Multiple Magnetic Resonance Imaging Sequences Research Articles

A geometric approach to robust medical image segmentation

Robustness of deep learning segmentation models is crucial for their safe incorporation into clinical practice. However, these models can falter when faced with distributional changes. This challenge is evident in magnetic resonance imaging (MRI) scans due to the diverse acquisition protocols across various domains, leading to differences in image characteristics such as textural appearances. We posit that the restricted anatomical differences between subjects could be harnessed to refine the latent space into a set of shape components. The learned set then aims to encompass the relevant anatomical shape variation found within the patient population.We explore this by utilising multiple MRI sequences to learn texture invariant and shape equivariant features which are used to construct a shape dictionary using vector quantisation. We investigate shape equivariance to a number of different types of groups. We hypothesise and prove that the greater the group order, i.e., the denser the constraint, the better becomes the model robustness. We achieve shape equivariance either with a contrastive based approach or by imposing equivariant constraints on the convolutional kernels. The resulting shape equivariant dictionary is then sampled to compose the segmentation output. Our method achieves state-of-the-art performance for the task of single domain generalisation for prostate and cardiac MRI segmentation. Code is available at https://github.com/AinkaranSanthi/A_Geometric_Perspective_For_Robust_Segmentation.

Convolutional neural network and glioma genotyping: A systematic review.

e14038 Background: Gliomas are the most common primary intracranial tumors. Overall, there are 23 different glioma types, and the grading and prognosis of each type are determined by their specific mutations. Several identified mutations are IDH gene status, CDKN2A/B homozygote delegation, PTEN mutation, p53 mutation, TERT promotor mutation, H3K27M mutation, MGMT promotor methylation, trisomy 7/monozomy10, EGFR amplification, ATRX mutation, 1p/19q codeletion, and BRAF mutation. These specific mutations and genomic features can have their distinctive radiologic features seen in different magnetic resonance imaging (MRI) sequences. By using a Convolutional Neural Network (CNN) for feature extraction, the genotyping power of MRI for glioma could be increased. To address this, we systematically reviewed the use of different CNN-based models to detect the genetic context of glioma based on the MRI images. Methods: We conducted a systematic search in PubMed, Embase, Scopus, Web of Science, and Google Scholar. We included original peer-reviewed human studies (retrospective or prospective), focusing on the use of CNN in individuals with glioma. We excluded animal studies, studies with no CNN-based models, reviews, book chapters, conference abstracts, and case reports. JBI checklists were utilized for quality assessment. Results: The initial search resulted in 973 articles. after removing duplicates, 563 studies entered the screening process by title and abstract. 59 studies met our inclusion criteria (56 retrospective and 3 prospective cohort studies) with a total of 26434 glioma patients included (studies ranged from 21 – 2648 individuals). 31 studies used CNN methods only in different stages of image processing (tumor segmentation, feature extraction, image encoder) and 28 studies used CNN as a classifier and mutation predictor. ResNet was the most commonly used CNN-based model (n = 14). U-net (n = 7) and DenseNet models (n = 4) were other commonly used CNN-based models. Included studies targeted IDH status (n=44), 1p/19q codeletion (n=18), MGMT promotor methylation (n=11), ATRX mutation (n=5), TERT promotor mutation (n=5), CDKN2A/B homozygote deletion (n=3), PTEN mutation (n=3), p53 mutation (n=3), H3K27M mutation (n=2), trisomy 7/monozomy10 (n=2), EGFR amplification (n=2), and BRAF mutation (n=1). Highly severe methodological heterogeneity among these studies excluded the possibility of meta-analysis. Conclusions: Current studies have shown that, as a non-invasive method, using CNN-based models on multiple MRI sequences increases the power of early detection and prediction of glioma genetic types before pathological confirmation and has promising results. However, DL performance varies among mutations. Combining multiple DL models with clinical and radiomic features in different and larger databases can further increase the final results.

