A score combining baseline neutrophilia and primary tumor SUVpeak measured from FDG PET is associated with outcome in locally advanced cervical cancer.

PURPOSE: We investigated whether a score combining baseline neutrophilia and a PET biomarker could predict outcome in patients with locally advanced cervical cancer (LACC). METHODS: Patients homogeneously treated with definitive chemoradiation plus image-guided adaptive brachytherapy (IGABT) between 2006 and 2013 were analyzed retrospectively. We divided patients into two groups depending on the PET device used: a training set (TS) and a validation set (VS). Primary tumors were semi-automatically delineated on PET images, and 11 radiomics features were calculated (LIFEx software). A PET radiomic index was selected using the time-dependent area under the curve (td-AUC) for 3-year local control (LC). We defined the neutrophil SUV grade (NSG = 0, 1 or 2) score as the number of risk factors among (i) neutrophilia (neutrophil count >7 G/L) and (ii) high risk defined from the PET radiomic index. The NSG prognostic value was evaluated for LC and overall survival (OS). RESULTS: Data from 108 patients were analyzed. Estimated 3-year LC was 72% in the TS (n = 69) and 65% in the VS (n = 39). In the TS, SUVpeak was selected as the most LC-predictive biomarker (td-AUC = 0.75), and was independent from neutrophilia (p = 0.119). Neutrophilia (HR = 2.6), high-risk SUVpeak (SUVpeak > 10, HR = 4.4) and NSG = 2 (HR = 9.2) were associated with low probability of LC in TS. In multivariate analysis, NSG = 2 was independently associated with low probability of LC (HR = 7.5, p < 0.001) and OS (HR = 5.8, p = 0.001) in the TS. Results obtained in the VS (HR = 5.2 for OS and 3.5 for LC, p < 0.02) were promising. CONCLUSION: This innovative scoring approach combining baseline neutrophilia and a PET biomarker provides an independent prognostic factor to consider for further clinical investigations.

Development of a Machine Learning Classifier Based on Radiomic Features Extracted From Post-Contrast 3D T1-Weighted MR Images to Distinguish Glioblastoma From Solitary Brain Metastasis.

OBJECTIVES: To differentiate Glioblastomas (GBM) and Brain Metastases (BM) using a radiomic features-based Machine Learning (ML) classifier trained from post-contrast three-dimensional T1-weighted (post-contrast 3DT1) MR imaging, and compare its performance in medical diagnosis versus human experts, on a testing cohort. METHODS: We enrolled 143 patients (71 GBM and 72 BM) in a retrospective bicentric study from January 2010 to May 2019 to train the classifier. Post-contrast 3DT1 MR images were performed on a 3-Tesla MR unit and 100 radiomic features were extracted. Selection and optimization of the Machine Learning (ML) classifier was performed using a nested cross-validation. Sensitivity, specificity, balanced accuracy, and area under the receiver operating characteristic curve (AUC) were calculated as performance metrics. The model final performance was cross-validated, then evaluated on a test set of 37 patients, and compared to human blind reading using a McNemar's test. RESULTS: The ML classifier had a mean [95% confidence interval] sensitivity of 85% [77; 94], a specificity of 87% [78; 97], a balanced accuracy of 86% [80; 92], and an AUC of 92% [87; 97] with cross-validation. Sensitivity, specificity, balanced accuracy and AUC were equal to 75, 86, 80 and 85% on the test set. Sphericity 3D radiomic index highlighted the highest coefficient in the logistic regression model. There were no statistical significant differences observed between the performance of the classifier and the experts' blinded examination. CONCLUSIONS: The proposed diagnostic support system based on radiomic features extracted from post-contrast 3DT1 MR images helps in differentiating solitary BM from GBM with high diagnosis performance and generalizability.

Standardization of brain MR images across machines and protocols: bridging the gap for MRI-based radiomics.

