Assessing robustness of radiomic features by image perturbation

Abstract

Image features need to be robust against differences in positioning, acquisition and segmentation to ensure reproducibility. Radiomic models that only include robust features can be used to analyse new images, whereas models with non-robust features may fail to predict the outcome of interest accurately. Test-retest imaging is recommended to assess robustness, but may not be available for the phenotype of interest. We therefore investigated 18 combinations of image perturbations to determine feature robustness, based on noise addition (N), translation (T), rotation (R), volume growth/shrinkage (V) and supervoxel-based contour randomisation (C). Test-retest and perturbation robustness were compared for combined total of 4032 morphological, statistical and texture features that were computed from the gross tumour volume in two cohorts with computed tomography imaging: I) 31 non-small-cell lung cancer (NSCLC) patients; II): 19 head-and-neck squamous cell carcinoma (HNSCC) patients. Robustness was determined using the 95% confidence interval (CI) of the intraclass correlation coefficient (1, 1). Features with CI ≥ 0:90 were considered robust. The NTCV, TCV, RNCV and RCV perturbation chain produced similar results and identified the fewest false positive robust features (NSCLC: 0.2–0.9%; HNSCC: 1.7–1.9%). Thus, these perturbation chains may be used as an alternative to test-retest imaging to assess feature robustness.

Impact of image preprocessing methods on reproducibility of radiomic features in multimodal magnetic resonance imaging in glioblastoma

“…This can be easily interpreted due to shape features are not or less correlated to the gray level intensity distribution. Shape features reproducibility were predominantly reported by multiple radiomics studies, mostly in CT modality and in the limited literature for MR images …”

“…Radiomics features, reliability and reproducibility can be affected by various aspects of radiomics processing (e.g., image acquisition parameters and protocols, image preprocessing algorithms, tumor segmentation, and software used for processing and feature extractions). Major of radiomics studies by concerning a different aspect of radiomics reproducibility and repeatability issue was done in computed tomography (CT) and PET modalities for limited cancer types, and a few studies have been reported in MRI …”

Application of radiomics for the prediction of HPV status for patients with head and neck cancers

“…Up sampling may also introduce image aliasing artifacts as well that requires anti-aliasing filters prior to filtering. 55,56,85 Multiple-scaling strategies potentially offer a good trade-off. 86 However, more studies are necessary to confirm the best sampling strategy.…”

Technical Note: An IBEX adaption toward image biomarker standardization

“…Currently, many free software packages are available for radiomic feature calculation in the form of stand‐alone programs, modules and libraries such as MaZda, CGITA, IBEX, LIFEx, MITK Phenotyping, RaCaT, CERR radiomic extension, and Pyradiomics . Moreover, several studies report in‐house developed radiomic programs …”

“…8 Moreover, several studies report in-house developed radiomic programs. [9][10][11][12][13][14][15] Many challenges hamper the reproducibility and validation of radiomic studies, 16,17 including ambiguous feature nomenclature and variability of feature definitions between software platforms. Indeed, standardization of radiomic software is a fundamental step to ease the comparison and validation of studies across different institutions and for a possible translation of radiomics into clinical practice.…”