A discrepancy-aware self-distillation method for multi-modal glioma grading

Gliomas are the most common intracranial primary tumors. Accurate grading of gliomas is crucial in determining treatment options and prognosis. Clinicians conventionally rely on multiple Magnetic Resonance Imaging (MRI) sequences for accurate glioma assessment. Traditional deep learning methods typically utilize the individual MRI sequence with the annotations of Region of Interest (ROIs), incurring substantial manual effort while losing the complementary information provided by other sequences. This task is inherently a standard multi-modal problem. However, existing methods often adopt complex multi-stream networks for feature extraction and fusion, leading to high resource demands and potential feature redundancy. To address the above challenge, a discrepancy-aware self-distillation method is proposed for multi-modal glioma grading, which requires only a single-stream network to concurrently analyze multiple MRI sequences without auxiliary ROIs. The first Modality Discrepancy-aware Fusion (MDF) module considers the MRI imaging differentiation and widens the inter-modal contrasts, allowing the modality-specific features to be highlighted and the modality-invariant features to be diminished. Furthermore, the proposed Class Activation Self-Distillation (CASD) strategy leverages the generated Class Activation Maps (CAMs) as dark knowledge for the distillation process. This guides shallow layers to focus on the category discriminative features specifically within the lesion region. Extensive experiments were conducted on the BraTS2018 and BraTS2019 datasets to evaluate the effectiveness of our method. The results show that our method outperforms other relevant glioma grading and multi-modal fusion methods on both datasets.

Cancer Radiomic and Perfusion Imaging Automated Framework: Validation on Musculoskeletal Tumors.

Limitations from commercial software applications prevent the implementation of a robust and cost-efficient high-throughput cancer imaging radiomic feature extraction and perfusion analysis workflow. This study aimed to develop and validate a cancer research computational solution using open-source software for vendor- and sequence-neutral high-throughput image processing and feature extraction. The Cancer Radiomic and Perfusion Imaging (CARPI) automated framework is a Python-based software application that is vendor- and sequence-neutral. CARPI uses contour files generated using an application of the user's choice and performs automated radiomic feature extraction and perfusion analysis. This workflow solution was validated using two clinical data sets, one consisted of 40 pelvic chondrosarcomas and 42 sacral chordomas with a total of 82 patients, and a second data set consisted of 26 patients with undifferentiated pleomorphic sarcoma (UPS) imaged at multiple points during presurgical treatment. Three hundred sixteen volumetric contour files were processed using CARPI. The application automatically extracted 107 radiomic features from multiple magnetic resonance imaging sequences and seven semiquantitative perfusion parameters from time-intensity curves. Statistically significant differences (P < .00047) were found in 18 of 107 radiomic features in chordoma versus chondrosarcoma, including six first-order and 12 high-order features. In UPS postradiation, the apparent diffusion coefficient mean increased 41% in good responders (P = .0017), while firstorder_10Percentile (P = .0312) was statistically significant between good and partial/nonresponders. The CARPI processing of two clinical validation data sets confirmed the software application's ability to differentiate between different types of tumors and help predict patient response to treatment on the basis of radiomic features. Benchmark comparison with five similar open-source solutions demonstrated the advantages of CARPI in the automated perfusion feature extraction, relational database generation, and graphic report export features, although lacking a user-friendly graphical user interface and predictive model building.

Comparative evaluation of multiparametric lumbar MRI radiomic models for detecting osteoporosis

BackgroundOsteoporosis is a serious global public health issue. Currently, there are few studies that explore the use of multiparametric MRI radiomics for osteoporosis detection. The purpose of this study was to compare the performance of radiomics features from multiple MRI sequences (T1WI, T2WI and T1WI combined with T2WI) for detecting osteoporosis in patients.MethodsA retrospective analysis was performed on 160 patients who had undergone dual-energy X-ray absorptiometry(DXA) and lumbar magnetic resonance imaging (MRI) at our hospital. Among them, 86 patients were diagnosed with abnormal bone mass (osteoporosis or low bone mass), and 74 patients were diagnosed with normal bone mass based on the DXA results. Sagittal T1-and T2-weighted images of all patients were imported into the uAI Research Portal (United Imaging Intelligence) for image delineation and radiomics analysis, where a series of radiomic features were obtained. A radiomic model that included T1WI, T2WI, and T1WI+T2WI was established using features selected by LASSO regression. We used ROC curve analysis to evaluate the predictive efficacy of each model for identifying bone abnormalities and conducted decision curve analysis (DCA) to evaluate the net benefit of each model. Finally, we validated the model in a sample of 35 patients from different health care institution.ResultsThe T1WI + T2WI radiomics model showed better screening performance for patients with abnormal bone mass. In the training group, the sensitivity was 0.758, the specificity was 0.78, and the accuracy was 0.768 (AUC =0.839, 95% CI=0.757-0.901). In the validation group, the sensitivity was 0.792, the specificity was 0.875, and the accuracy was 0.833 (AUC =0.86, 95% CI=0.73-0.943).The DCA also showed that the combined model had better net benefits. In the external validation group, the sensitivity was 0.764, the specificity was 0.833, and the accuracy was 0.8 (AUC =0.824, 95% CI 0.678-0.969).ConclusionsRadiomics-based multiparametric MRI can be used for the quantitative analysis of lumbar MRI and for accurately screening patients with abnormal bone mass.

