105 research outputs found
Semi-supervised assessment of incomplete LV coverage in cardiac MRI using generative adversarial nets
No full textCardiac magnetic resonance (CMR) images play a growing role in diagnostic imaging of cardiovascular diseases. Ensuring full coverage of the Left Ventricle (LV) is a basic criteria of CMR image quality. Complete LV coverage, from base to apex, precedes accurate cardiac volume and functional assessment. Incomplete coverage of the LV is identified through visual inspection, which is time-consuming and usually done retrospectively in large imaging cohorts. In this paper, we propose a novel semi-supervised method to check the coverage of LV from CMR images by using generative adversarial networks (GAN), we call it Semi-Coupled-GANs (SCGANs). To identify missing basal and apical slices in a CMR volume, a two-stage framework is proposed. First, the SCGANs generate adversarial examples and extract high-level features from the CMR images; then these image attributes are used to detect missing basal and apical slices. We constructed extensive experiments to validate the proposed method on UK Biobank with more than 6000 independent volumetric MR scans, which achieved high accuracy and robust results for missing slice detection, comparable with those of state of the art deep learning methods. The proposed method, in principle, can be adapted to other CMR image data for LV coverage assessment
Simulation and synthesis in medical imaging:Second international workshop, SASHIMI 2017 held in conjunction with MICCAI 2017 Québec city, QC, Canada, September 10, 2017 proceedings
No full textSimulation and synthesis in medical imaging:Second international workshop, SASHIMI 2017 held in conjunction with MICCAI 2017 Québec city, QC, Canada, September 10, 2017 proceedings
No full text
Automated Quality Assessment of Cardiac MR Images Using Convolutional Neural Networks
No full textImage quality assessment (IQA) is crucial in large-scale population imaging so that high-throughput image analysis can extract meaningful imaging biomarkers at scale. Specifically, in this paper, we address a seemingly basic yet unmet need: the automatic detection of missing (apical and basal) slices in Cardiac Magnetic Resonance Imaging (CMRI) scans, which is currently performed by tedious visual assessment. We cast the problem as classification tasks, where the bottom and top slices are tested for the presence of typical basal and apical patterns. Inspired by the success of deep learning methods, we train Convolutional Neural Networks (CNN) to construct a set of discriminative features. We evaluated our approach on a subset of the UK Biobank datasets. Precision and Recall figures for detecting missing apical slice (MAS) (81.61% and 88.73 %) and missing basal slice (MBS) (74.10% and 88.75 %) are superior to other state-of-the-art deep learning architectures. Cross-dataset experiments show the generalization ability of our approach.</p
Physics-informed brain MRI segmentation
No full textMagnetic Resonance Imaging (MRI) is one of the most flexible and powerful medical imaging modalities. This flexibility does however come at a cost; MRI images acquired at different sites and with different parameters exhibit significant differences in contrast and tissue appearance, resulting in downstream issues when quantifying brain anatomy or the presence of pathology. In this work, we propose to combine multiparametric MRI-based static-equation sequence simulations with segmentation convolutional neural networks (CNN), to make these networks robust to variations in acquisition parameters. Results demonstrate that, when given both the image and their associated physics acquisition parameters, CNNs can produce segmentations that exhibit robustness to acquisition variations. We also show that the proposed physics-informed methods can be used to bridge multi-centre and longitudinal imaging studies where imaging acquisition varies across a site or in time
Adversarial Image Synthesis for Unpaired Multi-Modal Cardiac Data
No full textThis paper demonstrates the potential for synthesis of medical images in one modality (e.g. MR) from images in another (e.g. CT) using a CycleGAN [24] architecture. The synthesis can be learned from unpaired images, and applied directly to expand the quantity of available training data for a given task. We demonstrate the application of this approach in synthesising cardiac MR images from CT images, using a dataset of MR and CT images coming from different patients. Since there can be no direct evaluation of the synthetic images, as no ground truth images exist, we demonstrate their utility by leveraging our synthetic data to achieve improved results in segmentation. Specifically, we show that training on both real and synthetic data increases accuracy by 15% compared to real data. Additionally, our synthetic data is of sufficient quality to be used alone to train a segmentation neural network, that achieves 95% of the accuracy of the same model trained on real data.</p
Simulation and Synthesis in Medical Imaging: 4th International Workshop, SASHIMI 2019, Held in Conjunction with MICCAI 2019, Shenzhen, China, October 13, 2019, Proceedings
No full textInternational audienc
Data augmentation using synthetic lesions improves machine learning detection of microbleeds from MRI
No full textMachine learning applied to medical imaging for lesions detection, such as cerebral microbleeds (CMB) from Magnetic Resonance Imaging (MRI), is challenged by the relatively small datasets available for which only subjective and tedious visual reading is available, and by the low prevalence of lesions (a few in ~10% of a typical elderly cohort) resulting in unbalanced classes. Moreover, the lack of actual ground truth might limit the performance of any machine learning method to that of human performance. Yet, the automatic identification of those lesions is relevant to quantify cerebrovascular burden associated with dementia, such as identifying co-morbidity for Alzheimer’s disease. In this paper, we investigated a novel approach consisting of simulating synthetic CMB on SWI MRI scans from healthy individuals to create a large and well characterized training dataset, as a data augmentation strategy. Firstly, we characterized actual CMBs from MRI SWI scans and designed a method to create realistic synthetic CMBs whose location, shape, appearance, and size are similar to actual CMBs. We then tested a supervised neural network classifier using various combinations of actual CMB and synthetic CMBs for training. Augmenting data with synthetic CMBs resulted in a large improvement over training on only actual CMBs only when tested on unseen lesions, and provided better results than other standard data augmentation approaches. Our results suggest that data augmentation using synthetic lesions can address the lack of ground truth and low prevalence limitations for medical imaging analysis allowing the deployment of data hungry supervised learning techniques such as deep learning.No Full Tex
- …
