Minimizing acquisition-related radiomics variability by image resampling and batch effect correction to allow for large-scale data analysis - PubMed (original) (raw)

. 2021 Mar;31(3):1460-1470.

doi: 10.1007/s00330-020-07174-0. Epub 2020 Sep 9.

Olivia Jordi-Ollero 2, Kinga Bernatowicz 1, Alonso Garcia-Ruiz 1, Eric Delgado-Muñoz 1, David Leiva 3, Richard Mast 4, Cristina Suarez 5, Roser Sala-Llonch 6, Nahum Calvo 3, Manuel Escobar 4, Arturo Navarro-Martin 7, Guillermo Villacampa 8, Rodrigo Dienstmann 8, Raquel Perez-Lopez 9 10

Affiliations

Marta Ligero et al. Eur Radiol. 2021 Mar.

Abstract

Objective: To identify CT-acquisition parameters accounting for radiomics variability and to develop a post-acquisition CT-image correction method to reduce variability and improve radiomics classification in both phantom and clinical applications.

Methods: CT-acquisition protocols were prospectively tested in a phantom. The multi-centric retrospective clinical study included CT scans of patients with colorectal/renal cancer liver metastases. Ninety-three radiomics features of first order and texture were extracted. Intraclass correlation coefficients (ICCs) between CT-acquisition protocols were evaluated to define sources of variability. Voxel size, ComBat, and singular value decomposition (SVD) compensation methods were explored for reducing the radiomics variability. The number of robust features was compared before and after correction using two-proportion z test. The radiomics classification accuracy (K-means purity) was assessed before and after ComBat- and SVD-based correction.

Results: Fifty-three acquisition protocols in 13 tissue densities were analyzed. Ninety-seven liver metastases from 43 patients with CT from two vendors were included. Pixel size, reconstruction slice spacing, convolution kernel, and acquisition slice thickness are relevant sources of radiomics variability with a percentage of robust features lower than 80%. Resampling to isometric voxels increased the number of robust features when images were acquired with different pixel sizes (p < 0.05). SVD-based for thickness correction and ComBat correction for thickness and combined thickness-kernel increased the number of reproducible features (p < 0.05). ComBat showed the highest improvement of radiomics-based classification in both the phantom and clinical applications (K-means purity 65.98 vs 73.20).

Conclusion: CT-image post-acquisition processing and radiomics normalization by means of batch effect correction allow for standardization of large-scale data analysis and improve the classification accuracy.

Key points: • The voxel size (accounting for the pixel size and slice spacing), slice thickness, and convolution kernel are relevant sources of CT-radiomics variability. • Voxel size resampling increased the mean percentage of robust CT-radiomics features from 59.50 to 89.25% when comparing CT scans acquired with different pixel sizes and from 71.62 to 82.58% when the scans were acquired with different slice spacings. • ComBat batch effect correction reduced the CT-radiomics variability secondary to the slice thickness and convolution kernel, improving the capacity of CT-radiomics to differentiate tissues (in the phantom application) and the primary tumor type from liver metastases (in the clinical application).

Keywords: Image processing; Metastasis; Radiologic phantom; X-ray computed tomography.

PubMed Disclaimer

Conflict of interest statement

The authors of this manuscript declare no relationships with any companies whose products or services may be related to the subject matter of the article.

Figures

Fig. 1

Fig. 1

Methodology flowchart

Fig. 2

Fig. 2

Axial CT of the Gammex 467 Tissue Characterization Phantom showing the thirteen tissue and water materials (a) with the segmented volumes of interest (VOI) for the different rod materials (b). Axial enhanced CT of the abdomen showing the target liver metastases (green and red masks) of a patient with clear cell renal carcinoma (c) and a patient with colorectal adenocarcinoma (d)

Fig. 3

Fig. 3

Intraclass correlation coefficients (ICCs) of the radiomics features of first order and texture matrices (gray-level co-occurrence matrix [GLCM], gray-level dependence matrix [GLDM], gray-level run length matrix [GLRLM], gray-level size zone matrix [GLSZM], neighboring gray-tone different matrix [NGTDM]) between extreme CT-acquisition parameters in the phantom application

Fig. 4

Fig. 4

Principal component analysis (PCA) before and after resampling to 1 × 1 × 1 mm3 voxels of CT images acquired with different pixel sizes (a) and slice spacings (b). PC4 (explaining 7.14% of the radiomics data variance) is associated with the different acquisition pixel sizes before resampling. PC2 (18.42%) is associated with the distribution of the different acquisition pixel heights. After resampling, the acquisition voxel size (accounting for pixel size and slice spacing) is not associated with the variance explained by the PCA

Fig. 5

Fig. 5

Principal component analysis (PCA) of the brain and liver material radiomics distribution before and after convolution kernel–slice thickness ComBat correction. The distance between the radiomics data of the brain and liver materials from CT scans with different acquisitions protocols increases after applying batch correction (i.e., the radiomics distribution better reflects differences between materials and not due to the CT-acquisition parameters)

Fig. 6

Fig. 6

Principal component analysis (PCA) of the liver metastasis radiomics distribution from CT scans of patients with colorectal adenocarcinoma and clear cell renal carcinoma. PCA before and after convolution kernel ComBat correction. The distribution of the groups of patients with different tumor types differs more after batch correction. The first component (PC1 [%]) of data variance can differentiate better between groups (colorectal versus renal) after correction

References

    1. Lambin P, Leijenaar RTH, Deist TM, et al. Radiomics: the bridge between medical imaging and personalized medicine. Nat Rev Clin Oncol. 2017;14:749–762. doi: 10.1038/nrclinonc.2017.141. - DOI - PubMed
    1. Rizzo S, Botta F, Raimondi S, et al. Radiomics: the facts and the challenges of image analysis. Eur Radiol Exp. 2018;2:36. doi: 10.1186/s41747-018-0068-z. - DOI - PMC - PubMed
    1. Sun R, Limkin EJ, Vakalopoulou M, et al. A radiomics approach to assess tumour-infiltrating CD8 cells and response to anti-PD-1 or anti-PD-L1 immunotherapy: an imaging biomarker, retrospective multicohort study. Lancet Oncol. 2018;19:1180–1191. doi: 10.1016/S1470-2045(18)30413-3. - DOI - PubMed
    1. Trebeschi S, Drago SG, Birkbak NJ et al (2019) Predicting response to cancer immunotherapy using non-invasive radiomic biomarkers. Ann Oncol. 10.1093/annonc/mdz108 - PMC - PubMed
    1. Aerts HJWL, Velazquez ER, Leijenaar RT et al (2014) Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach. Nat Commun 5:4006 - PMC - PubMed

MeSH terms

LinkOut - more resources