The PREMISE database of 20 Macaca fascicularis PET/MRI brain images available for research

Non-human primate studies are unique in translational research, especially in neurosciences where neuroimaging approaches are the preferred methods used for cross-species comparative neurosciences. In this regard, neuroimaging database development and sharing are encouraged to increase the number of subjects available to the community, while limiting the number of animals used in research. Here we present a simultaneous positron emission tomography (PET)/magnetic resonance (MR) dataset of 20 Macaca fascicularis images structured according to the Brain Imaging Data Structure standards. This database contains multiple MR imaging sequences (anatomical, diffusion and perfusion imaging notably), as well as PET perfusion and inflammation imaging using respectively [15O]H2O and [11C]PK11195 radiotracers. We describe the pipeline method to assemble baseline data from various cohorts and qualitatively assess all the data using signal-to-noise and contrast-to-noise ratios as well as the median of intensity and the pseudo-noise-equivalent-count rate (dynamic and at maximum) for PET data. Our study provides a detailed example for quality control integration in preclinical and translational PET/MR studies with the aim of increasing reproducibility. The PREMISE database is stored and available through the PRIME-DE consortium repository.

Identifying core MRI sequences for reliable automatic brain metastasis segmentation

BackgroundMany automatic approaches to brain tumor segmentation employ multiple magnetic resonance imaging (MRI) sequences. The goal of this project was to compare different combinations of input sequences to determine which MRI sequences are needed for effective automated brain metastasis (BM) segmentation. MethodsWe analyzed preoperative imaging (T1-weighted sequence ± contrast-enhancement (T1/T1-CE), T2-weighted sequence (T2), and T2 fluid-attenuated inversion recovery (T2-FLAIR) sequence) from 339 patients with BMs from seven centers. A baseline 3D U-Net with all four sequences and six U-Nets with plausible sequence combinations (T1-CE, T1, T2-FLAIR, T1-CE + T2-FLAIR, T1-CE + T1 + T2-FLAIR, T1-CE + T1) were trained on 239 patients from two centers and subsequently tested on an external cohort of 100 patients from five centers. ResultsThe model based on T1-CE alone achieved the best segmentation performance for BM segmentation with a median Dice similarity coefficient (DSC) of 0.96. Models trained without T1-CE performed worse (T1-only: DSC = 0.70 and T2-FLAIR-only: DSC = 0.73). For edema segmentation, models that included both T1-CE and T2-FLAIR performed best (DSC = 0.93), while the remaining four models without simultaneous inclusion of these both sequences reached a median DSC of 0.81–0.89. ConclusionsA T1-CE-only protocol suffices for the segmentation of BMs. The combination of T1-CE and T2-FLAIR is important for edema segmentation. Missing either T1-CE or T2-FLAIR decreases performance. These findings may improve imaging routines by omitting unnecessary sequences, thus allowing for faster procedures in daily clinical practice while enabling optimal neural network-based target definitions.

Ensemble learning-based radiomics with multi-sequence magnetic resonance imaging for benign and malignant soft tissue tumor differentiation.

