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

Abdalla Ibrahim,Henry C Woodruff,Raymond J Acciavatti,Andrew D A Maidment,Roland Hustinx,Joachim E Wildberger,Turkey Refaee,Bruno Barufaldi,Sergey Primakov,Philippe Lambin,Renée W Y Granzier,Felix M Mottaghy

doi:10.3390/cancers13081848

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.

Highlights

In recent years, quantitative medical imaging research using handcrafted radiomic features (HRFs) has been growing exponentially [1,2]
We found that linear interpolation is not a good candidate for the purpose of improving the reproducibility of HRFs when compared to other available interpolation methods (IMs), and the performance of an IM is dependent on the original in-plane spatial resolution (IPR) values and the chosen new unified in-plane resolution (NUIR) as well as the imaging vendor
Other HRFs were found to be concordant across different scanner models and IPRs. These findings indicate the need for performing reproducibility studies depending on the data under study, and the fact that at this level, we are unable to provide a list of HRFs that can be used regardless of the acquisition and reconstruction parameters and scanner models used

Summary

Introduction

Quantitative medical imaging research using handcrafted radiomic features (HRFs) has been growing exponentially [1,2]. Radiomics refers to the high throughput extraction of quantitative imaging features that are expected to correlate with clinical and biological characteristics of patients [3,4]. It has been hypothesized that image texture analysis could potentially extract more information from a region of interest (ROI) than that solely perceived by the human eye [5,6]. Many studies have reported on the potential of radiomics to predict various clinical endpoints [9,10]. Major challenges, including the reproducibility of the HRFs across different acquisition and reconstruction parameters, have hindered the incorporation of radiomics in clinical decision support systems [11,12]

Objectives

Methods

Results

Discussion

Conclusion