Quality Of Synthetic Images Research Articles

Multi-task reconstruction network for synthetic diffusion kurtosis imaging: Predicting neoadjuvant chemoradiotherapy response in locally advanced rectal cancer

PurposeTo assess the feasibility and clinical value of synthetic diffusion kurtosis imaging (DKI) generated from diffusion weighted imaging (DWI) through multi-task reconstruction network (MTR-Net) for tumor response prediction in patients with locally advanced rectal cancer (LARC). MethodsIn this retrospective study, 120 eligible patients with LARC were enrolled and randomly divided into training and testing datasets with a 7:3 ratio. The MTR-Net was developed for reconstructing Dapp and Kapp images from apparent diffusion coefficient (ADC) images. Tumor regions were manually segmented on both true and synthetic DKI images. The synthetic image quality and manual segmentation agreement were quantitatively assessed. The support vector machine (SVM) classifier was used to construct radiomics models based on the true and synthetic DKI images for pathological complete response (pCR) prediction. The prediction performance for the models was evaluated by the receiver operating characteristic (ROC) curve analysis. ResultsThe mean squared error (MSE), peak signal-to-noise ratio (PSNR), and structural similarity index measure (SSIM) for tumor regions were 0.212, 24.278, and 0.853, respectively, for the synthetic Dapp images and 0.516, 24.883, and 0.804, respectively, for the synthetic Kapp images. The Dice similarity coefficient (DSC), positive predictive value (PPV), sensitivity (SEN), and Hausdorff distance (HD) for the manually segmented tumor regions were 0.786, 0.844, 0.755, and 0.582, respectively. For predicting pCR, the true and synthetic DKI-based radiomics models achieved area under the curve (AUC) values of 0.825 and 0.807 in the testing datasets, respectively. ConclusionsGenerating synthetic DKI images from DWI images using MTR-Net is feasible, and the efficiency of synthetic DKI images in predicting pCR is comparable to that of true DKI images.

The image quality and diagnostic accuracy of T1-mapping-based synthetic late gadolinium enhancement imaging: comparison with conventional late gadolinium enhancement imaging in real-life clinical situation

BackgroundsSynthetic late gadolinium enhancement (LGE) images are less sensitive to inversion time (TI) and robust to motion artifact, because it is generated retrospectively by post-contrast T1-mapping images. To explore the clinical applicability of synthetic LGE, we investigated the image quality and diagnostic accuracy of synthetic LGE images, in comparison to that of conventional LGE for various disease groups.Method and materialsFrom July to November 2019, a total of 98 patients who underwent cardiovascular magnetic resonance imaging (CMR), including LGE and T1-mapping sequences, with suspicion of myocardial abnormality were retrospectively included. Synthetic magnitude inversion-recovery (IR) and phase-sensitive IR (PSIR) images were generated through calculations based on the post-contrast T1-mapping sequence. Three cardiothoracic radiologists independently analyzed the image quality of conventional and synthetic LGE images on an ordinal scale with per-segment basis and the image qualities were compared with chi-square test. The agreement of LGE detection was analyzed on per-patient and per-segment basis with Cohen’s kappa test. In addition, the LGE area and percentage were semi-quantitatively analyzed for LGE positive ischemic (n = 14) and hypertrophic cardiomyopathy (n = 13) subgroups by two cardiothoracic radiologists. The difference of quantified LGE area and percentage between conventional and synthetic LGE images were assessed with Mann–Whitney U-test and the inter-reader agreement was assessed with Bland–Altman analysis.ResultsThe image quality of synthetic images was significantly better than conventional images in both magnitude IR and PSIR through all three observers (P < 0.001, all). The agreements of per-patient and per-segment LGE detection rates were excellent (kappa = 0.815–0.864). The semi-quantitative analysis showed no significant difference in the LGE area and percentage between conventional and synthetic LGE images. In the inter-reader agreement showed only small systematic differences in both magnitude IR and PSIR and synthetic LGE images showed smaller systematic biases compared to conventional LGE images.ConclusionCompared to conventional LGE images, synthetic LGE images have better image quality in real-life clinical situation.

Overcoming barriers to data sharing with medical image generation: a comprehensive evaluation

Privacy concerns around sharing personally identifiable information are a major barrier to data sharing in medical research. In many cases, researchers have no interest in a particular individual’s information but rather aim to derive insights at the level of cohorts. Here, we utilise generative adversarial networks (GANs) to create medical imaging datasets consisting entirely of synthetic patient data. The synthetic images ideally have, in aggregate, similar statistical properties to those of a source dataset but do not contain sensitive personal information. We assess the quality of synthetic data generated by two GAN models for chest radiographs with 14 radiology findings and brain computed tomography (CT) scans with six types of intracranial haemorrhages. We measure the synthetic image quality by the performance difference of predictive models trained on either the synthetic or the real dataset. We find that synthetic data performance disproportionately benefits from a reduced number of classes. Our benchmark also indicates that at low numbers of samples per class, label overfitting effects start to dominate GAN training. We conducted a reader study in which trained radiologists discriminate between synthetic and real images. In accordance with our benchmark results, the classification accuracy of radiologists improves with an increasing resolution. Our study offers valuable guidelines and outlines practical conditions under which insights derived from synthetic images are similar to those that would have been derived from real data. Our results indicate that synthetic data sharing may be an attractive alternative to sharing real patient-level data in the right setting.

