Retinal image synthesis from multiple-landmarks input with generative adversarial networks.

BACKGROUND: Medical datasets, especially medical images, are often imbalanced due to the different incidences of various diseases. To address this problem, many methods have been proposed to synthesize medical images using generative adversarial networks (GANs) to enlarge training datasets for facilitating medical image analysis. For instance, conventional methods such as image-to-image translation techniques are used to synthesize fundus images with their respective vessel trees in the field of fundus image. METHODS: In order to improve the image quality and details of the synthetic images, three key aspects of the pipeline are mainly elaborated: the input mask, architecture of GANs, and the resolution of paired images. We propose a new preprocessing pipeline named multiple-channels-multiple-landmarks (MCML), aiming to synthesize color fundus images from a combination of vessel tree, optic disc, and optic cup images. We compared both single vessel mask input and MCML mask input on two public fundus image datasets (DRIVE and DRISHTI-GS) with different kinds of Pix2pix and Cycle-GAN architectures. A new Pix2pix structure with ResU-net generator is also designed, which has been compared with the other models. RESULTS AND CONCLUSION: As shown in the results, the proposed MCML method outperforms the single vessel-based methods for each architecture of GANs. Furthermore, we find that our Pix2pix model with ResU-net generator achieves superior PSNR and SSIM performance than the other GANs. High-resolution paired images are also beneficial for improving the performance of each GAN in this work. Finally, a Pix2pix network with ResU-net generator using MCML and high-resolution paired images are able to generate good and realistic fundus images in this work, indicating that our MCML method has great potential in the field of glaucoma computer-aided diagnosis based on fundus image.

Spatial resolution enhancement for pushbroom-based microscopic hyperspectral imaging.

Hyperspectral imaging (HSI) is a promising tool for microscopic histopathology studies. Pushbroom microscopic hyperspectral imaging systems are widely used because of their low cost and easy implementation. However, the spatial resolution of pushbroom HSI systems is limited by the width of the optical entrance slit. A narrower slit leads to longer exposure time and slower imaging speed. In this paper, we explored several spatial resolution enhancement algorithms, originally designed for remote-sensing hyperspectral imaging, for pushbroom microscopic HSI systems. Our results demonstrate that those algorithms could effectively achieve a higher spatial resolution without sacrificing imaging speed.

A super-resolution method-based pipeline for fundus fluorescein angiography imaging.

BACKGROUND: Fundus fluorescein angiography (FFA) imaging is a standard diagnostic tool for many retinal diseases such as age-related macular degeneration and diabetic retinopathy. High-resolution FFA images facilitate the detection of small lesions such as microaneurysms, and other landmark changes, in the early stages; this can help an ophthalmologist improve a patient's cure rate. However, only low-resolution images are available in most clinical cases. Super-resolution (SR), which is a method to improve the resolution of an image, has been successfully employed for natural and remote sensing images. To the best of our knowledge, no one has applied SR techniques to FFA imaging so far. METHODS: In this work, we propose a SR method-based pipeline for FFA imaging. The aim of this pipeline is to enhance the image quality of FFA by using SR techniques. Several SR frameworks including neighborhood embedding, sparsity-based, locally-linear regression and deep learning-based approaches are investigated. Based on a clinical FFA dataset collected from Second Affiliated Hospital to Xuzhou Medical University, each SR method is implemented and evaluated for the pipeline to improve the resolution of FFA images. RESULTS AND CONCLUSION: As shown in our results, most SR algorithms have a positive impact on the enhancement of FFA images. Super-resolution forests (SRF), a random forest-based SR method has displayed remarkable high effectiveness and outperformed other methods. Hence, SRF should be one potential way to benefit ophthalmologists by obtaining high-resolution FFA images in a clinical setting.

In vivo long-term investigation of tumor bearing mKate2 by an in-house fluorescence molecular imaging system.

