Deep learning reconstruction improves radiomics feature stability and discriminative power in abdominal CT imaging: a phantom study

Abstract

Objectives To compare image quality of deep learning reconstruction (AiCE) for radiomics feature extraction with filtered back projection (FBP), hybrid iterative reconstruction (AIDR 3D), and model-based iterative reconstruction (FIRST). Methods Effects of image reconstruction on radiomics features were investigated using a phantom that realistically mimicked a 65-year-old patient’s abdomen with hepatic metastases. The phantom was scanned at 18 doses from 0.2 to 4 mGy, with 20 repeated scans per dose. Images were reconstructed with FBP, AIDR 3D, FIRST, and AiCE. Ninety-three radiomics features were extracted from 24 regions of interest, which were evenly distributed across three tissue classes: normal liver, metastatic core, and metastatic rim. Features were analyzed in terms of their consistent characterization of tissues within the same image (intraclass correlation coefficient ≥ 0.75), discriminative power (Kruskal-Wallis test p value < 0.05), and repeatability (overall concordance correlation coefficient ≥ 0.75). Results The median fraction of consistent features across all doses was 6%, 8%, 6%, and 22% with FBP, AIDR 3D, FIRST, and AiCE, respectively. Adequate discriminative power was achieved by 48%, 82%, 84%, and 92% of features, and 52%, 20%, 17%, and 39% of features were repeatable, respectively. Only 5% of features combined consistency, discriminative power, and repeatability with FBP, AIDR 3D, and FIRST versus 13% with AiCE at doses above 1 mGy and 17% at doses ≥ 3 mGy. AiCE was the only reconstruction technique that enabled extraction of higher-order features. Conclusions AiCE more than doubled the yield of radiomics features at doses typically used clinically. Inconsistent tissue characterization within CT images contributes significantly to the poor stability of radiomics features. Key Points • Image quality of CT images reconstructed with filtered back projection and iterative methods is inadequate for the majority of radiomics features due to inconsistent tissue characterization, low discriminative power, or low repeatability. • Deep learning reconstruction enhances image quality for radiomics and more than doubled the feature yield at doses that are typically used in clinical CT imaging. • Image reconstruction algorithms can optimize image quality for more reliable quantification of tissues in CT images.

Deep learning-based reconstruction on cardiac CT yields distinct radiomic features compared to iterative and filtered back projection reconstructions

“…Our study showed that most radiomic features obtained from LV myocardium differ depending on the reconstruction methods and are presumably affected by the reconstruction method. This result is in line with a previous study showing that applying DLR yielded superior feature consistency, discriminative power and repeatability to IR, and FBP for radiomic features on abdominal CT 30 . DLR uses a deep neural network to enhance image quality by removing noise from signal without changing the noise texture itself and is thought to produce adequate images for feature extraction and diagnostic modelling.…”

“…The IR algorithm tended to produce CT images with a more “plastic-looking” texture than FBP reconstructions, so the effects of IR on quantitative radiomic features make sense 29 . Although DLR aims to produce a more natural texture similar to FBP reconstruction, few studies to date have investigated the effects of DLR on quantitative radiomic features 30 . Our study showed that most radiomic features obtained from LV myocardium differ depending on the reconstruction methods and are presumably affected by the reconstruction method.…”

Evaluation of radiomics feature stability in abdominal monoenergetic photon counting CT reconstructions

“…Furthermore, standardization initiatives such as the image biomarker standardization initiative 40 can facilitate increasing the comparability of quantitative imaging data as it was shown that extraction algorithms can have an influence on radiomics features. Deep learning-based reconstruction models can be used 41 to standardize images and increase feature stability. Now, the implementation of PCCT with its intrinsic capability of spectral reconstructions may be regarded as an additional tool to address the issue of feature stability, which remains a relevant impediment to the clinical application of radiomics, by utilization of energy-specific reconstruction techniques.…”

Identification of CT Imaging Phenotypes of Colorectal Liver Metastases from Radiomics Signatures—Towards Assessment of Interlesional Tumor Heterogeneity

“…A widely proposed solution to further increase the stability of quantitative imaging approaches is the application of deep learning reconstruction methods to imaging data [ 33 ]. While this approach has shown promising results in the prediction of known phenotypes [ 34 ], it has some limitations in an unsupervised approach, because a ground truth for training cannot be defined easily.…”