate model performance across multiple datasets.
How It Works
The framework combines mixed-effects modeling, which accounts for variability across different datasets, with a hierarchical Pareto-based strategy to identify optimal scan parameters. This approach allows for quantifying the influence of various acquisition parameters on model performance, leading to the selection of configurations that maintain high accuracy and F1 scores.