The Effects of In-Plane Spatial Resolution on CT-Based Radiomic Features’ Stability with and without ComBat Harmonization

Abstract

While handcrafted radiomic features (HRFs) have shown promise in the field of personalized medicine, many hurdles hinder its incorporation into clinical practice, including but not limited to their sensitivity to differences in acquisition and reconstruction parameters. In this study, we evaluated the effects of differences in in-plane spatial resolution (IPR) on HRFs, using a phantom dataset (n = 14) acquired on two scanner models. Furthermore, we assessed the effects of interpolation methods (IMs), the choice of a new unified in-plane resolution (NUIR), and ComBat harmonization on the reproducibility of HRFs. The reproducibility of HRFs was significantly affected by variations in IPR, with pairwise concordant HRFs, as measured by the concordance correlation coefficient (CCC), ranging from 42% to 95%. The number of concordant HRFs (CCC > 0.9) after resampling varied depending on (i) the scanner model, (ii) the IM, and (iii) the NUIR. The number of concordant HRFs after ComBat harmonization depended on the variations between the batches harmonized. The majority of IMs resulted in a higher number of concordant HRFs compared to ComBat harmonization, and the combination of IMs and ComBat harmonization did not yield a significant benefit. Our developed framework can be used to assess the reproducibility and harmonizability of RFs.

MRI-Based Radiomics Analysis for the Pretreatment Prediction of Pathologic Complete Tumor Response to Neoadjuvant Systemic Therapy in Breast Cancer Patients: A Multicenter Study

“…Despite the promising potential of radiomics, several hurdles that impede the clinical implementation of radiomics models have been identified. One of these is the sensitivity of radiomics features to the variations in acquisition and reconstruction parameters across different imaging modalities [ 22 , 23 , 24 , 25 , 26 ], and some features were found not to be reproducible even in test-retest scenarios [ 27 , 28 , 29 ].…”

Reply to Orlhac, F.; Buvat, I. Comment on “Ibrahim et al. The Effects of In-Plane Spatial Resolution on CT-Based Radiomic Features’ Stability with and without ComBat Harmonization. Cancers 2021, 13, 1848”

“…We would like to thank Orlhac and Buvat [ 1 ] for their commentary on our article [ 2 ]. Orlhac and Buvat present the opinion that we “misused” ComBat harmonisation to assess radiomic features in a computed tomography (CT) phantom by evaluating the phantom as a whole.…”

“…They state that we must apply ComBat harmonisation separately to each layer of the phantom, akin to restricting a radiomics study to either liver or tumour. However, the main aim of our work [ 2 ] was not to address a specific radiomics task, but to use CT phantom data to evaluate the robustness of 91 radiomics features to changes in voxel size, either alone or with two harmonisation methods—interpolation and ComBat.…”

“…We believe that this is overly prescriptive. In our usage, the sample is the phantom, which is intended to represent a range of tissue types, because we sought to understand how acquisition differences affect each measure over a range of materials [ 2 ]. This is consistent with the use of Combat by Fortin et al [ 4 ].…”

“…The aim of the design matrix of the biologic covariates in ComBat is to preserve biologic information while harmonising the features [ 3 , 8 , 9 ]. However, as we have stated [ 2 ], we performed this study to examine the impact of pixel interpolation on radiomic features in a phantom, and no biologic covariates were appropriate for our study. We clearly state in the discussion that anthropomorphic phantom scans provide some evidence of the reproducibility of features, but that they cannot completely represent features extracted from human images, and human or cadaveric reproducibility studies are encouraged when ethical.…”