The STOIC2021 COVID-19 AI challenge: Applying reusable training methodologies to private data.

Challenges drive the state-of-the-art of automated medical image analysis. The quantity of public training data that they provide can limit the performance of their solutions. Public access to the training methodology for these solutions remains absent. This study implements the Type Three (T3) challenge format, which allows for training solutions on private data and guarantees reusable training methodologies. With T3, challenge organizers train a codebase provided by the participants on sequestered training data. T3 was implemented in the STOIC2021 challenge, with the goal of predicting from a computed tomography (CT) scan whether subjects had a severe COVID-19 infection, defined as intubation or death within one month. STOIC2021 consisted of a Qualification phase, where participants developed challenge solutions using 2000 publicly available CT scans, and a Final phase, where participants submitted their training methodologies with which solutions were trained on CT scans of 9724 subjects. The organizers successfully trained six of the eight Final phase submissions. The submitted codebases for training and running inference were released publicly. The winning solution obtained an area under the receiver operating characteristic curve for discerning between severe and non-severe COVID-19 of 0.815. The Final phase solutions of all finalists improved upon their Qualification phase solutions.

Translational Frontiers and Clinical Opportunities of Immunologically Fitted Radiotherapy.

Ionizing radiation can have a wide range of impacts on tumor-immune interactions, which are being studied with the greatest interest and at an accelerating pace by the medical community. Despite its undeniable immunostimulatory potential, it clearly appears that radiotherapy as it is prescribed and delivered nowadays often alters the host's immunity toward a suboptimal state. This may impair the full recovery of a sustained and efficient antitumor immunosurveillance posttreatment. An emerging concept is arising from this awareness and consists of reconsidering the way of designing radiation treatment planning, notably by taking into account the individualized risks of deleterious radio-induced immune alteration that can be deciphered from the planned beam trajectory through lymphocyte-rich organs. In this review, we critically appraise key aspects to consider while planning immunologically fitted radiotherapy, including the challenges linked to the identification of new dose constraints to immune-rich structures. We also discuss how pharmacologic immunomodulation could be advantageously used in combination with radiotherapy to compensate for the radio-induced loss, for example, with (i) agonists of interleukin (IL)2, IL4, IL7, IL9, IL15, or IL21, similarly to G-CSF being used for the prophylaxis of severe chemo-induced neutropenia, or with (ii) myeloid-derived suppressive cell blockers.

Standardizing and improving dose predictions for head and neck cancers using complete sets of OAR contours.

BACKGROUND: Radiotherapy dose predictions have been trained with data from previously treated patients of similar sites and prescriptions. However, clinical datasets are often inconsistent and do not contain the same number of organ at risk (OAR) structures. The effects of missing contour data in deep learning-based dose prediction models have not been studied. PURPOSE: The purpose of this study was to investigate the impacts of incomplete contour sets in the context of deep learning-based radiotherapy dose prediction models trained with clinical datasets and to introduce a novel data substitution method that utilizes automated contours for undefined structures. METHODS: We trained Standard U-Nets and Cascade U-Nets to predict the volumetric dose distributions of patients with head and neck cancers (HNC) using three input variations to evaluate the effects of missing contours, as well as a novel data substitution method. Each architecture was trained with the original contour (OC) inputs, which included missing information, hybrid contour (HC) inputs, where automated OAR contours generated in software were substituted for missing contour data, and automated contour (AC) inputs containing only automated OAR contours. 120 HNC treatments were used for model training, 30 were used for validation and tuning, and 44 were used for evaluation and testing. Model performance and accuracy were evaluated with global whole body dose agreement, PTV coverage accuracy, and OAR dose agreement. The differences in these values between dataset variations were used to determine the effects of missing data and automated contour substitutions. RESULTS: Automated contours used as substitutions for missing data were found to improve dose prediction accuracy in the Standard U-Net and Cascade U-Net, with a statistically significant difference in some global metrics and/or OAR metrics. For both models, PTV coverage between input variations was unaffected by the substitution technique. Automated contours in HC and AC datasets improved mean dose accuracy for some OAR contours, including the mandible and brainstem, with a greater improvement seen with HC datasets. Global dose metrics, including mean absolute error, mean error, and percent error were different for the Standard U-Net but not for the Cascade U-Net. CONCLUSION: Automated contours used as a substitution for contour data improved prediction accuracy for some but not all dose prediction metrics. Compared to the Standard U-Net models, the Cascade U-Net achieved greater precision.