Radiomics relies on the extraction of a wide variety of quantitative image-based features to provide decision support. Magnetic resonance imaging (MRI) contributes to the personalization of patient care but suffers from being highly dependent on acquisition and reconstruction parameters. Today, there are no guidelines regarding the optimal pre-processing of MR images in the context of radiomics, which is crucial for the generalization of published image-based signatures. This study aims to assess the impact of three different intensity normalization methods (Nyul, WhiteStripe, Z-Score) typically used in MRI together with two methods for intensity discretization (fixed bin size and fixed bin number). The impact of these methods was evaluated on first- and second-order radiomics features extracted from brain MRI, establishing a unified methodology for future radiomics studies. Two independent MRI datasets were used. The first one (DATASET1) included 20 institutional patients with WHO grade II and III gliomas who underwent post-contrast 3D axial T1-weighted (T1w-gd) and axial T2-weighted fluid attenuation inversion recovery (T2w-flair) sequences on two different MR devices (1.5 T and 3.0 T) with a 1-month delay. Jensen-Shannon divergence was used to compare pairs of intensity histograms before and after normalization. The stability of first-order and second-order features across the two acquisitions was analysed using the concordance correlation coefficient and the intra-class correlation coefficient. The second dataset (DATASET2) was extracted from the public TCIA database and included 108 patients with WHO grade II and III gliomas and 135 patients with WHO grade IV glioblastomas. The impact of normalization and discretization methods was evaluated based on a tumour grade classification task (balanced accuracy measurement) using five well-established machine learning algorithms. Intensity normalization highly improved the robustness of first-order features and the performances of subsequent classification models. For the T1w-gd sequence, the mean balanced accuracy for tumour grade classification was increased from 0.67 (95% CI 0.61-0.73) to 0.82 (95% CI 0.79-0.84, P = .006), 0.79 (95% CI 0.76-0.82, P = .021) and 0.82 (95% CI 0.80-0.85, P = .005), respectively, using the Nyul, WhiteStripe and Z-Score normalization methods compared to no normalization. The relative discretization makes unnecessary the use of intensity normalization for the second-order radiomics features. Even if the bin number for the discretization had a small impact on classification performances, a good compromise was obtained using the 32 bins considering both T1w-gd and T2w-flair sequences. No significant improvements in classification performances were observed using feature selection. A standardized pre-processing pipeline is proposed for the use of radiomics in MRI of brain tumours. For models based on first- and second-order features, we recommend normalizing images with the Z-Score method and adopting an absolute discretization approach. For second-order feature-based signatures, relative discretization can be used without prior normalization. In both cases, 32 bins for discretization are recommended. This study may pave the way for the multicentric development and validation of MR-based radiomics biomarkers.

Influence of Magnetic Field Strength on Magnetic Resonance Imaging Radiomics Features in Brain Imaging, an In Vitro and In Vivo Study.

BACKGROUND: The development and clinical adoption of quantitative imaging biomarkers (radiomics) has established the need for the identification of parameters altering radiomics reproducibility. The aim of this study was to assess the impact of magnetic field strength on magnetic resonance imaging (MRI) radiomics features in neuroradiology clinical practice. METHODS: T1 3D SPGR sequence was acquired on two phantoms and 10 healthy volunteers with two clinical MR devices from the same manufacturer using two different magnetic fields (1.5 and 3T). Phantoms varied in terms of gadolinium concentrations and textural heterogeneity. 27 regions of interest were segmented (phantom: 21, volunteers: 6) using the LIFEX software. 34 features were analyzed. RESULTS: In the phantom dataset, 10 (67%) out of 15 radiomics features were significantly different when measured at 1.5T or 3T (student's t-test, p < 0.05). Gray levels resampling, and pixel size also influence part of texture features. These findings were validated in healthy volunteers. CONCLUSIONS: According to daily used protocols for clinical examinations, radiomic features extracted on 1.5T should not be used interchangeably with 3T when evaluating texture features. Such confounding factor should be adjusted when adapting the results of a study to a different platform, or when designing a multicentric trial.

Dosimetry-Driven Quality Measure of Brain Pseudo Computed Tomography Generated From Deep Learning for MRI-Only Radiation Therapy Treatment Planning.

PURPOSE: This study aims to evaluate the impact of key parameters on the pseudo computed tomography (pCT) quality generated from magnetic resonance imaging (MRI) with a 3-dimensional (3D) convolutional neural network. METHODS AND MATERIALS: Four hundred two brain tumor cases were retrieved, yielding associations between 182 computed tomography (CT) and T1-weighted MRI (T1) scans, 180 CT and contrast-enhanced T1-weighted MRI (T1-Gd) scans, and 40 CT, T1, and T1-Gd scans. A 3D CNN was used to map T1 or T1-Gd onto CT scans and evaluate the importance of different components. First, the training set size's influence on testing set accuracy was assessed. Moreover, we evaluated the MRI sequence impact, using T1-only and T1-Gd-only cohorts. We then investigated 4 MRI standardization approaches (histogram-based, zero-mean/unit-variance, white stripe, and no standardization) based on training, validation, and testing cohorts composed of 242, 81, and 79 patients cases, respectively, as well as a bias field correction influence. Finally, 2 networks, namely HighResNet and 3D UNet, were compared to evaluate the architecture's impact on the pCT quality. The mean absolute error, gamma indices, and dose-volume histograms were used as evaluation metrics. RESULTS: Generating models using all the available cases for training led to higher pCT quality. The T1 and T1-Gd models had a maximum difference in gamma index means of 0.07 percentage point. The mean absolute error obtained with white stripe was 78 ± 22 Hounsfield units, which slightly outperformed histogram-based, zero-mean/unit-variance, and no standardization (P < .0001). Regarding the network architectures, 3%/3 mm gamma indices of 99.83% ± 0.19% and 99.74% ± 0.24% were obtained for HighResNet and 3D UNet, respectively. CONCLUSIONS: Our best pCTs were generated using more than 200 samples in the training data set. Training with T1 only and T1-Gd only did not significantly affect performance. Regardless of the preprocessing applied, the dosimetry quality remained equivalent and relevant for potential use in clinical practice.

