Reproducibility of CT-based radiomic features against image resampling and perturbations for tumour and healthy kidney in renal cancer patients

Margherita Mottola,Leonardo Rundo,Evis Sala,Stephan Ursprung,Iosif Mendichovszky,Grant D Stewart,Lorena Escudero Sanchez,Tobias Klatte,Alessandro Bevilacqua

doi:10.1038/s41598-021-90985-y

Abstract

Computed Tomography (CT) is widely used in oncology for morphological evaluation and diagnosis, commonly through visual assessments, often exploiting semi-automatic tools as well. Well-established automatic methods for quantitative imaging offer the opportunity to enrich the radiologist interpretation with a large number of radiomic features, which need to be highly reproducible to be used reliably in clinical practice. This study investigates feature reproducibility against noise, varying resolutions and segmentations (achieved by perturbing the regions of interest), in a CT dataset with heterogeneous voxel size of 98 renal cell carcinomas (RCCs) and 93 contralateral normal kidneys (CK). In particular, first order (FO) and second order texture features based on both 2D and 3D grey level co-occurrence matrices (GLCMs) were considered. Moreover, this study carries out a comparative analysis of three of the most commonly used interpolation methods, which need to be selected before any resampling procedure. Results showed that the Lanczos interpolation is the most effective at preserving original information in resampling, where the median slice resolution coupled with the native slice spacing allows the best reproducibility, with 94.6% and 87.7% of features, in RCC and CK, respectively. GLCMs show their maximum reproducibility when used at short distances.

Highlights

Computed Tomography (CT) is widely used in oncology for morphological evaluation and diagnosis, commonly through visual assessments, often exploiting semi-automatic tools as well
To allows readers to assess the visual differences of these three methods, some exemplifying images are shown in Supplementary Figure S1 for three patients where linear, Akima, and Lanczos methods were respectively the best methods, in either upsampling and downsampling (v sl ), or both
Feature robustness depends on the tumour phenotype and is not g eneralizable14, this study focuses on the need for analysing feature robustness on renal cell carcinomas (RCCs) in CT using one of the largest datasets22

Summary

Introduction

Computed Tomography (CT) is widely used in oncology for morphological evaluation and diagnosis, commonly through visual assessments, often exploiting semi-automatic tools as well. A different approach has been recently p roposed, where the authors test different perturbation chains on NSCLC and head and neck cancer datasets, to find the chain better reproducing the outcome of a test-retest procedure, to be used when such method is not applicable All these studies, carried out on different tumours, analyse the reproducibility mainly against varying ROI segmentation, or a set of perturbations. To the best of our knowledge, this is the first work assessing robustness of first order (FO) and 2D and 3D second order texture features in CT imaging of renal cell carcinoma (RCC) and normal kidney (CK), by addressing three types of perturbations induced by Added White Gussian Noise (AWGN) (N), different voxel-size (V) and varying ROI (R). Results can provide practical operating guidelines to choose the proper voxel size in case of datasets with heterogeneous in-plane resolutions and to aggregate information derived from grey level (GL) co-occurrence matrices (GLCMs), improving standardisation of radiomic studies

Objectives

Methods

Results

Conclusion