A multi-centric evaluation of self-learning GAN based pseudo-CT generation software for low field pelvic magnetic resonance imaging.

Purpose/objectives: An artificial intelligence-based pseudo-CT from low-field MR images is proposed and clinically evaluated to unlock the full potential of MRI-guided adaptive radiotherapy for pelvic cancer care. Materials and method: In collaboration with TheraPanacea (TheraPanacea, Paris, France) a pseudo-CT AI-model was generated using end-to-end ensembled self-supervised GANs endowed with cycle consistency using data from 350 pairs of weakly aligned data of pelvis planning CTs and TrueFisp-(0.35T)MRIs. The image accuracy of the generated pCT were evaluated using a retrospective cohort involving 20 test cases coming from eight different institutions (US: 2, EU: 5, AS: 1) and different CT vendors. Reconstruction performance was assessed using the organs at risk used for treatment. Concerning the dosimetric evaluation, twenty-nine prostate cancer patients treated on the low field MR-Linac (ViewRay) at Montpellier Cancer Institute were selected. Planning CTs were non-rigidly registered to the MRIs for each patient. Treatment plans were optimized on the planning CT with a clinical TPS fulfilling all clinical criteria and recalculated on the warped CT (wCT) and the pCT. Three different algorithms were used: AAA, AcurosXB and MonteCarlo. Dose distributions were compared using the global gamma passing rates and dose metrics. Results: The observed average scaled (between maximum and minimum HU values of the CT) difference between the pCT and the planning CT was 33.20 with significant discrepancies across organs. Femoral heads were the most reliably reconstructed (4.51 and 4.77) while anal canal and rectum were the less precise ones (63.08 and 53.13). Mean gamma passing rates for 1%1mm, 2%/2mm, and 3%/3mm tolerance criteria and 10% threshold were greater than 96%, 99% and 99%, respectively, regardless the algorithm used. Dose metrics analysis showed a good agreement between the pCT and the wCT. The mean relative difference were within 1% for the target volumes (CTV and PTV) and 2% for the OARs. Conclusion: This study demonstrated the feasibility of generating clinically acceptable an artificial intelligence-based pseudo CT for low field MR in pelvis with consistent image accuracy and dosimetric results.

The use of image synthesis techniques in target and roi delineation in the upright position.

The use of multi-modality imaging technologies such as CT, MRI, and PET imaging is state of the art for radiation therapy treatment planning. Except for a limited number of low magnetic field MR scanners the majority of such imaging technologies can only image the patient in a recumbent position. Delivering radiation therapy treatments with the patient in an upright orientation has many benefits and several companies are now developing upright patient positioners combined with upright diagnostic helical CT scanners to facilitate upright radiation therapy treatments. Due to the directional changes in the gravitational forces on the patient's body, most structures and organs will change position and shape between the recumbent and upright positions. Detailed knowledge about such structures and organs are therefore often only available in the recumbent position. The problem statement is therefore well defined, that is, how do we know where such structures and organs, that is, the target or region at risk volumes, are in the upright position if those cannot be identified and or delineated accurately enough using the upright diagnostic quality CT images only? This paper outlines two methods based on synthetic CT or MR images to overcome this problem.

Automatic AI-based contouring of prostate MRI for online adaptive radiotherapy.