Many previous studies focused on differentiating between benign and malignant soft tissue tumors using radiomics model based on various magnetic resonance imaging (MRI) sequences, but it is still unclear how to set up the input radiomic features from multiple MRI sequences. Here, we evaluated two types of radiomics models generated using different feature incorporation strategies. In order to differentiate between benign and malignant soft tissue tumors (STTs), we compared the diagnostic performance of an ensemble of random forest (R) models with single-sequence MRI inputs to R models with pooled multi-sequence MRI inputs. One-hundred twenty-five STT patients with preoperative MRI were retrospectively included and consisted of training (n = 100) and test (n = 25) sets. MRI included T1-weighted (T1-WI), T2-weighted (T2-WI), contrast-enhanced (CE)-T1-WI, diffusion-weighted images (DWIs, b = 800 sec/mm2) and apparent diffusion coefficient (ADC) maps. After tumor segmentation on each sequence, 100 original radiomic features were extracted from each sequence image and divided into three-feature sets: T features from T1- and T2-WI, CE features from CE-T1-WI, and D features from DWI and ADC maps. Four radiomics models were built using Lasso and R with four combinations of three-feature sets as inputs: T features (R-T), T+CE features (R-C), T+D features (R-D), and T+CE+D features (R-A) (Type-1 model). An ensemble model was built by soft voting of five, single-sequence-based R models (Type-2 model). AUC, sensitivity, specificity, and accuracy of each model was calculated with five-fold cross validation. In Type-1 model, AUC, sensitivity, specificity, and accuracy were 0.752, 71.8%, 61.1%, and 67.2% in R-T; 0.756, 76.1%, 70.4%, and 73.6% in R-C; 0.750, 77.5%, 63.0%, and 71.2% in R-D; and 0.749, 74.6%, 61.1%, and 68.8% R-A models, respectively. AUC, sensitivity, specificity, and accuracy of Type-2 model were 0.774, 76.1%, 68.5%, and 72.8%. In conclusion, an ensemble method is beneficial to incorporate features from multi-sequence MRI and showed diagnostic robustness for differentiating malignant STTs.

Prediction of microvascular invasion in hepatocellular carcinoma with magnetic resonance imaging using models combining deep attention mechanism with clinical features

To investigate the consistency and diagnostic performance of magnetic resonance imaging (MRI) for detecting microvascular invasion (MVI) of hepatocellular carcinoma (HCC) and the validity of deep learning attention mechanisms and clinical features for MVI grade prediction. This retrospective study was conducted among 158 patients with HCC treated in Shunde Hospital Affiliated to Southern Medical University between January, 2017 and February, 2020. The imaging data and clinical data of the patients were collected to establish single sequence deep learning models and fusion models based on the EfficientNetB0 and attention modules. The imaging data included conventional MRI sequences (T1WI, T2WI, and DWI), enhanced MRI sequences (AP, PP, EP, and HBP) and synthesized MRI sequences (T1mapping-pre and T1mapping-20 min), and the high-risk areas of MVI were visualized using deep learning visualization techniques. The fusion model based on T1mapping-20min sequence and clinical features outperformed other fusion models with an accuracy of 0.8376, a sensitivity of 0.8378, a specificity of 0.8702, and an AUC of 0.8501 for detecting MVI. The deep fusion models were also capable of displaying the high-risk areas of MVI. The fusion models based on multiple MRI sequences can effectively detect MVI in patients with HCC, demonstrating the validity of deep learning algorithm that combines attention mechanism and clinical features for MVI grade prediction.

The Effective 3D MRI Reconstruction Method Driven by the Fusion Strategy in NSST Domain

The 3D reconstruction of medical images plays an important role in modern clinical diagnosis. Although the analytic-based, the iterative-based and the deep learning-based methods have been popularly used, there are still many problems to deal with. The analysis-based methods are not accurate enough, the iteration-based methods are computationally intensive, and the deep learning based methods are heavily dependent on the training of the data. To solve the default that only the single scan sequence is included in the traditional methods, a reconstruction method driven by the non-subsampled shearlet transform (NSST) and the algebraic reconstruction technique (ART) is proposed. Firstly, the multiple magnetic resonance imaging (MRI) sequences are decomposed into high-frequency and low-frequency components by NSST. Secondly, the low-frequency parts are fused with the weighted average fusion scheme and the high-frequency parts are fused with the weighted coefficient scheme that guided by the regional average gradient and energy. Finally, the 3D reconstruction is performed by using the ART algorithm. Compared with the traditional reconstruction methods, the proposed method is able to capture more information from the multiple MRI sequences, which makes the reconstruction results much clearer and more accurate. By comparing with the single-sequence reconstruction model without fusion, the experiments fully prove the accuracy and effectiveness.

Classification of low- and high-grade gliomas using radiomic analysis of multiple sequences of MRI brain.