BACKGROUND: Optical imaging is one of the most common, low-cost imaging tools used for investigating the tumor biological behavior in vivo. This study explores the feasibility and sensitivity of a near infrared fluorescent protein mKate2 for a long-term non-invasive tumor imaging in BALB/c nude mice, by using a low-power optical imaging system. METHODS: In this study, breast cancer cell line MDA-MB-435s expressing mKate2 and MDA-MB-231 expressing a dual reporter gene firefly luciferase (fLuc)-GFP were used as cell models. Tumor cells were implanted in different animal body compartments including subcutaneous, abdominal and deep tissue area and closely monitored in real-time. A simple and low-power optical imaging system was set up to image both fluorescence and bioluminescence in live animals. RESULTS: The presence of malignant tissue was further confirmed by histopathological assay. Considering its lower exposure time and no need of substrate injection, mKate2 is considered a superior choice for subcutaneous imaging compared with fLuc. On the contrary, fLuc has shown to be a better option when monitoring the tumor in a diffusive area such as abdominal cavity. Furthermore, both reporter genes have shown good stability and sensitivity for deep tissue imaging, i.e. tumor within the liver. In addition, fLuc has shown to be an excellent method for detecting tumor cells in the lung. CONCLUSIONS: The combination of mKate2 and fLuc offers a superior choice for long-term non-invasive real-time investigation of tumor biological behavior in vivo.

Graphene/Intermetallic PtPb Nanoplates Composites for Boosting Electrochemical Detection of H2O2 Released from Cells.

Rational design and construction of electrocatalytic nanomaterials is vital for improving the sensitivity and selectivity of nonenzymatic electrochemical sensors. Here, we report a novel graphene supported intermetallic PtPb nanoplates (PtPb/G) nanocomposite as an enhanced electrochemical sensing platform for high-sensitivity detection of H2O2 in neutral solution and also released from the cells. The intermetallic PtPb nanoplates are first synthesized via a simple wet-chemistry process and subsequently assembled on graphene via a solution-phase self-assembly approach. The obtained nanocomposite exhibits excellent electrocatalytic activity for the electrochemical reduction of H2O2 in a half-cell test and can detect H2O2 with a wide linear detection range of 2 nM to 2.5 mM and a very low detection limit of 2 nM. Under the same conditions, the sensitivity of PtPb/G for the detection of H2O2 is more than 12.7 times higher than that of commercial Pt/C. The high-density of electrocatalytic active sites on the unique PtPb nanoplates and the synergistic effect between PtPb nanoplates and graphene appear to be the main factors in contributing to the outstanding electroanalytical performance. The PtPb/G can be also used for the practical detection of H2O2 released from Raw 264.7 cells.

Three-dimensional photoacoustic imaging via scanning a one dimensional linear unfocused ultrasound array.

Three-dimensional (3D) photoacoustic imaging can be achieved by a two-dimensional (2D) ultrasound array matrix. However, the 2D matrix consisting of hundreds or thousands of transducer elements makes it not only expensive, but also a big technical challenge for both probe manufacturing and parallel data acquisition. In this study, we performed the photoacoustic imaging by scanning an unfocused linear ultrasound array probe over a planar geometry, resulting in an equivalent 2D matrix probe. The phantom study demonstrated that this method substantially increased imaging quality, which has great potential for animal and clinical photoacoustic imaging.

Adaptive classifier allows enhanced flow contrast in OCT angiography using a histogram-based motion threshold and 3D Hessian analysis-based shape filtering.

In this Letter, we propose an adaptive digital classifier for flow contrast enhancement in optical coherence tomography angiography (OCTA). To solve the depth dependence in the initial motion-based classification, a depth-adaptive motion threshold was determined by performing a histogram analysis of an en-face image at each depth and identifying the static and dynamic voxel populations through fitting. In the follow-up shape-based classification, to adapt to the deformed vessel shapes in OCTA, a modified vesselness function along with an anisotropic Gaussian probe kernel was defined, and then a three-dimensional (3D) Hessian analysis-based shape filtering was utilized for effectively removing the residual static voxels. The experimental outcomes validated that the proposed adaptive digital classifier enabled a superior flow contrast by combining both the motion and 3D shape information.

Establishing Reliable Cu-64 Production Process: From Target Plating to Molecular Specific Tumor Micro-PET Imaging.

