AutoRadiomics

AutoRadiomics performs end-to-end radiomic analysis of medical imaging data to extract radiomic features and apply machine learning for standardized, reproducible clinical prediction and evaluation.

Key Features:

• Comprehensive Workflow: Includes image preprocessing, feature extraction, feature selection, modeling, and model evaluation.
• Optimization of Parameters: Automatically selects optimal parameters for specific tasks.
• Reproducibility and Standardization: Facilitates standardized workflows to improve reproducibility across radiomic studies.

Scientific Applications:

• Clinical Datasets Evaluated: Evaluated on eight open-source clinical datasets, including six from the WORC database and two prostate MRI datasets (Prostate-UCLA and PROSTATEx).
• Performance Metrics: Reported AUCs ranged from 0.56 for lung melanoma metastases detection to 0.93 for liposarcoma detection in WORC datasets, and 0.51–0.77 for prostate cancer detection tasks.
• Reproducibility of Results: Replicated previously reported results without significant overfitting between training and test sets.

Methodology:

Employs machine learning on radiomic features extracted from medical imaging data and automates selection of optimal parameters within a standardized workflow.