BACKGROUND AND PURPOSE: MR-guided radiotherapy (MRgRT) online plan adaptation accounts for tumor volume changes, interfraction motion and thus allows daily sparing of relevant organs at risk. Due to the high interfraction variability of bladder and rectum, patients with tumors in the pelvic region may strongly benefit from adaptive MRgRT. Currently, fast automatic annotation of anatomical structures is not available within the online MRgRT workflow. Therefore, the aim of this study was to train and validate a fast, accurate deep learning model for automatic MRI segmentation at the MR-Linac for future implementation in a clinical MRgRT workflow. MATERIALS AND METHODS: For a total of 47 patients, T2w MRI data were acquired on a 1.5 T MR-Linac (Unity, Elekta) on five different days. Prostate, seminal vesicles, rectum, anal canal, bladder, penile bulb, body and bony structures were manually annotated. These training data consisting of 232 data sets in total was used for the generation of a deep learning based autocontouring model and validated on 20 unseen T2w-MRIs. For quantitative evaluation the validation set was contoured by a radiation oncologist as gold standard contours (GSC) and compared in MATLAB to the automatic contours (AIC). For the evaluation, dice similarity coefficients (DSC), and 95% Hausdorff distances (95% HD), added path length (APL) and surface DSC (sDSC) were calculated in a caudal-cranial window of ± 4 cm with respect to the prostate ends. For qualitative evaluation, five radiation oncologists scored the AIC on the possible usage within an online adaptive workflow as follows: (1) no modifications needed, (2) minor adjustments needed, (3) major adjustments/ multiple minor adjustments needed, (4) not usable. RESULTS: The quantitative evaluation revealed a maximum median 95% HD of 6.9 mm for the rectum and minimum median 95% HD of 2.7 mm for the bladder. Maximal and minimal median DSC were detected for bladder with 0.97 and for penile bulb with 0.73, respectively. Using a tolerance level of 3 mm, the highest and lowest sDSC were determined for rectum (0.94) and anal canal (0.68), respectively. Qualitative evaluation resulted in a mean score of 1.2 for AICs over all organs and patients across all expert ratings. For the different autocontoured structures, the highest mean score of 1.0 was observed for anal canal, sacrum, femur left and right, and pelvis left, whereas for prostate the lowest mean score of 2.0 was detected. In total, 80% of the contours were rated be clinically acceptable, 16% to require minor and 4% major adjustments for online adaptive MRgRT. CONCLUSION: In this study, an AI-based autocontouring was successfully trained for online adaptive MR-guided radiotherapy on the 1.5 T MR-Linac system. The developed model can automatically generate contours accepted by physicians (80%) or only with the need of minor corrections (16%) for the irradiation of primary prostate on the clinically employed sequences.

Single-timepoint low-dimensional characterization and classification of acute versus chronic multiple sclerosis lesions using machine learning.

Multiple sclerosis (MS) is a chronic inflammatory and neurodegenerative disease characterized by the appearance of focal lesions across the central nervous system. The discrimination of acute from chronic MS lesions may yield novel biomarkers of inflammatory disease activity which may support patient management in the clinical setting and provide endpoints in clinical trials. On a single timepoint and in the absence of a prior reference scan, existing methods for acute lesion detection rely on the segmentation of hyperintense foci on post-gadolinium T1-weighted magnetic resonance imaging (MRI), which may underestimate recent acute lesion activity. In this paper, we aim to improve the sensitivity of acute MS lesion detection in the single-timepoint setting, by developing a novel machine learning approach for the automatic detection of acute MS lesions, using single-timepoint conventional non-contrast T1- and T2-weighted brain MRI. The MRI input data are supplemented via the use of a convolutional neural network generating "lesion-free" reconstructions from original "lesion-present" scans using image inpainting. A multi-objective statistical ranking module evaluates the relevance of textural radiomic features from the core and periphery of lesion sites, compared within "lesion-free" versus "lesion-present" image pairs. Then, an ensemble classifier is optimized through a recursive loop seeking consensus both in the feature space (via a greedy feature-pruning approach) and in the classifier space (via model selection repeated after each pruning operation). This leads to the identification of a compact textural signature characterizing lesion phenotype. On the patch-level task of acute versus chronic MS lesion classification, our method achieves a balanced accuracy in the range of 74.3-74.6% on fully external validation cohorts.

Joint Deformable Image Registration and ADC Map Regularization: Application to DWI-Based Lymphoma Classification.

The Apparent Diffusion Coefficient (ADC) is considered an importantimaging biomarker contributing to the assessment of tissue microstructure and pathophy- siology. It is calculated from Diffusion-Weighted Magnetic Resonance Imaging (DWI) by means of a diffusion model, usually without considering any motion during image acquisition. We propose a method to improve the computation of the ADC by coping jointly with both motion artifacts in whole-body DWI (through group-wise registration) and possible instrumental noise in the diffusion model. The proposed deformable registration method yielded on average the lowest ADC reconstruction error on data with simulated motion and diffusion. Moreover, our approach was applied on whole-body diffusion weighted images obtained with five different b-values from a cohort of 38 patients with histologically confirmed lymphomas of three different types (Hodgkin, diffuse large B-cell lymphoma and follicular lymphoma). Evaluation on the real data showed that ADC-based features, extracted using our joint optimization approach classified lymphomas with an accuracy of approximately 78.6% (yielding a 11% increase in respect to the standard features extracted from unregistered diffusion-weighted images). Furthermore, the correlation between diffusion characteristics and histopathological findings was higher than any other previous approach of ADC computation.

