Image Contrast, Image Pre-Processing, and T(1) Mapping Affect MRI Radiomic Feature Repeatability in Patients with Colorectal Cancer Liver Metastases

SIMPLE SUMMARY: Medical images are data. They contain more information than is routinely identified by radiologists reading scans. Many scientists are investigating if extracting shape and grey-scale features from images can predict which oncology patients will respond to therapy. This approach, termed ‘radiomics’, must be validated before being ready for clinical use. One step is to determine measurement repeatability to ensure that radiomic features are robust, and that changes in features reflect genuine changes in tumours. In this study patients had two repeated sets of magnetic resonance imaging scans. We found that radiomic feature repeatability varied depending on scan acquisition parameters and the use of an administered contrast agent. We also compared how different repeatability assessment methods can best reveal these findings. We conclude that measuring radiomic feature repeatability is essential, but is also complex and prone to pitfalls. Overall, our study provides several insights into how radiomic feature repeatability is best assessed. ABSTRACT: Imaging biomarkers require technical, biological, and clinical validation to be translated into robust tools in research or clinical settings. This study contributes to the technical validation of radiomic features from magnetic resonance imaging (MRI) by evaluating the repeatability of features from four MR sequences: pre-contrast T [Formula: see text]- and T [Formula: see text]-weighted images, pre-contrast quantitative T [Formula: see text] maps (qT [Formula: see text]), and contrast-enhanced T [Formula: see text]-weighted images. Fifty-one patients with colorectal cancer liver metastases were scanned twice, up to 7 days apart. Repeatability was quantified using the intraclass correlation coefficient (ICC) and repeatability coefficient (RC), and the impact of non-Gaussian feature distributions and image normalisation was evaluated. Most radiomic features had non-Gaussian distributions, but Box–Cox transformations enabled ICCs and RCs to be calculated appropriately for an average of 97% of features across sequences. ICCs ranged from [Formula: see text] to [Formula: see text] , with volume and other shape features tending to be most repeatable; volume ICC > 0.98 for all sequences. 19% of features from non-normalised images exhibited significantly different ICCs in pair-wise sequence comparisons. Normalisation tended to increase ICCs for pre-contrast T [Formula: see text]- and T [Formula: see text]-weighted images, and decrease ICCs for qT [Formula: see text] maps. RCs tended to vary more between sequences than ICCs, showing that evaluations of feature performance depend on the chosen metric. This work suggests that feature-specific repeatability, from specific combinations of MR sequence and pre-processing steps, should be evaluated to select robust radiomic features as biomarkers in specific studies. In addition, as different repeatability metrics can provide different insights into a specific feature, consideration of the appropriate metric should be taken in a study-specific context.