The complexity of tumor shape, spiculatedness, correlates with tumor radiomic shape features.

Radiomics extracts high-throughput quantitative data from medical images to contribute to precision medicine. Radiomic shape features have been shown to correlate with patient outcomes. However, how radiomic shape features vary in function of tumor complexity and tumor volume, as well as with method used for meshing and voxel resampling, remains unknown. The aims of this study are to create tumor models with varying degrees of complexity, or spiculatedness, and evaluate their relationship with quantitatively extracted shape features. Twenty-eight tumor models were mathematically created using spherical harmonics with the spiculatedness degree d being increased by increments of 3 (d = 11 to d = 92). Models were 3D printed with identical bases of 5 cm, imaged with a CT scanner with two different slice thicknesses, and semi-automatically delineated. Resampling of the resulting masks on a 1 × 1 × 1 mm3 grid was performed, and the voxel size of each model was then calculated to eliminate volume differences. Four MATLAB-based algorithms (isosurface (M1), isosurface filter (M2), isosurface remeshing (M3), and boundary (M4)) were used to extract nine 3D features (Volume, Surface area, Surface-to-volume, Compactness1, Compactness2, Compactness3, Spherical Disproportion, Sphericity and Fractional Concavity). To quantify the impact of 3D printing, acquisition, segmentation and meshing, features were computed directly from the stereolithography (STL) file format that was used for 3D printing, and compared to those computed. Changes in feature values between 0.6 and 2 mm slice acquisitions were also compared. Spearman's rank-order correlation coefficients were computed to determine the relationship of each shape feature with spiculatedness for each of the four meshing algorithms. Percent changes were calculated between shape features extracted from the original and resampled contoured images to evaluate the influence of spatial resampling. Finally, the percent change in shape features when the volume was changed from 25% to 150% of their original volume was quantified for three distinct tumor models and compared to the percent change observed when modifying the spiculatedness of the model from d = 11 to d = 92. Values extracted using isosurface remeshing method are the closest to the STL reference ones, with mean differences less than 10.8% (Compactness2) for all features. Seven of the eight features had strong significant correlations with tumor model complexity irrespective of the meshing algorithm (r > 0.98, p < 10-4), with fractional concavity having the lowest correlation coefficient (r = 0.83, p < 10-4, M2). Comparisons of features extracted from the 0.6 and 2 mm slice thicknesses showed that mean differences were from 2.1% (Compactness3) to 12.7% (Compactness2) for the isosurface remeshing method. Resampling on a 1 × 1 × 1 mm3 grid resulted in between 1.3% (Compactness3) to 9.5% (Fractional Concavity) mean changes in feature values. Compactness2, Compactness3, Spherical Disproportion, Sphericity and Fractional Concavity were the features least affected by volume changes. Compactness1 had a 90.4% change with volume, which was greater than the change between the least and most spiculated models. This is the first methodological study that directly demonstrates the relationship of tumor spiculatedness with radiomic shape features, that also produced 3D tumor models, which may serve as reference phantoms for future radiomic studies. Surface Area, Surface-to-volume, and Spherical Disproportion had direct relationships with spiculatedness while the three formulas for Compactness, Sphericity and Fractional Concavity had inverse relationships. The features Compactness2, Compactness3, Spherical Disproportion, and Sphericity should be prioritized as these have minimal variations with volume changes, slice thickness and resampling.

Increased bone marrow SUVmax on 18F-FDG PET is associated with higher pelvic treatment failure in patients with cervical cancer treated by chemoradiotherapy and brachytherapy.

The aim of this study was to evaluate if bone marrow (BM) SUVmax measured on pre-treatment 18F-FDG PET/CT predicts the clinical outcome of locally advanced cervical cancer (LACC). We recruited retrospectively patients with LACC who underwent staging 18F-FDG PET/CT and had baseline blood tests, then treated by chemoradiation therapy (CRT), followed by image-guided adaptive brachytherapy (IGABT). BM SUVmax was calculated and correlated to inflammatory blood markers. Tumor size and pelvic lymph node involvement were evaluated on baseline MRI. Prognostic value of SUV uptake and blood markers regarding overall survival (OS), pelvic and extra-pelvic recurrence-free survival (PRFS and EPRFS respectively) was assessed using Cox models with adjusted p-values. 116 patients with FIGO stage Ib-IVa cervical cancer, treated between 2005 and 2014, were analyzed. The median follow-up was 75.5 months. BM SUVmax was significantly correlated to tumor SUVmax. In multivariate analysis, PRFS was significantly poorer in patients with high BM SUVmax (>2.8) and neutrophilia (p < .05). Tumor size (>5 vs ≤5 cm) could predict PRFS, EPRFS and OS (p < .05). In our cohort, FIGO stage (I-II vs III-IV), pelvic lymph node involvement and tumor SUVmax (>12 vs ≤12) were not prognostic for OS or pelvic and extra-pelvic relapses. Patients with LACC and high BM SUVmax on 18F-FDG PET have worse PFRS following CRT plus IGABT. These results can be potentially explained by the pro-inflammatory role of the tumor microenvironment and G-CSF expressed by tumor cells. These data support the role of PET as a potential indicator of disease aggressiveness beyond tumor staging.

