Impact of Preprocessing and Harmonization Methods on the Removal of Scanner Effects in Brain MRI Radiomic Features

Li, Yingping; Ammari, Samy; Balleyguier, Corinne; Lassau, Nathalie; Chouzenoux, Émilie

doi:10.3390/cancers13123000

Cited by 57 publications

(37 citation statements)

References 57 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To obtain robust features, linear interpolation was first adopted to resample the voxel size of the image to an isovolumetric voxel (1 × 1 × 1 mm 3 ) before feature extraction ( 21 ). The Z-score method was used to standardize the image, and image intensity discretization was applied with a fixed bin width of 5 ( 22 ).…”

Section: Methodsmentioning

confidence: 99%

Multi-Parametric Magnetic Resonance Imaging-Based Radiomics Analysis of Cervical Cancer for Preoperative Prediction of Lymphovascular Space Invasion

et al. 2022

View full text Add to dashboard Cite

BackgroundDetection of lymphovascular space invasion (LVSI) in early cervical cancer (CC) is challenging. To date, no standard clinical markers or screening tests have been used to detect LVSI preoperatively. Therefore, non-invasive risk stratification tools are highly desirable.ObjectiveTo train and validate a multi-parametric magnetic resonance imaging (mpMRI)-based radiomics model to detect LVSI in patients with CC and investigate its potential as a complementary tool to enhance the efficiency of risk assessment strategies.Materials and MethodsThe model was developed from the tumor volume of interest (VOI) of 125 patients with CC. A total of 1037 radiomics features obtained from conventional magnetic resonance imaging (MRI), including a small field-of-view (sFOV) high-resolution (HR)-T2-weighted MRI (T2WI), apparent diffusion coefficient (ADC), T2WI, fat-suppressed (FS)-T2WI, as well as axial and sagittal contrast-enhanced T1-weighted MRI (T1c). We conducted a radiomics-based characterization of each tumor region using pretreatment image data. Feature selection was performed using the least absolute shrinkage and selection operator method on the training set. The predictive performance was compared with single variates (clinical data and single-layer radiomics signatures) analyzed using a receiver operating characteristic (ROC) curve. Three-fold cross-validation performed 20 times was used to evaluate the accuracy of the trained classifiers and the stability of the selected features. The models were validated by using a validation set.ResultsFeature selection extracted the six most important features (3 from sFOV HR-T2WI, 1 T2WI, 1 FS-T2WI, and 1 T1c) for model construction. The mpMRI-combined radiomics model (area under the curve [AUC]: 0.940) reached a significantly higher performance (better than the clinical parameters [AUC: 0.730]), including any single-layer model using sFOV HR-T2WI (AUC: 0.840), T2WI (AUC: 0.770), FS-T2WI (AUC: 0.710), ADC maps (AUC: 0.650), sagittal, and axial T1c values (AUC: 0.710, 0.680) in the validation set.ConclusionBiomarkers using multi-parametric radiomics features derived from preoperative MR images could predict LVSI in patients with CC.

show abstract

Section: Methodsmentioning

confidence: 99%

Multi-Parametric Magnetic Resonance Imaging-Based Radiomics Analysis of Cervical Cancer for Preoperative Prediction of Lymphovascular Space Invasion

et al. 2022

View full text Add to dashboard Cite

show abstract

“…The TCGA-GBM and TCGA-LGG datasets used in our paper include patients collected from different institutions with different image scanners or clinical protocols; thus, the nonbiological scanner effects might be introduced. We applied Z-score normalization on the MRI images as a preprocessing step; however, according to the conclusion of our recent study [ 43 ], only applying the intensity normalization might not be enough for radiomics models. Thus, we tested the ComBat method to see whether it helped improve the feature ability of the radiomic models.…”

Section: Resultsmentioning

confidence: 99%

“…As stated previously, the clinical MRI scans have already been preprocessed, including reorientation, co-registration, resampling, and skull-stripping. Li et al [ 43 ] demonstrated that the intensity normalization would help relieve the nonbiological variations at the image level, which were introduced by different MRI scanners or protocols when collecting the images. Thus, we show the image histograms and compare the radiomics model performance before and after applying Z-score intensity normalization in the following section.…”

Section: Materials and Methodsmentioning

confidence: 99%

Radiomics-Based Method for Predicting the Glioma Subtype as Defined by Tumor Grade, IDH Mutation, and 1p/19q Codeletion