COMBING: Clustering in Oncology for Mathematical and Biological Identification of Novel Gene Signatures.

Precision medicine is a paradigm shift in healthcare relying heavily on genomics data. However, the complexity of biological interactions, the large number of genes as well as the lack of comparisons on the analysis of data, remain a tremendous bottleneck regarding clinical adoption. In this paper, we introduce a novel, automatic and unsupervised framework to discover low-dimensional gene biomarkers. Our method is based on the LP-Stability algorithm, a high dimensional center-based unsupervised clustering algorithm. It offers modularity as concerns metric functions and scalability, while being able to automatically determine the best number of clusters. Our evaluation includes both mathematical and biological criteria to define a quantitative metric. The recovered signature is applied to a variety of biological tasks, including screening of biological pathways and functions, and characterization relevance on tumor types and subtypes. Quantitative comparisons among different distance metrics, commonly used clustering methods and a referential gene signature used in the literature, confirm state of the art performance of our approach. In particular, our signature, based on 27 genes, reports at least 30 times better mathematical significance (average Dunn's Index) and 25% better biological significance (average Enrichment in Protein-Protein Interaction) than those produced by other referential clustering methods. Finally, our signature reports promising results on distinguishing immune inflammatory and immune desert tumors, while reporting a high balanced accuracy of 92% on tumor types classification and averaged balanced accuracy of 68% on tumor subtypes classification, which represents, respectively 7% and 9% higher performance compared to the referential signature.

Deep learning for lung disease segmentation on CT: Which reconstruction kernel should be used?

PURPOSE: The purpose of this study was to determine whether a single reconstruction kernel or both high and low frequency kernels should be used for training deep learning models for the segmentation of diffuse lung disease on chest computed tomography (CT). MATERIALS AND METHODS: Two annotated datasets of COVID-19 pneumonia (323,960 slices) and interstitial lung disease (ILD) (4,284 slices) were used. Annotated CT images were used to train a U-Net architecture to segment disease. All CT slices were reconstructed using both a lung kernel (LK) and a mediastinal kernel (MK). Three different trainings, resulting in three different models were compared for each disease: training on LK only, MK only or LK+MK images. Dice similarity scores (DSC) were compared using the Wilcoxon signed-rank test. RESULTS: Models only trained on LK images performed better on LK images than on MK images (median DSC = 0.62 [interquartile range (IQR): 0.54, 0.69] vs. 0.60 [IQR: 0.50, 0.70], P < 0.001 for COVID-19 and median DSC = 0.62 [IQR: 0.56, 0.69] vs. 0.50 [IQR 0.43, 0.57], P < 0.001 for ILD). Similarly, models only trained on MK images performed better on MK images (median DSC = 0.62 [IQR: 0.53, 0.68] vs. 0.54 [IQR: 0.47, 0.63], P < 0.001 for COVID-19 and 0.69 [IQR: 0.61, 0.73] vs. 0.63 [IQR: 0.53, 0.70], P < 0.001 for ILD). Models trained on both kernels performed better or similarly than those trained on only one kernel. For COVID-19, median DSC was 0.67 (IQR: =0.59, 0.73) when applied on LK images and 0.67 (IQR: 0.60, 0.74) when applied on MK images (P < 0.001 for both). For ILD, median DSC was 0.69 (IQR: 0.63, 0.73) when applied on LK images (P = 0.006) and 0.68 (IQR: 0.62, 0.72) when applied on MK images (P > 0.99). CONCLUSION: Reconstruction kernels impact the performance of deep learning-based models for lung disease segmentation. Training on both LK and MK images improves the performance.

Deep Multi-Instance Learning Using Multi-Modal Data for Diagnosis of Lymphocytosis.