Radiomics in Nuclear Medicine Applied to Radiation Therapy: Methods, Pitfalls, and Challenges.

Radiomics is a recent area of research in precision medicine and is based on the extraction of a large variety of features from medical images. In the field of radiation oncology, comprehensive image analysis is crucial to personalization of treatments. A better characterization of local heterogeneity and the shape of the tumor, depicting individual cancer aggressiveness, could guide dose planning and suggest volumes in which a higher dose is needed for better tumor control. In addition, noninvasive imaging features that could predict treatment outcome from baseline scans could help the radiation oncologist to determine the best treatment strategies and to stratify patients as at low risk or high risk of recurrence. Nuclear medicine molecular imaging reflects information regarding biological processes in the tumor thanks to a wide range of radiotracers. Many studies involving 18F-fluorodeoxyglucose positron emission tomography suggest an added value of radiomics compared with the use of conventional PET metrics such as standardized uptake value for both tumor diagnosis and prediction of recurrence or treatment outcome. However, these promising results should not hide technical difficulties that still currently prevent the approach from being widely studied or clinically used. These difficulties mostly pertain to the variability of the imaging features as a function of the acquisition device and protocol, the robustness of the models with respect to that variability, and the interpretation of the radiomic models. Addressing the impact of the variability in acquisition and reconstruction protocols is needed, as is harmonizing the radiomic feature calculation methods, to ensure the reproducibility of studies in a multicenter context and their implementation in a clinical workflow. In this review, we explain the potential impact of positron emission tomography radiomics for radiation therapy and underline the various aspects that need to be carefully addressed to make the most of this promising approach.

LIFEx: A Freeware for Radiomic Feature Calculation in Multimodality Imaging to Accelerate Advances in the Characterization of Tumor Heterogeneity.

Textural and shape analysis is gaining considerable interest in medical imaging, particularly to identify parameters characterizing tumor heterogeneity and to feed radiomic models. Here, we present a free, multiplatform, and easy-to-use freeware called LIFEx, which enables the calculation of conventional, histogram-based, textural, and shape features from PET, SPECT, MR, CT, and US images, or from any combination of imaging modalities. The application does not require any programming skills and was developed for medical imaging professionals. The goal is that independent and multicenter evidence of the usefulness and limitations of radiomic features for characterization of tumor heterogeneity and subsequent patient management can be gathered. Many options are offered for interactive textural index calculation and for increasing the reproducibility among centers. The software already benefits from a large user community (more than 800 registered users), and interactions within that community are part of the development strategy.Significance: This study presents a user-friendly, multi-platform freeware to extract radiomic features from PET, SPECT, MR, CT, and US images, or any combination of imaging modalities. Cancer Res; 78(16); 4786-9. ©2018 AACR.

Prediction of cervical cancer recurrence using textural features extracted from 18F-FDG PET images acquired with different scanners.

OBJECTIVES: To identify an imaging signature predicting local recurrence for locally advanced cervical cancer (LACC) treated by chemoradiation and brachytherapy from baseline 18F-FDG PET images, and to evaluate the possibility of gathering images from two different PET scanners in a radiomic study. METHODS: 118 patients were included retrospectively. Two groups (G1, G2) were defined according to the PET scanner used for image acquisition. Eleven radiomic features were extracted from delineated cervical tumors to evaluate: (i) the predictive value of features for local recurrence of LACC, (ii) their reproducibility as a function of the scanner within a hepatic reference volume, (iii) the impact of voxel size on feature values. RESULTS: Eight features were statistically significant predictors of local recurrence in G1 (p < 0.05). The multivariate signature trained in G2 was validated in G1 (AUC=0.76, p<0.001) and identified local recurrence more accurately than SUVmax (p=0.022). Four features were significantly different between G1 and G2 in the liver. Spatial resampling was not sufficient to explain the stratification effect. CONCLUSION: This study showed that radiomic features could predict local recurrence of LACC better than SUVmax. Further investigation is needed before applying a model designed using data from one PET scanner to another.