Stability of radiomics features in apparent diffusion coefficient maps from a multi-centre test-retest trial

Area of Science:

• Radiology
• Medical Imaging
• Oncology

Background:

• Quantitative radiomics features from medical images offer prognostic value for clinical outcomes.
• The stability of radiomics features from apparent diffusion coefficient (ADC)-maps is crucial for reliable correlation with tumor pathology and clinical applications.

Purpose of the Study:

• To establish a method for analyzing radiomics features from ADC-maps in a multicentre, multi-vendor trial.
• To identify stable radiomics features across different tumor types, imaging vendors, and magnetic field strengths for improved prognostic modeling.

Main Methods:

• Retrospective analysis of 1322 radiomics features (shape, texture, intensity) from ADC-maps of ovarian, lung, and colorectal liver metastasis patients.
• Features were extracted from diffusion-weighted imaging acquired at 1.5T and 3T.
• Feature stability was assessed using the concordance correlation coefficient (CCC > 0.85).

Main Results:

• 122 radiomics features were found to be stable across all tumor entities, vendors, and field strengths.
• Some features were specific to tissue type or respiratory motion.
• A significant proportion of features demonstrated stability across different vendors and field strengths.

Conclusions:

• Extraction of stable phenotypic features from ADC-maps reduces dimensionality and enables the creation of reliable prognostic models.
• This approach facilitates the clinical implementation of ADC-based radiomics for cancer outcome prediction.