Lawrance

Quillent

et al. 2022

Cancers

Self Cite

View full text Add to dashboard Cite

Gliomas are among the most common types of central nervous system (CNS) tumors. A prompt diagnosis of the glioma subtype is crucial to estimate the prognosis and personalize the treatment strategy. The objective of this study was to develop a radiomics pipeline based on the clinical Magnetic Resonance Imaging (MRI) scans to noninvasively predict the glioma subtype, as defined based on the tumor grade, isocitrate dehydrogenase (IDH) mutation status, and 1p/19q codeletion status. A total of 212 patients from the public retrospective The Cancer Genome Atlas Low Grade Glioma (TCGA-LGG) and The Cancer Genome Atlas Glioblastoma Multiforme (TCGA-GBM) datasets were used for the experiments and analyses. Different settings in the radiomics pipeline were investigated to improve the classification, including the Z-score normalization, the feature extraction strategy, the image filter applied to the MRI images, the introduction of clinical information, ComBat harmonization, the classifier chain strategy, etc. Based on numerous experiments, we finally reached an optimal pipeline for classifying the glioma tumors. We then tested this final radiomics pipeline on the hold-out test data with 51 randomly sampled random seeds for reliable and robust conclusions. The results showed that, after tuning the radiomics pipeline, the mean AUC improved from 0.8935 (±0.0351) to 0.9319 (±0.0386), from 0.8676 (±0.0421) to 0.9283 (±0.0333), and from 0.6473 (±0.1074) to 0.8196 (±0.0702) in the test data for predicting the tumor grade, IDH mutation, and 1p/19q codeletion status, respectively. The mean accuracy for predicting the five glioma subtypes also improved from 0.5772 (±0.0816) to 0.6716 (±0.0655). Finally, we analyzed the characteristics of the radiomic features that best distinguished the glioma grade, the IDH mutation, and the 1p/19q codeletion status, respectively. Apart from the promising prediction of the glioma subtype, this study also provides a better understanding of the radiomics model development and interpretability. The results in this paper are replicable with our python codes publicly available in github.

show abstract

“…Diffusion and perfusion maps were then co-registered to anatomical images in the BraTS space with ANTs. Intensity normalization was performed for all anatomical images (T1w, CE-T1w, T2w, and FLAIR) using the WhiteStripe (WS) method [ 23 ] implemented in the Python library found at (last accessed on 4 December 2021) [ 24 ]. The normalization was performed on the preprocessed images, i.e., on the images co-registered in the BraTS space.…”

Section: Methodsmentioning

confidence: 99%

Radiomics-Based Detection of Radionecrosis Using Harmonized Multiparametric MRI

Acquitter

Piram

Sabatini

et al. 2022

Cancers

View full text Add to dashboard Cite

In this study, a radiomics analysis was conducted to provide insights into the differentiation of radionecrosis and tumor progression in multiparametric MRI in the context of a multicentric clinical trial. First, the sensitivity of radiomic features to the unwanted variability caused by different protocol settings was assessed for each modality. Then, the ability of image normalization and ComBat-based harmonization to reduce the scanner-related variability was evaluated. Finally, the performances of several radiomic models dedicated to the classification of MRI examinations were measured. Our results showed that using radiomic models trained on harmonized data achieved better predictive performance for the investigated clinical outcome (balanced accuracy of 0.61 with the model based on raw data and 0.72 with ComBat harmonization). A comparison of several models based on information extracted from different MR modalities showed that the best classification accuracy was achieved with a model based on MR perfusion features in conjunction with clinical observation (balanced accuracy of 0.76 using LASSO feature selection and a Random Forest classifier). Although multimodality did not provide additional benefit in predictive power, the model based on T1-weighted MRI before injection provided an accuracy close to the performance achieved with perfusion.

show abstract

Impact of Preprocessing and Harmonization Methods on the Removal of Scanner Effects in Brain MRI Radiomic Features

Cited by 57 publications

References 57 publications

Multi-Parametric Magnetic Resonance Imaging-Based Radiomics Analysis of Cervical Cancer for Preoperative Prediction of Lymphovascular Space Invasion

Multi-Parametric Magnetic Resonance Imaging-Based Radiomics Analysis of Cervical Cancer for Preoperative Prediction of Lymphovascular Space Invasion

Radiomics-Based Method for Predicting the Glioma Subtype as Defined by Tumor Grade, IDH Mutation, and 1p/19q Codeletion

Radiomics-Based Detection of Radionecrosis Using Harmonized Multiparametric MRI

Contact Info

Product

Resources

About