Cardiac SPECT radiomic features repeatability and reproducibility: A multi-scanner phantom study

Abstract

Background

The aim of this work was to assess the robustness of cardiac SPECT radiomic features against changes in imaging settings, including acquisition, and reconstruction parameters.

Methods

Four commercial SPECT and SPECT/CT cameras were used to acquire images of a static cardiac phantom mimicking typical myorcardial perfusion imaging using 185 MBq of ^99mTc. The effects of different image acquisition and reconstruction parameters, including number of views, view matrix size, attenuation correction, as well as image reconstruction related parameters (algorithm, number of iterations, number of subsets, type of post-reconstruction filter, and its associated parameters, including filter order and cut-off frequency) were studied. In total, 5,063 transverse views were reconstructed by varying the aforementioned factors. Eighty-seven radiomic features including first-, second-, and high-order textures were extracted from these images. To assess reproducibility and repeatability, the coefficient of variation (COV), as a widely adopted metric, was measured for each of the radiomic features over the different imaging settings.

Results

The Inverse Difference Moment Normalized (IDMN) and Inverse Difference Normalized (IDN) features from the Gray Level Co-occurrence Matrix (GLCM), Run Percentage (RP) from the Gray Level Co-occurrence Matrix (GLRLM), Zone Entropy (ZE) from the Gray Level Size Zone Matrix (GLSZM), and Dependence Entropy (DE) from the Gray Level Dependence Matrix (GLDM) feature sets were the only features that exhibited high reproducibility (COV ≤ 5%) against changes in all imaging settings. In addition, Large Area Low Gray Level Emphasis (LALGLE), Small Area Low Gray Level Emphasis (SALGLE) and Low Gray Level Zone Emphasis (LGLZE) from GLSZM, and Small Dependence Low Gray Level Emphasis (SDLGLE) from GLDM feature sets turned out to be less reproducible (COV > 20%) against changes in imaging settings. The GLRLM (31.88%) and GLDM feature set (54.2%) had the highest (COV < 5%) and lowest (COV > 20%) number of the reproducible features, respectively. Matrix size had the largest impact on feature variability as most of the features were not repeatable when matrix size was modified with 82.8% of them having a COV > 20%.

Conclusion

The repeatability and reproducibility of SPECT/CT cardiac radiomic features under different imaging settings is feature-dependent. Different image acquisition and reconstruction protocols have variable effects on radiomic features. The radiomic features exhibiting low COV are potential candidates for future clinical studies.

Introduction

As one of the major causes of mortality worldwide, cardiovascular disease is among the main concerns in public health.¹ Myocardial perfusion imaging (MPI) is a valuable non-invasive clinical tool enabling the functional assessment of coronary artery disease (CAD) to identify patients at risk with the aim to improve patient management.² In this regards, molecular imaging modalities, including single-photon emission computed tomography (SPECT) and positron emission tomography (PET), remain the most common procedures for the evaluation and risk stratification of patients with known or suspected CAD.³ Previous studies indicated that SPECT and SPECT/CT provide high image quality, low radiation exposure, and high diagnostic accuracy for the management of CAD.^4,5 Advances in nuclear cardiac imaging instrumentation and clinically validated software, including novel resolution recovery reconstruction algorithms incorporating correction for detector response, have enhanced image quality and quantitative accuracy in nuclear cardiovascular imaging.^6,7

Recently, quantitative radiomic studies have opened new horizons for better management of a number of diseases, including cancer and CAD.8, 9, 10, 11, 12 The aim of radiomics is to extract quantitative features from medical images using data-mining algorithms for survival, prognosis, and therapeutic response prediction and assessment.^8,13 In this context, radiomics could provide valuable information for personalized medicine. Previous studies have suggested that radiomic features could act as biomarkers to characterize and predict disease to provide support for patient management.^8,14

Radiomics have been widely used for the detection, diagnosis and prognosis of a number of diseases, including brain disorders,^15,16 various cancer types,^8,17 and more recently in cardiac diseases18, 19, 20, 21 using MRI, CT, and PET imaging modalities. With respect to the use of SPECT radiomics, this imaging modality has not been exploited to its full potential owing to its low spatial resolution and sensitivity. However, a number of recent studies demonstrated promising results using SPECT radiomic analysis in brain^15,16 and cardiac^18,19,22,23 disease management.

Medical image analysis remains a human enterprise. Owing to limitations of the human eyes and medical monitor display systems, a substantial amount of imperceptible but feature-rich information is unwittingly discarded.²⁴ To address this issue, radiomics quantifies medical images through mathematical extraction of different feature sets, thus enabling potential retrieval of the hidden information.

Based on biomarker discovery guidelines and studies, repeatability and reproducibility assessment of biomarkers are essential ingredients prior to clinical decision-making.²⁵ Regarding repeatability and reproducibility assessment, a reliable radiomic feature remains stable between two measurements when certain conditions change. Ideally, the features should also remain the same while the experimental settings, including equipment, software, processor, or operator vary.²⁶ When these conditions are fulfilled, the feature may be considered as a good biomarker for clinical setting. Hence, a considerable amount of literature has been published on radiomic features repeatability and reproducibility against changes in the radiomics generation process, such as image acquisition, reconstruction, pre-processing, segmentation, and data analysis.27, 28, 29 Nuclear radiomic studies have tested the repeatability and reproducibility of imaging features over various imaging parameters including reconstruction algorithms, matrix size, iteration number, number of subsets, and post-filtering using both phantom and clinical studies.^26,29,30 To overcome such vulnerability, it was suggested to consider the reproducibility and repeatability of radiomic features as a feasible measure to preselect features for further analysis⁸ These analyses (repeatability and reproducibility) are main frontiers, critical, and important tests in image biomarkers development to find robust features as imaging biomarkers based on the recommendations of the Quantitative Imaging Biomarker Alliance (QIBA) Technical Performance Working Group.³¹

To date, little evidence has been reported on cardiac SPECT imaging repeatability and reproducibility over different imaging settings. The present study aims to assess the repeatability and reproducibility of radiomic features using a dedicated cardiac phantom against variations in image acquisition and reconstruction protocols.