Copper-64 is a useful radioisotope for positron emission tomography (PET). Due to the wide range of applications, the demand of 64Cu with low metallic impurities is increasing. Here we report a simple method for the efficient production of high specific activity 64Cu using a cyclotron for biomedical application. We designed new equipment based on the plating of enriched 64Ni as the target, and used automated ion exchange chromatography to purify copper-64 efficiently after irradiation and dissolution of the target in good radiochemical and chemical yield and purity. The 64Cu radionuclide produced using 99.32% enriched 64Ni with a density of 61.4 ± 5.0 mg/cm², reaching a total radioactivity greater than 200 mCi, with specific activity up to 5.6 GBq/µmoL. It was further incorporated into modified monoclonal antibody DOTA-rituximab to synthesize 64Cu-DOTA-rituximab, which was used successfully for micro-PET imaging.

A Novel Scatter Correction Method for Cone-Beam Computed Tomography.

In this article, a novel scatter correction approach was proposed based on the Klein-Nishina formulation. Through a series of deductions from this formulation, a principle was proposed that the photon intensity distribution was determined by the attenuation coefficient µ and the path length l. This means if 2 pencil beams pass through 2 objects with the same µl, even if the attenuation coefficient µ and the path length l of the objects are different, they will still achieve the same photon intensity distribution, that is, the same point spread function. Subsequently, a novel scatter correction approach was established after a series of deductions based on this principle. The simulations and experiments demonstrated the correctness of our principle and the comparable correction effect of our scatter correction approach compared with the beam stop array method. Furthermore, because of the character of our method, the program has very high parallel computing features, which can dramatically increase the computation speed.

A New Method for Cone-Beam Computed Tomography Geometric Parameters Estimation.

This work demonstrates a new approach for geometric parameters estimation of cone-beam computed tomography system from the coordinates of the centroids of 2 projected point sources sampled over 360 degrees. Nonlinear object expression was derived for the coordinates of the centroids in terms of the geometric parameters after a slice of reasonable simplification, which aims to improve the convergence and robustness of the nonlinear object expression. All of the geometric parameters could be precisely estimated from the nonlinear object expression using the annealing algorithm. The simulations and experiments indicate more excellent convergence, robustness, and precision of our approach compared with other methods. Furthermore, our approach is insensitive to the initial value; namely, we do not need to set the value close to the true value to guarantee the convergence of the approach.

Evaluation of a Wobbling Method Applied to Correcting Defective Pixels of CZT Detectors in SPECT Imaging.

In this paper, we propose a wobbling method to correct bad pixels in cadmium zinc telluride (CZT) detectors, using information of related images. We build up an automated device that realizes the wobbling correction for small animal Single Photon Emission Computed Tomography (SPECT) imaging. The wobbling correction method is applied to various constellations of defective pixels. The corrected images are compared with the results of conventional interpolation method, and the correction effectiveness is evaluated quantitatively using the factor of peak signal-to-noise ratio (PSNR) and structural similarity (SSIM). In summary, the proposed wobbling method, equipped with the automatic mechanical system, provides a better image quality for correcting defective pixels, which could be used for all pixelated detectors for molecular imaging.

[Research progress of visual prosthesis].

Vision is one of the most important human sensations about the surrounding world. Visual deprivation not only markedly affects the life of blind people, but also gives a heavy burden to their family and the society. A visual prosthesis is an electronic device that helps the blinds to regain visual perception by directly stimulating the visual pathway using the microelectrodes implanted into the body. In recent years, visual prostheses have been developed rapidly and some devices have already become clinically available. In this paper, we reviewed the history of visual prosthesis, introduced different visual prostheses classified according to the location of the implanted stimulating electrodes. Clinical study results as well as the functional status of the currently available visual prosthesis devices were also summarized.

Bloodletting at Jing-well points decreases interstitial fluid flow in the thalamus of rats.