We investigate the use of recent advances in deep learning and propose an end-to-end trainable multi-instance convolutional neural network within a mixture-of-experts formulation that combines information from two types of data-images and clinical attributes-for the diagnosis of lymphocytosis. The convolutional network learns to extract meaningful features from images of blood cells using an embedding level approach and aggregates them. Moreover, the mixture-of-experts model combines information from these images as well as clinical attributes to form an end-to-end trainable pipeline for diagnosis of lymphocytosis. Our results demonstrate that even the convolutional network by itself is able to discover meaningful associations between the images and the diagnosis, indicating the presence of important unexploited information in the images. The mixture-of-experts formulation is shown to be more robust while maintaining performance via. a repeatability study to assess the effect of variability in data acquisition on the predictions. The proposed methods are compared with different methods from literature based both on conventional handcrafted features and machine learning, and on recent deep learning models based on attention mechanisms. Our method reports a balanced accuracy of [Formula: see text] and outperfroms the handcrafted feature-based and attention-based approaches as well that of biologists which scored [Formula: see text], [Formula: see text] and [Formula: see text] respectively. These results give insights on the potentials of the applicability of the proposed method in clinical practice. Our code and datasets can be found at https://github.com/msahasrabudhe/lymphoMIL.

Reinventing radiation therapy with machine learning and imaging bio-markers (radiomics): State-of-the-art, challenges and perspectives.

Radiation therapy is a pivotal cancer treatment that has significantly progressed over the last decade due to numerous technological breakthroughs. Imaging is now playing a critical role on deployment of the clinical workflow, both for treatment planning and treatment delivery. Machine-learning analysis of predefined features extracted from medical images, i.e. radiomics, has emerged as a promising clinical tool for a wide range of clinical problems addressing drug development, clinical diagnosis, treatment selection and implementation as well as prognosis. Radiomics denotes a paradigm shift redefining medical images as a quantitative asset for data-driven precision medicine. The adoption of machine-learning in a clinical setting and in particular of radiomics features requires the selection of robust, representative and clinically interpretable biomarkers that are properly evaluated on a representative clinical data set. To be clinically relevant, radiomics must not only improve patients' management with great accuracy but also be reproducible and generalizable. Hence, this review explores the existing literature and exposes its potential technical caveats, such as the lack of quality control, standardization, sufficient sample size, type of data collection, and external validation. Based upon the analysis of 165 original research studies based on PET, CT-scan, and MRI, this review provides an overview of new concepts, and hypotheses generating findings that should be validated. In particular, it describes evolving research trends to enhance several clinical tasks such as prognostication, treatment planning, response assessment, prediction of recurrence/relapse, and prediction of toxicity. Perspectives regarding the implementation of an AI-based radiotherapy workflow are presented.

AI-driven quantification, staging and outcome prediction of COVID-19 pneumonia.

Coronavirus disease 2019 (COVID-19) emerged in 2019 and disseminated around the world rapidly. Computed tomography (CT) imaging has been proven to be an important tool for screening, disease quantification and staging. The latter is of extreme importance for organizational anticipation (availability of intensive care unit beds, patient management planning) as well as to accelerate drug development through rapid, reproducible and quantified assessment of treatment response. Even if currently there are no specific guidelines for the staging of the patients, CT together with some clinical and biological biomarkers are used. In this study, we collected a multi-center cohort and we investigated the use of medical imaging and artificial intelligence for disease quantification, staging and outcome prediction. Our approach relies on automatic deep learning-based disease quantification using an ensemble of architectures, and a data-driven consensus for the staging and outcome prediction of the patients fusing imaging biomarkers with clinical and biological attributes. Highly promising results on multiple external/independent evaluation cohorts as well as comparisons with expert human readers demonstrate the potentials of our approach.

Elastic Registration-driven Deep Learning for Longitudinal Assessment of Systemic Sclerosis Interstitial Lung Disease at CT.

