A Two-Stage GAN for High-Resolution Retinal Image Generation and Segmentation

In this paper, we use Generative Adversarial Networks (GANs) to synthesize high-quality retinal images along with the corresponding semantic label-maps, instead of real images during training of a segmentation network. Different from other previous proposals, we employ a two-step approach: first, a progressively growing GAN is trained to generate the semantic label-maps, which describes the blood vessel structure (i.e., the vasculature); second, an image-to-image translation approach is used to obtain realistic retinal images from the generated vasculature. The adoption of a two-stage process simplifies the generation task, so that the network training requires fewer images with consequent lower memory usage. Moreover, learning is effective, and with only a handful of training samples, our approach generates realistic high-resolution images, which can be successfully used to enlarge small available datasets. Comparable results were obtained by employing only synthetic images in place of real data during training. The practical viability of the proposed approach was demonstrated on two well-established benchmark sets for retinal vessel segmentation—both containing a very small number of training samples—obtaining better performance with respect to state-of-the-art techniques.

Dual Attention Multiscale Network for Vessel Segmentation in Fundus Photography

Background: Automatic vessel structure segmentation is an essential step towards an automatic disease diagnosis system. The task is challenging due to the variance shapes and sizes of vessels across populations.Methods: A multiscale network with dual attention is proposed to segment vessels in different sizes. The network injects spatial attention module and channel attention module on feature map which size is 1 8 of the input size. The network also uses multiscale input to receive multi-level information, and the network uses the multiscale output to gain more supervision. Results: The proposed method is tested on two publicly available datasets: DRIVE and CHASEDB1. The accuracy, AUC, sensitivity, specificity on DRIVE dataset is 0.9615, 0.9866, 0.7693, and 0.9851, respectively. On the CHASEDB1 dataset, the metrics are 0.9797, 0.9895, 0.8432, and 0.9863 respectively. The ablative study further shows effectiveness for each part of the network. Conclusions: Multiscale and dual attention mechanism both improves the performance. The proposed architecture is simple and effective. The inference time is 12ms on a GPU and has potential for real-world applications. The code will be made publicly available.

RFARN: Retinal vessel segmentation based on reverse fusion attention residual network

Accurate segmentation of retinal vessels is critical to the mechanism, diagnosis, and treatment of many ocular pathologies. Due to the poor contrast and inhomogeneous background of fundus imaging and the complex structure of retinal fundus images, this makes accurate segmentation of blood vessels from retinal images still challenging. In this paper, we propose an effective framework for retinal vascular segmentation, which is innovative mainly in the retinal image pre-processing stage and segmentation stage. First, we perform image enhancement on three publicly available fundus datasets based on the multiscale retinex with color restoration (MSRCR) method, which effectively suppresses noise and highlights the vessel structure creating a good basis for the segmentation phase. The processed fundus images are then fed into an effective Reverse Fusion Attention Residual Network (RFARN) for training to achieve more accurate retinal vessel segmentation. In the RFARN, we use Reverse Channel Attention Module (RCAM) and Reverse Spatial Attention Module (RSAM) to highlight the shallow details of the channel and spatial dimensions. And RCAM and RSAM are used to achieve effective fusion of deep local features with shallow global features to ensure the continuity and integrity of the segmented vessels. In the experimental results for the DRIVE, STARE and CHASE datasets, the evaluation metrics were 0.9712, 0.9822 and 0.9780 for accuracy (Acc), 0.8788, 0.8874 and 0.8352 for sensitivity (Se), 0.9803, 0.9891 and 0.9890 for specificity (Sp), area under the ROC curve(AUC) was 0.9910, 0.9952 and 0.9904, and the F1-Score was 0.8453, 0.8707 and 0.8185. In comparison with existing retinal image segmentation methods, e.g. UNet, R2UNet, DUNet, HAnet, Sine-Net, FANet, etc., our method in three fundus datasets achieved better vessel segmentation performance and results.