OBJECTIVE: To investigate the changes in the neuronal microenvironment of the middle cerebral artery (MCA) territory induced by Jing-well points bloodletting acupuncture (WPBA) and to explore the neuroprotective mechanism of WPBA in stroke. METHODS: Adult male Sprague Dawley (n = 32) rats were randomly divided into four groups of eight animals each: WPBA-thalamus group (WT), WPBA-caudate nucleus group (WC), sham-control thalamus group (ST) and sham-control caudate nucleus group (SC). Animals in the WT and WC groups received 2 µL of the extracellular tracer gadolinium-diethylene triamine pentaacetic acid (Gd-DTPA) injected into the thalamus or caudate nucleus, respectively, and 12 Jing-well points in the distal ends of the rats' digits were used for WPBA. Although 2 µL of Gd-DTPA was injected into the thalamus or caudate nucleus, respectively, for animals in the two sham groups (ST and SC), no acupuncture or bloodletting was performed. Brain extracellular space and interstitial fluid flow parameters were measured using Gd-DTPA-enhanced magnetic resonance imaging. RESULTS: The brain interstitial fluid flow speed was decreased in the thalamus after WPBA, with a significantly lower Gd-DTPA clearance rate and longer half-life of Gd-DTPA in the thalamus of treated rats than those in sham-control rats [WPBA-treated rats' clearance rate, (7.47 ± 3.15) x 10(-5)/s (P.= 0.009); half-life, (1.52 ± 0.13) h, P = 0.000]. By contrast, no significant changes in brain extracellular space and interstitial fluid flow parameters were detected in the caudate nucleus after WPBA (P = 0.649). In addition, no differences in the morphology of the brain extracellular space or the final distribution of the traced brain interstitial fluid were demonstrated between the WT and WC groups (P = 0.631, P = 0.970, respectively). CONCLUSION: The WPBA decreased the speed of the local thalamic ISF flow in rats, which is assumed to be a beneficial protection by down-modulated the metabolic rate of the attacked neurons under stroke.

Properties of electrically evoked potentials activated by optic nerve stimulation with penetrating electrodes of different modes in rabbits.

PURPOSE: To investigate the effect of different stimulation modes on cortical electrically evoked potentials (EEPs) by intraorbital optic nerve (ON) stimulation with penetrating electrodes. METHODS: A stimulating electrode array with three electrodes arranged linearly was inserted into the ON along its axis. EEPs were recorded using a 4 × 4 silver-ball electrode array in response to monopolar and bipolar stimulation mode, respectively. RESULTS: The simultaneous monopolar stimulation mode had a lower threshold than the individual monopolar stimulation mode, but elicited smaller cortical response when a fixed charge was injected. The threshold of the bipolar stimulation mode was comparable to that of individual monopolar stimulation mode. The response to the smaller spacing (150 µm) bipolar stimulation mode was similar in amplitude to that of the individual monopolar stimulation mode, but spread wider. The larger spacing (500 µm) bipolar stimulation mode elicited stronger and wider response than the individual monopolar stimulation mode. For the individual monopolar stimulation mode, stimulation with different electrodes can be differentiated even when the spacing of the two electrodes was 150 µm. CONCLUSIONS: For ON stimulation with penetrating electrodes, the monopolar stimulation mode could induce more localized cortical responses than the bipolar stimulation mode with comparable threshold and had a high stimulation selectivity. These findings may provide valuable information for the design of stimulation strategy of the penetrative ON visual prosthesis.

A modularly designed fluorescence molecular tomography system for multi-modality imaging.

Nowadays multi-modality imaging has gained great interest in biology research by offering complementary information. In this paper, a modularly designed fluorescence molecular tomography (FMT) system has been developed, which can be not only used as a standalone imaging device, but also feasibly integrated with other imaging modalities, such as X-ray computed tomography (X-CT), single photon emission computed tomography (SPECT) and positron emission tomography (PET), to perform multi-modality imaging in a sequential manner. The system rotates the CCD camera and the excitation light source in the vertical plane, while the animal is stationed on a horizontally moveable transparent animal holder at its natural prone position. FMT and other imaging modalities are co-registered automatically. Phantom and animal experiments have been carried out to demonstrate the performance of the system. The accurate results show that this innovative flexible FMT system has a great potential to be a powerful tool for the study of small animal disease models.

High-resolution dual-modality photoacoustic ocular imaging.

We have developed a prototype ocular imaging system that integrates optical-resolution photoacoustic microscopy and high-frequency ultrasound imaging. The system can perform high-resolution ocular imaging from the anterior region down to the fundus. It has successfully imaged murine eyes in vivo, including iris, lens, retina, and retinal pigment epithelium. Our results demonstrate that this system shows strong potential for the diagnosis of ophthalmic diseases.

Air puff induced corneal vibrations: theoretical simulations and clinical observations.