Gliomas are frequent tumors of brain parenchyma, which have histology similar to that of glial cells. Accurate glioma grading is required for determining clinical management. The background of this study is to investigate the accuracy of magnetic resonance imaging (MRI)-based radiomic features extracted from multiple MRI sequences in differentiating low and high-grade gliomas. This is a retrospective study. It includes two groups. Group A includes patients with confirmed histopathological diagnosis of low (23) and high-grade (58) gliomas from 2012 to 2020 were included. The MRI images were acquired using a Signa HDxt 1.5 Tesla MRI (GE Healthcare, Milwaukee, USA). Group B includes an external test set consisting of low- (20) and high-grade gliomas (20) obtained from The Cancer Genome Atlas (TCGA). The radiomic features were extracted from axial T2, apparent diffusion coefficient map, axial T2 fluid-attenuated inversion recovery, and axial T1 post-contrast sequences for both the groups. The Mann - Whitney U test was performed to assess the significant radiomic features useful for distinguishing the glioma grades for Group A. To determine the accuracy of radiomic features for differentiating gliomas, AUC was calculated from receiver operating characteristic curve analysis for both groups. Our study noticed in Group A, fourteen MRI-based radiomic features from four MRI sequences showed a significant difference (p < 0.001) in differentiating gliomas. In Group A, we noticed T1 post-contrast radiomic features such as first-order variance (FOV) (sensitivity - 94.56%, specificity - 97.51%, AUC - 0.969) and GLRLM long-run gray-level emphasis (sensitivity - 97.54%), specificity - 96.53%, AUC - 0.972) had the highest discriminative power for distinguishing the histological subtypes of gliomas. Our study noticed no statistical significant difference between ROC curves of significant radiomic features for both groups. In Group B, the T1 post-contrast radiomic features such as FOV (AUC-0.933) and GLRLM long-run gray-level emphasis (AUC-0.981) had also shown high discriminative power for distinguishing the gliomas. Our study concludes that MRI-based radiomic features extracted from multiple MRI sequences provide a non-invasive diagnosis of low- and high-grade gliomas and can be implemented in clinical settings for diagnosing the glioma grades.

Pathologic Complete Response Prediction after Neoadjuvant Chemoradiation Therapy for Rectal Cancer Using Radiomics and Deep Embedding Network of MRI

Assessment of magnetic resonance imaging (MRI) after neoadjuvant chemoradiation therapy (nCRT) is essential in rectal cancer staging and treatment planning. However, when predicting the pathologic complete response (pCR) after nCRT for rectal cancer, existing works either rely on simple quantitative evaluation based on radiomics features or partially analyze multi-parametric MRI. We propose an effective pCR prediction method based on novel multi-parametric MRI embedding. We first seek to extract volumetric features of tumors that can be found only by analyzing multiple MRI sequences jointly. Specifically, we encapsulate multiple MRI sequences into multi-sequence fusion images (MSFI) and generate MSFI embedding. We merge radiomics features, which capture important characteristics of tumors, with MSFI embedding to generate multi-parametric MRI embedding and then use it to predict pCR using a random forest classifier. Our extensive experiments demonstrate that using all given MRI sequences is the most effective regardless of the dimension reduction method. The proposed method outperformed any variants with different combinations of feature vectors and dimension reduction methods or different classification models. Comparative experiments demonstrate that it outperformed four competing baselines in terms of the AUC and F1-score. We use MRI sequences from 912 patients with rectal cancer, a much larger sample than in any existing work.

Radiomics signature as a new biomarker for preoperative prediction of neoadjuvant chemoradiotherapy response in locally advanced rectal cancer.