Background Longitudinal follow-up of interstitial lung diseases (ILDs) at CT mainly relies on the evaluation of the extent of ILD, without accounting for lung shrinkage. Purpose To develop a deep learning-based method to depict worsening of ILD based on lung shrinkage detection from elastic registration of chest CT scans in patients with systemic sclerosis (SSc). Materials and Methods Patients with SSc evaluated between January 2009 and October 2017 who had undergone at least two unenhanced supine CT scans of the chest and pulmonary function tests (PFTs) performed within 3 months were retrospectively included. Morphologic changes on CT scans were visually assessed by two observers and categorized as showing improvement, stability, or worsening of ILD. Elastic registration between baseline and follow-up CT images was performed to obtain deformation maps of the whole lung. Jacobian determinants calculated from the deformation maps were given as input to a deep learning-based classifier to depict morphologic and functional worsening. For this purpose, the set was randomly split into training, validation, and test sets. Correlations between mean Jacobian values and changes in PFT measurements were evaluated with the Spearman correlation. Results A total of 212 patients (median age, 53 years; interquartile range, 45-62 years; 177 women) were included as follows: 138 for the training set (65%), 34 for the validation set (16%), and 40 for the test set (21%). Jacobian maps demonstrated lung parenchyma shrinkage of the posterior lung bases in patients found to have worsened ILD at visual assessment. The classifier detected morphologic and functional worsening with an accuracy of 80% (32 of 40 patients; 95% confidence interval [CI]: 64%, 91%) and 83% (33 of 40 patients; 95% CI: 67%, 93%), respectively. Jacobian values correlated with changes in forced vital capacity (R = -0.38; 95% CI: -0.25, -0.49; P < .001) and diffusing capacity for carbon monoxide (R = -0.42; 95% CI: -0.27, -0.54; P < .001). Conclusion Elastic registration of CT scans combined with a deep learning classifier aided in the diagnosis of morphologic and functional worsening of interstitial lung disease in patients with systemic sclerosis. © RSNA, 2020 Online supplemental material is available for this article. See also the editorial by Verschakelen in this issue.

Radiomics to predict outcomes and abscopal response of patients with cancer treated with immunotherapy combined with radiotherapy using a validated signature of CD8 cells.

BACKGROUND: Combining radiotherapy (RT) with immuno-oncology (IO) therapy (IORT) may enhance IO-induced antitumor response. Quantitative imaging biomarkers can be used to provide prognosis, predict tumor response in a non-invasive fashion and improve patient selection for IORT. A biologically inspired CD8 T-cells-associated radiomics signature has been developed on previous cohorts. We evaluated here whether this CD8 radiomic signature is associated with lesion response, whether it may help to assess disease spatial heterogeneity for predicting outcomes of patients treated with IORT. We also evaluated differences between irradiated and non-irradiated lesions. METHODS: Clinical data from patients with advanced solid tumors in six independent clinical studies of IORT were investigated. Immunotherapy consisted of 4 different drugs (antiprogrammed death-ligand 1 or anticytotoxic T-lymphocyte-associated protein 4 in monotherapy). Most patients received stereotactic RT to one lesion. Irradiated and non-irradiated lesions were delineated from baseline and the first evaluation CT scans. Radiomic features were extracted from contrast-enhanced CT images and the CD8 radiomics signature was applied. A responding lesion was defined by a decrease in lesion size of at least 30%. Dispersion metrices of the radiomics signature were estimated to evaluate the impact of tumor heterogeneity in patient's response. RESULTS: A total of 94 patients involving multiple lesions (100 irradiated and 189 non-irradiated lesions) were considered for a statistical interpretation. Lesions with high CD8 radiomics score at baseline were associated with significantly higher tumor response (area under the receiving operating characteristic curve (AUC)=0.63, p=0.0020). Entropy of the radiomics scores distribution on all lesions was shown to be associated with progression-free survival (HR=1.67, p=0.040), out-of-field abscopal response (AUC=0.70, p=0.014) and overall survival (HR=2.08, p=0.023), which remained significant in a multivariate analysis including clinical and biological variables. CONCLUSIONS: These results enhance the predictive value of the biologically inspired CD8 radiomics score and suggests that tumor heterogeneity should be systematically considered in patients treated with IORT. This CD8 radiomics signature may help select patients who are most likely to benefit from IORT.

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.

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.

Deep Learning-Based Concurrent Brain Registration and Tumor Segmentation.

Image registration and segmentation are the two most studied problems in medical image analysis. Deep learning algorithms have recently gained a lot of attention due to their success and state-of-the-art results in variety of problems and communities. In this paper, we propose a novel, efficient, and multi-task algorithm that addresses the problems of image registration and brain tumor segmentation jointly. Our method exploits the dependencies between these tasks through a natural coupling of their interdependencies during inference. In particular, the similarity constraints are relaxed within the tumor regions using an efficient and relatively simple formulation. We evaluated the performance of our formulation both quantitatively and qualitatively for registration and segmentation problems on two publicly available datasets (BraTS 2018 and OASIS 3), reporting competitive results with other recent state-of-the-art methods. Moreover, our proposed framework reports significant amelioration (p < 0.005) for the registration performance inside the tumor locations, providing a generic method that does not need any predefined conditions (e.g., absence of abnormalities) about the volumes to be registered. Our implementation is publicly available online at https://github.com/TheoEst/joint_registration_tumor_segmentation.