PURPOSE: To investigate air puff induced corneal vibrations and their relationship to the intraocular pressure (IOP), viscoelasticity, mass, and elasticity of the cornea based on theoretical simulations and preliminary clinical observations. METHODS: To simulate the corneal movement during air puff deformation, a kinematic viscoelastic corneal model was developed involving the factors of corneal mass, damping coefficient, elasticity, and IOP. Different parameter values were taken to investigate how factors would affect the corneal movements. Two clinical ocular instruments, CorVis ST (Oculus Optikgeräte GmbH, Wetzlar, Germany) and the Ocular Response Analyzer (ORA; Reichert, Inc., Buffalo, NY), were employed to observe the corneal dynamical behaviors. RESULTS: Numerical results showed that during the air puff deformation, there would be vibrations along with the corneal deformation, and the damping viscoelastic response of the cornea had the potential to reduce the vibration amplitude. With consistent IOP, the overall vibration amplitude and inward motion depths were smaller with a stiffer cornea. CONCLUSIONS: A kinematic viscoelastic model of the cornea is presented to illustrate how the vibrations are associated with factors such as corneal mass, viscoelasticity, and IOP. Also, the predicted corneal vibrations during air puff deformation were confirmed by clinical observation.

Dual-channel spectral-domain optical-coherence tomography system based on 3 × 3 fiber coupler for extended imaging range.

We have demonstrated a dual-channel multiplexing spectral-domain optical-coherence tomography (SD-OCT) system based on a 3×3 fiber coupler for extended imaging range of whole human eye depth, with a single light source and spectrometer. OCT images of anterior segments of a human eye were sequentially performed and constructed to demonstrate an extended depth range as large as 15 mm in air. A good quality OCT image of the whole anterior segment of an eye was present. Furthermore, whole eye segmental imaging was performed and ocular distances were calculated to show the validation of the system for whole eye morphological measurement.

Boosting Weakly Supervised Object Localization and Segmentation With Domain Adaption.

Weakly supervised object localization (WSOL), adopting only image-level annotations to learn the pixel-level localization model, can release human resources in the annotation process. Most one-stage WSOL methods learn the localization model with multi-instance learning, making them only activate discriminative object parts rather than the whole object. In our work, we attribute this problem to the domain shift between the training and test process of WSOL and provide a novel perspective that views WSOL as a domain adaption (DA) task. Under this perspective, a DA-WSOL pipeline is elaborated to better assist WSOL with DA approaches by considering the specificities for the adaption of WSOL. Our DA-WSOL pipeline can discern the source-related and the Universum samples from other target samples based on a proposed target sampling strategy and then utilize them to solve the sample unbalancing and label unmatching between the source and target domain of WSOL. Experiments show that our pipeline outperforms SOTA methods on three WSOL benchmarks and can improve the performance of downstream weakly supervised semantic segmentation tasks. Codes are available at https://github.com/zh460045050/dawsol.

Suppressing label noise in medical image classification using mixup attention and self-supervised learning.

Deep neural networks (DNNs) have been widely applied in medical image classification and achieve remarkable classification performance. These achievements heavily depend on large-scale accurately annotated training data. However, label noise is inevitably introduced in the medical image annotation, as the labeling process heavily relies on the expertise and experience of annotators. Meanwhile, DNNs suffer from overfitting noisy labels, degrading the performance of models. Therefore, in this work, we innovatively devise a noise-robust training approach to mitigate the adverse effects of noisy labels in medical image classification. Specifically, we incorporate contrastive learning and intra-group mixup attention strategies into vanilla supervised learning. The contrastive learning for feature extractor helps to enhance visual representation of DNNs. The intra-group mixup attention module constructs groups and assigns self-attention weights for group-wise samples, and subsequently interpolates massive noisy-suppressed samples through weighted mixup operation. We conduct comparative experiments on both synthetic and real-world noisy medical datasets under various noise levels. Rigorous experiments validate that our noise-robust method with contrastive learning and mixup attention can effectively handle with label noise, and is superior to state-of-the-art methods. An ablation study also shows that both components contribute to boost model performance. The proposed method demonstrates its capability of curb label noise and has certain potential toward real-world clinic applications.