Whether radiomics methods are useful in prediction of therapeutic response to neoadjuvant chemoradiotherapy (nCRT) is unclear. This study aimed to investigate multiple magnetic resonance imaging (MRI) sequence-based radiomics methods in evaluating therapeutic response to nCRT in patients with locally advanced rectal cancer (LARC). This retrospective study enrolled patients with LARC (06/2014-08/2017) and divided them into nCRT-sensitive and nCRT-resistant groups according to postoperative tumor regression grading results. Radiomics features from preoperative MRI were extracted, followed by dimension reduction using the minimum redundancy maximum relevance filter. Three machine-learning classifiers and an ensemble classifier were used for therapeutic response prediction. Radiomics nomogram incorporating clinical parameters were constructed using logistic regression. The receiver operating characteristic (ROC), decision curves analysis (DCA) and calibration curves were also plotted to evaluate the prediction performance. The machine learning classifiers showed good prediction performance for therapeutic responses in LARC patients (n=189). The ROC curve showed satisfying performance (area under the curve [AUC], 0.830; specificity, 0.794; sensitivity, 0.815) in the validation group. The radiomics signature included 30 imaging features derived from axial T1-weighted imaging with contrast and sagittal T2-weighted imaging and exhibited good predictive power for nCRT. A radiomics nomogram integrating carcinoembryonic antigen levels and tumor diameter showed excellent performance with an AUC of 0.949 (95% confidence interval, 0.892-0.997; specificity, 0.909; sensitivity, 0.879) in the validation group. DCA confirmed the clinical usefulness of the nomogram model. The radiomics method using multiple MRI sequences can be used to achieve individualized prediction of nCRT in patients with LARC before treatment.

Comparison of Image Quality of Multiple Magnetic Resonance Imaging Sequences in Multiple Myeloma

Multiple myeloma is a refractory malignant disease characterized by clonal hyper proliferation of plasma cells in the bone marrow microenvironment. In recent years, its incidence has gradually increased and it is younger. Magnetic resonance imaging is a medical imaging technology that has developed rapidly in recent years. Its application and promotion have greatly improved the level of medical services and scientific research. It has become one of the most important examination methods for myeloma. Magnetic resonance imaging has a high soft tissue resolution and has a high detection rate for multiple myeloma. However, there are few studies on the MM magnetic resonance scanning protocol, and the initial or follow-up examination methods have not been unified. Therefore, this paper subjectively and objectively evaluates the clinical images of MM patients with multiple sequences of magnetic resonance of different devices, and hopes to provide more advantageous examination methods for clinicians and patients. The experimental results show that magnetic resonance multisequence imaging can be ideal for clinical diagnosis and follow-up of MM patients.

Emerging Imaging Biomarkers in Crohn Disease.

In this review article, we present the latest developments in quantitative imaging biomarkers based on magnetic resonance imaging (MRI), applied to the diagnosis, assessment of response to therapy, and assessment of prognosis of Crohn disease. We also discuss the biomarkers' limitations and future prospects. We performed a literature search of clinical and translational research in Crohn disease using diffusion-weighted MRI (DWI-MRI), dynamic contrast-enhanced MRI (DCE-MRI), motility MRI, and magnetization transfer MRI, as well as emerging topics such as T1 mapping, radiomics, and artificial intelligence. These techniques are integrated in and combined with qualitative image assessment of magnetic resonance enterography (MRE) examinations. Quantitative MRI biomarkers add value to MRE qualitative assessment, achieving substantial diagnostic performance (area under receiver-operating curve = 0.8-0.95). The studies reviewed show that the combination of multiple MRI sequences in a multiparametric quantitative fashion provides rich information that may help for better diagnosis, assessment of severity, prognostication, and assessment of response to biological treatment. However, the addition of quantitative sequences to MRE examinations has potential drawbacks, including increased scan time and the need for further validation before being used in therapeutic drug trials as well as the clinic.

MRI-Based Radiomics of Rectal Cancer: Assessment of the Local Recurrence at the Site of Anastomosis

To investigate the significance of magnetic resonance imaging (MRI)-based radiomics model in differentiating local recurrence of rectal cancer from nonrecurrence lesions at the site of anastomosis. A total of 80 patients with clinically suspected lesions of anastomosis underwent 3.0T pelvic MRI consisting of T2-weighted imaging (T2WI), diffusion-weighted imaging (DWI), and contrast-enhanced T1-weighted volume interpolated body examination (VIBE) imaging. Radiomics features were extracted from volumes of interest (VOIs), delineated manually on multiple MRI sequences. Subsequently, principal component analysis reduced the dimensionality of features for T2WI, DWI, VIBE, and combined multisequences, respectively. On this basis, the extreme gradient boosting (XGBoost) classifier was trained to build ModelT2WI, ModelDWI, ModelVIBE, and Modelcombination. Receiver operating characteristic curves were generated to determine the diagnostic performance of various models. Principal component analysis selected eight, four, seven, and six principal components to construct the radiomics model for T2WI, DWI, VIBE, and combined multisequences, respectively. Modelcombination had an area under the receiver operating characteristic curve of 0.864, with sensitivity and specificity of 81.82% and 75.86% in the validation set, demonstrating a more optimal performance compared to other models (p< 0.05). The decision curve analysis confirmed the clinical usefulness of the model. This study demonstrated that MRI-based radiomics is a sophisticated and noninvasive tool for accurately distinguishing LR from nonrecurrence lesions at the site of anastomosis. Combining multiple sequences significantly improves its performance.