103 Coronary plaque healing and diabetes: insights from optical coherence tomography imaging

Aims Atherosclerotic plaque healing is a dynamic process developing after plaque rupture or erosion, which aims to prevent lasting occlusive thrombus formation and to promote plaque repair. We hypothesized that diabetes mellitus, one of the major conventional cardiovascular risk factors, may influence the healing capacity after plaque destabilization. Methods and results In this single-centre observational cohort study, patients with acute coronary syndrome (ACS) or chronic coronary syndrome (CCS) who underwent optical coherence tomography (OCT) imaging at Fondazione Policlinico A. Gemelli–IRCCS, Rome, were included. Patients were divided into two groups (i.e. diabetes vs. no diabetes), and stratified based on diabetes medications (i.e. insulin, vs. oral antidiabetic drugs). OCT analysis of non-culprit coronary segments was performed. 105 patients were included (44 diabetes, 61 no diabetes). Prevalence of HCPs was not significantly different between patients with and without diabetes (3.6% vs. 3.8%, P = 0.854). However, patients with diabetes on insulin showed a lower prevalence of HCPs both at patient-based (7.1% vs. 26.4%, P = 0.116) and at segment-based analysis (1.2% vs. 4.2%, P = 0.020). When comparing HbA1c levels based on the presence or absence of healed plaque at the non-culprit lesions, patients with healed plaque showed significantly lower levels of HbA1c compared to patients without healed plaques (43.5 ± 12.1% vs. 61.2 ± 10.4%, P < 0.001). At segment-based analysis, normal vessel structure, pathological intimal thickening (PIT), and spotty calcifications were significantly less prevalent in diabetic patients (2.1% vs. 5.1%, P = 0.001; 7.2% vs. 9.5%, P = 0.05; 9.9% vs. 13.6%, P = 0.02, respectively), whereas neovascularization was significantly higher (19.2% vs. 15.6%, P = 0.035). Conclusions Patients with diabetes have a distinct coronary non-culprit plaque phenotype. Healing capacity may be impaired in patients with advanced diabetes on insulin therapy and in those with a suboptimal control of the disease. Further prospective, larger scale studies are warranted to confirm these findings.

Deep Att-ResGAN: A Retinal Vessel Segmentation Network for Color Fundus Images

Retinal vessel segmentation plays a significant role in the diagnosis and treatment of ophthalmological diseases. Recent studies have proved that deep learning can effectively segment the retinal vessel structure. However, the existing methods have difficulty in segmenting thin vessels, especially when the original image contains lesions. Based on generative adversarial network (GAN), this paper proposes a deep network with residual module and attention module (Deep Att-ResGAN). The network consists of four identical subnetworks. The output of each subnetwork is imported to the next subnetwork as contextual features that guide the segmentation. Firstly, the problems of the original image, namely, low contrast, uneven illumination, and data insufficiency, were solved through image enhancement and preprocessing. Next, an improved U-Net was adopted to serve as the generator, which stacks the residual and attention modules. These modules optimize the weight of the generator, and enhance the generalizability of the network. Further, the segmentation was refined iteratively by the discriminator, which contributes to the performance of vessel segmentation. Finally, comparative experiments were carried out on two public datasets: Digital Retinal Images for Vessel Extraction (DRIVE) and Structured Analysis of the Retina (STARE). The experimental results show that Deep Att-ResGAN outperformed the equivalent models like U-Net and GAN in most metrics. Our network achieved accuracy of 0.9565 and F1 of 0.829 on DRIVE, and accuracy of 0.9690 and F1 of 0.841 on STARE.

Electrical Modelling of In-Vivo Impedance Spectroscopy of Nicotiana tabacum Plants

Electrical impedance spectroscopy has been suggested as a sensing method for plants. Here, a theoretical approach for electrical conduction via the plant stem is presented and validated, linking its living electrical characteristics to its internal structure. An electrical model for the alternating current conduction and the associated impedance in a live plant stem is presented. The model accounts for biological and geometrical attributes. It uses the electrically prevalent coupled transmission line model approach for a simplified description of the complicated vessel structure. It considers the electrode coupling to the plant stem (either Galvanic or Faradic), and accounts for the different interactions of the setup. Then the model is simplified using the lumped element approach. The model is then validated using a four-point probe impedance spectroscopy method, where the probes are galvanically coupled to the stem of Nicotiana tabacum plants. The electrical impedance data was collected continuously and the results exhibit an excellent fitting to the theoretical model, with a fitting error of less than 1.5% for data collected on various days and plants. A parametric evaluation of the fitting corresponds to the proposed physically based model, therefore providing a baseline for future plant sensor design.