Deep Learning-based Approach for Automated Assessment of Interstitial Lung Disease in Systemic Sclerosis on CT Images.

PURPOSE: To develop a deep learning algorithm for the automatic assessment of the extent of systemic sclerosis (SSc)-related interstitial lung disease (ILD) on chest CT images. MATERIALS AND METHODS: This retrospective study included 208 patients with SSc (median age, 57 years; 167 women) evaluated between January 2009 and October 2017. A multicomponent deep neural network (AtlasNet) was trained on 6888 fully annotated CT images (80% for training and 20% for validation) from 17 patients with no, mild, or severe lung disease. The model was tested on a dataset of 400 images from another 20 patients, independently partially annotated by three radiologist readers. The ILD contours from the three readers and the deep learning neural network were compared by using the Dice similarity coefficient (DSC). The correlation between disease extent obtained from the deep learning algorithm and that obtained by using pulmonary function tests (PFTs) was then evaluated in the remaining 171 patients and in an external validation dataset of 31 patients based on the analysis of all slices of the chest CT scan. The Spearman rank correlation coefficient (ρ) was calculated to evaluate the correlation between disease extent and PFT results. RESULTS: The median DSCs between the readers and the deep learning ILD contours ranged from 0.74 to 0.75, whereas the median DSCs between contours from radiologists ranged from 0.68 to 0.71. The disease extent obtained from the algorithm, by analyzing the whole CT scan, correlated with the diffusion lung capacity for carbon monoxide, total lung capacity, and forced vital capacity (ρ = -0.76, -0.70, and -0.62, respectively; P < .001 for all) in the dataset for the correlation with PFT results. The disease extents correlated with diffusion lung capacity for carbon monoxide, total lung capacity, and forced vital capacity were ρ = -0.65, -0.70, and -0.57, respectively, in the external validation dataset (P < .001 for all). CONCLUSION: The developed algorithm performed similarly to radiologists for disease-extent contouring, which correlated with pulmonary function to assess CT images from patients with SSc-related ILD.Supplemental material is available for this article.© RSNA, 2020.

Quantification of Cystic Fibrosis Lung Disease with Radiomics-based CT Scores.

PURPOSE: To develop radiomics-based CT scores for assessing lung disease severity and exacerbation risk in adult patients with cystic fibrosis (CF). MATERIALS AND METHODS: This two-center retrospective observational study was approved by an institutional ethics committee, and the need for patient consent was waived. A total of 215 outpatients with CF referred for unenhanced follow-up chest CT were evaluated in two different centers between January 2013 and December 2016. After lung segmentation, chest CT scans from center 1 (training cohort, 162 patients [median age, 29 years; interquartile range {IQR}, 24-36 years; 84 men]) were used to build CT scores from 38 extracted CT features, using five different machine learning techniques trained to predict a clinical prognostic score, the Nkam score. The correlations between the developed CT scores, two different clinical prognostic scores (Liou and CF-ABLE), forced expiratory volume in 1 second (FEV1), and risk of respiratory exacerbations were evaluated in the test cohort (center 2, 53 patients [median age, 27 years; IQR, 22-35 years; 34 men]) using the Spearman rank coefficient. RESULTS: In the test cohort, all radiomics-based CT scores showed moderate to strong correlation with the Nkam score (R = 0.57 to 0.63, P < .001) and Liou scores (R = -0.55 to -0.65, P < .001), whereas the correlation with CF-ABLE score was weaker (R = 0.28 to 0.38, P = .005 to .048). The developed CT scores showed strong correlation with predicted FEV1 (R = -0.62 to -0.66, P < .001) and weak to moderate correlation with the number of pulmonary exacerbations to occur in the 12 months after the CT examination (R = 0.38 to 0.55, P < .001 to P = .006). CONCLUSION: Radiomics can be used to build automated CT scores that correlate to clinical severity and exacerbation risk in adult patients with CF.Supplemental material is available for this article.See also the commentary by Elicker and Sohn in this issue.© RSNA, 2020.