A generative adversarial network-based (GAN-based) architecture for automatic fiducial marker detection in prostate MRI-only radiotherapy simulation images.

Clinical sites utilizing magnetic resonance imaging (MRI)-only simulation for prostate radiotherapy planning typically use fiducial markers for pretreatment patient positioning and alignment. Fiducial markers appear as small signal voids in MRI images and are often difficult to discern. Existing clinical methods for fiducial marker localization require multiple MRI sequences and/or manual interaction and specialized expertise. In this study, we develop a robust method for automatic fiducial marker detection in prostate MRI simulation images and quantify the pretreatment alignment accuracy using automatically detected fiducial markers in MRI. In this study, a deep learning-based algorithm was used to convert MRI images into labeled fiducial marker volumes. Seventy-seven prostate cancer patients who received marker implantation prior to MRI and CT simulation imaging were selected for this study. Multiple-Echo T1 -VIBE MRI images were acquired, and images were stratified (at the patient level) based on the presence of intraprostatic calcifications. Ground truth (GT) contours were defined by an expert on MRI using CT images. Training was done using the pix2pix generative adversarial network (GAN) image-to-image translation package and model testing was done using fivefold cross validation. For performance comparison, an experienced medical dosimetrist and a medical physicist each manually contoured fiducial markers in MRI images. The percent of correct detections and F1 classification scores are reported for markers detected using the automatic detection algorithm and human observers. The patient positioning errors were quantified by calculating the target registration errors (TREs) from fiducial marker driven rigid registration between MRI and CBCT images. Target registration errors were quantified for fiducial marker contours defined on MRI by the automatic detection algorithm and the two expert human observers. Ninety-six percent of implanted fiducial markers were correctly identified using the automatic detection algorithm. Two expert raters correctly identified 97% and 96% of fiducial markers, respectively. The F1 classification score was 0.68, 0.75, and 0.72 for the automatic detection algorithm and two human raters, respectively. The main source of false discoveries was intraprostatic calcifications. The mean TRE differences between alignments from automatic detection algorithm and human detected markers and GT were <1mm. We have developed a deep learning-based approach to automatically detect fiducial markers in MRI-only simulation images in a clinically representative patient cohort. The automatic detection algorithm-predicted markers can allow for patient setup with similar accuracy to independent human observers.

Subsampled brain MRI reconstruction by generative adversarial neural networks.

A main challenge in magnetic resonance imaging (MRI) is speeding up scan time. Beyond improving patient experience and reducing operational costs, faster scans are essential for time-sensitive imaging, such as fetal, cardiac, or functional MRI, where temporal resolution is important and target movement is unavoidable, yet must be reduced. Current MRI acquisition methods speed up scan time at the expense of lower spatial resolution and costlier hardware. We introduce a practical, software-only framework, based on deep learning, for accelerating MRI acquisition, while maintaining anatomically meaningful imaging. This is accomplished by MRI subsampling followed by estimating the missing k-space samples via generative adversarial neural networks. A generator-discriminator interplay enables the introduction of an adversarial cost in addition to fidelity and image-quality losses used for optimizing the reconstruction. Promising reconstruction results are obtained from feasible sampling patterns of up to a fivefold acceleration of diverse brain MRIs, from a large publicly available dataset of healthy adult scans as well as multimodal acquisitions of multiple sclerosis patients and dynamic contrast-enhanced MRI (DCE-MRI) sequences of stroke and tumor patients. Clinical usability of the reconstructed MRI scans is assessed by performing either lesion or healthy tissue segmentation and comparing the results to those obtained by using the original, fully sampled images. Reconstruction quality and usability of the DCE-MRI sequences is demonstrated by calculating the pharmacokinetic (PK) parameters. The proposed MRI reconstruction approach is shown to outperform state-of-the-art methods for all datasets tested in terms of the peak signal-to-noise ratio (PSNR), the structural similarity index (SSIM), as well as either the mean squared error (MSE) with respect to the PK parameters, calculated for the fully sampled DCE-MRI sequences, or the segmentation compatibility, measured in terms of Dice scores and Hausdorff distance. The code is available on GitHub.