MRI investigation of vascular remodeling for heterogeneous edema lesions in subacute ischemic stroke rat models: Correspondence between cerebral vessel structure and function

The spatial heterogeneity in the temporal occurrence of pseudo-normalization of MR apparent diffusion coefficient values for ischemic lesions may be related to morphological and functional vascular remodeling. As the area of accelerated pseudo-normalization tends to expand faster and more extensively into the chronic stage, detailed vascular characterization of such areas is necessary. During the subacute stage of transient middle cerebral artery occlusion rat models, the morphological size of the macrovasculature, microvascular vessel size index (VSI), and microvessel density (MVD) were quantified along with functional perfusion measurements of the relative cerebral blood flow (rCBF) and mean transit time (rMTT) of the corresponding areas (33 cases for each parameter). When compared with typical pseudo-normalization lesions, early pseudo-normalization lesions exhibited larger VSI and rCBF (p < 0.001) at reperfusion days 4 and 7, along with reduced MVD and elongated rMTT (p < 0.001) at reperfusion days 1, 4, and 7. The group median VSI and rCBF exhibited a strong positive correlation (r = 0.92), and the corresponding MVD and rMTT showed a negative correlation (r = −0.48). Light sheet fluorescence microscopy images were used to quantitatively validate the corresponding MRI-derived microvascular size, density, and cerebral blood volume.

In Vivo Simultaneous Nonlinear Absorption Raman and Fluorescence (SNARF) Imaging of Mouse Brain Cortical Structures

Two photon excited fluorescence (TPEF) microscopy is a widely used optical imaging technique that has revolutionized neurophotonics through a diverse palette of dyes, specialized transgenic models, easy implementation, and straightforward data analysis. However, in vivo TPEF imaging is often limited in the number of contrasts available to distinguish different cells, structures, or functions. We propose using two label free multiphoton microscopy techniques: stimulated Raman scattering (SRS) microscopy and transient absorption microscopy (TAM) as complementary and orthogonal imaging modalities to TPEF for in vivo brain imaging. In this study, we construct a simultaneous nonlinear absorption, Raman, and fluorescence (SNARF) microscope and image several cortical structures up to 250-300 μm below the pial surface, the highest reported in vivo imaging depth for SRS or TAM. We further demonstrate the capabilities of our SNARF microscope through the quantification of age-dependent myelination, hemodynamics, vessel structure, cell density, and cell identity in vivo. Using machine learning, we report the use of label free SRS and TAM features to predict capillary lining cell identities with 90% accuracy. The SNARF microscope and methodology outlined herein provide a powerful platform to study several research topics, including neurovascular coupling, blood brain barrier, neuronal and axonal degeneration in aging, and neurodegenerative diseases.

Establishment of Standard Human Blood Vessel Model Based on Image Registration and Fitting Technology

Purpose: The blood vessel gives key information for pathological changes in a variety of diseases. In view of the crucial role of blood vessel structure, the present study aims to establish a digital human blood vessel standard model for diagnosing blood vessel-related diseases. Methods: The present study recruited eight healthy volunteers, and reconstructed their bilateral upper extremity arteries according to CTA. The reconstructed vessels were segmented, registered, and merged into a bunch. After being cut by continuous cut planes, the dispersion of the blood vessel bunches on each cut plane were calculated. Results: The results demonstrated that the middle segment of the brachial artery, the proximal segment of the ulnar artery, and the middle and distal segments of the radial artery had a low degree of dispersion. A standard blood vessel model was finally established by the integral method using the low-dispersion segments above. The accuracy of the standard blood vessel model was also verified by an actual contralateral vessel, which revealed that the deviation between the model and the actual normal contralateral brachial artery was relatively small. Conclusion: The structure of the model was highly accordant with the real ones, which can be of great help in evaluating the blood vessel changes in blood vessel-related diseases, bone and soft-tissue tumors, and creating accurate surgical plans.