Accelerated Isotropic Multiparametric Imaging by High Spatial Resolution 3D-QALAS With Compressed Sensing: A Phantom, Volunteer, and Patient Study.

The aims of this study were to develop an accelerated multiparametric magnetic resonance imaging method based on 3D-quantification using an interleaved Look-Locker acquisition sequence with a T2 preparation pulse (3D-QALAS) combined with compressed sensing (CS) and to evaluate the effect of CS on the quantitative mapping, tissue segmentation, and quality of synthetic images. A magnetic resonance imaging system phantom, containing multiple compartments with standardized T1, T2, and proton density (PD) values; 10 healthy volunteers; and 12 patients with multiple sclerosis were scanned using the 3D-QALAS sequence with and without CS and conventional contrast-weighted imaging. The scan times of 3D-QALAS with and without CS were 5:56 and 11:11, respectively. For healthy volunteers, brain volumetry and myelin estimation were performed based on the measured T1, T2, and PD. For patients with multiple sclerosis, the mean T1, T2, PD, and the amount of myelin in plaques and contralateral normal-appearing white matter (NAWM) were measured. Simple linear regression analysis and Bland-Altman analysis were performed for each metric obtained from the datasets with and without CS. To compare overall image quality and structural delineations on synthetic and conventional contrast-weighted images, case-control randomized reading sessions were performed by 2 neuroradiologists in a blinded manner. The linearity of both phantom and volunteer measurements in T1, T2, and PD values obtained with and without CS was very strong (R2 = 0.9901-1.000). The tissue segmentation obtained with and without CS also had high linearity (R2 = 0.987-0.999). The quantitative tissue values of the plaques and NAWM obtained with CS showed high linearity with those without CS (R2 = 0.967-1.000). There were no significant differences in overall image quality between synthetic contrast-weighted images obtained with and without CS (P = 0.17-0.99). Multiparametric imaging of the whole brain based on 3D-QALAS can be accelerated using CS while preserving tissue quantitative values, tissue segmentation, and quality of synthetic images.

Identity-aware CycleGAN for face photo-sketch synthesis and recognition

Face photo-sketch synthesis and recognition has many applications in digital entertainment and law enforcement. Recently, generative adversarial networks (GANs) based methods have significantly improved the quality of image synthesis, but they have not explicitly considered the purpose of recognition. In this paper, we first propose an Identity-Aware CycleGAN (IACycleGAN) model that applies a new perceptual loss to supervise the image generation network. It improves CycleGAN on photo-sketch synthesis by paying more attention to the synthesis of key facial regions, such as eyes and nose, which are important for identity recognition. Furthermore, we develop a mutual optimization procedure between the synthesis model and the recognition model, which iteratively synthesizes better images by IACycleGAN and enhances the recognition model by the triplet loss of the generated and real samples. Extensive experiments are performed on both photo-to-sketch and sketch-to-photo tasks using the widely used CUFS and CUFSF databases. The results show that the proposed method performs better than several state-of-the-art methods in terms of both synthetic image quality and photo-sketch recognition accuracy.

No-reference synthetic image quality assessment with convolutional neural network and local image saliency

Depth-image-based rendering (DIBR) is widely used in 3DTV, free-viewpoint video, and interactive 3D graphics applications. Typically, synthetic images generated by DIBR-based systems incorporate various distortions, particularly geometric distortions induced by object dis-occlusion. Ensuring the quality of synthetic images is critical to maintaining adequate system service. However, traditional 2D image quality metrics are ineffective for evaluating synthetic images as they are not sensitive to geometric distortion. In this paper, we propose a novel no-reference image quality assessment method for synthetic images based on convolutional neural networks, introducing local image saliency as prediction weights. Due to the lack of existing training data, we construct a new DIBR synthetic image dataset as part of our contribution. Experiments were conducted on both the public benchmark IRCCyN/IVC DIBR image dataset and our own dataset. Results demonstrate that our proposed metric outperforms traditional 2D image quality metrics and state-of-the-art DIBR-related metrics.

A Texture Synthesis Method Based on Multi-resolution Matching

Texture synthesis is the field of computer graphics and related basic core issues. Studies have shown that the MRF model-based two-dimensional image texture synthesis algorithm, the synthesis results are directly affected by the magnitude of the impact of random images; that is consistent with a high degree of MRF model is better than other texture synthesis texture. The same time, the quality of synthetic images and synthesis algorithms is proportional to time-consuming. In this paper, the synthesis of the texture image decomposition strategy is introduced and the concept of multi-resolution atlas. It can be used to guide the texture partition, to improve the quality of texture synthesis. Experiments show that the proposed algorithm is effective.