External Fixation Devices Within the Magnetic Resonance Imaging Bore: A Safety and Radiologic Analysis.

To (1) report the thermal changes encountered at the pin/skin interface in a cadaver with a knee-spanning external fixator inside the magnetic resonance imaging (MRI) bore and (2) report on the quality of the MRI sequences collected. Three commonly used external fixation systems were placed on cadaveric lower extremities to simulate knee external fixation. Fiber optic thermal probes were placed at the pin/skin interface of a femoral and tibial pin. A control probe was embedded in the soft tissues of the thigh. Full knee MRI scans were performed using a 1.5-Tesla magnet. Real-time thermal data were collected. A clinically significant increase in temperature compared with the control was defined as 2°C. Two blinded radiologists evaluated the images for image quality and overall diagnostic utility using a standardized 5-point grading scale. There were statistically significant differences in the temperature changes between the femoral/tibial pin sites and the control probe sites during each phase of the MRI scan. However, there was only one clinically significant difference in temperature change during a single sequence of one MRI scan of one of the external fixator devices. Overall image quality was graded as a 4 for each image set with 100% interobserver agreement (k = 1.0). Despite significant differences in temperature changes between the pin sites and controls over multiple MRI sequences in commonly used external fixator devices, the differences in temperature change are likely not clinically relevant. Overall image quality and interpretability of the images were excellent.

Prostate cancer heterogeneity: texture analysis score based on multiple magnetic resonance imaging sequences for detection, stratification and selection of lesions at time of biopsy.

To undertake an early proof-of-concept study on a novel, semi-automated texture-based scoring system in order to enhance the association between magnetic resonance imaging (MRI) lesions and clinically significant prostate cancer (SPCa). With ethics approval, 536 imaging volumes were generated from 20 consecutive patients who underwent multiparametric MRI (mpMRI) at time of biopsy. Volumes of interest (VOIs) included zonal anatomy segmentation and suspicious MRI lesions for cancer (Likert Scale score >2). Entropy (E), measuring heterogeneity, was computed from VOIs and plotted as a multiparametric score defined as the entropy score (ES) = E ADC + E Ktrans + E Ve + E T2WI. The reference test that was used to define the ground truth comprised systematic saturation biopsies coupled with MRI-targeted sampling. This generated 422 cores in all that were individually labelled and oriented in three-dimensions. Diagnostic accuracy for detection of SPCa, defined as Gleason score ≥3 + 4 or >3 mm of any grade of cancer on a single core, was assessed using receiver operating characteristics, correlation, and descriptive statistics. The proportion of cancerous lesions detected by ES and visual scoring (VS) were statistically compared using the paired McNemar test. Any cancer (Gleason score 6-8) was found in 12 of the 20 (60%) patients, with a median PSA level of 8.22 ng/mL. SPCa (mean [95% confidence interval, CI] ES = 17.96 [0.72] NATural information unit [NAT]) had a significantly higher ES than non-SPCa (mean [95% CI] ES = 15.33 [0.76] NAT). The ES correlated with Gleason score (rs = 0.568, P = 0.033) and maximum cancer core length (ρ = 0.781; P < 0.001). The area under the curve for the ES (0.89) and VS (0.91) were not significantly different (P = 0.75) for the detection of SPCa amongst MRI lesions. Best ES estimated numerical threshold of 16.61 NAT led to a sensitivity of 100% and negative predictive value of 100%. The proportion of MRI lesions that were found to be positive for SPCa using this ES threshold (54%) was significantly higher (P < 0.001) than using the VS (24% of score 3, 4, 5) in a paired analysis using the McNemar test. In all, 53% of MRI lesions would have avoided biopsy sampling without missing significant disease. Capturing heterogeneity of prostate cancer across multiple MRI sequences with the ES yielded high performances for the detection and stratification of SPCa. The ES outperformed the VS in predicting positivity of lesions, holding promise in the selection of targets for biopsy and calling for further understanding of this association.