RESUMEN

The problem surrounding convolutional neural network robustness and noise immunity is currently of great interest. In this paper, we propose a technique that involves robustness estimation and stability improvement. We also examined the noise immunity of convolutional neural networks and estimated the influence of uncertainty in the training and testing datasets on recognition probability. For this purpose, we estimated the recognition accuracies of multiple datasets with different uncertainties; we analyzed these data and provided the dependence of recognition accuracy on the training dataset uncertainty. We hypothesized and proved the existence of an optimal (in terms of recognition accuracy) amount of uncertainty in the training data for neural networks working with undefined uncertainty data. We have shown that the determination of this optimum can be performed using statistical modeling. Adding an optimal amount of uncertainty (noise of some kind) to the training dataset can be used to improve the overall recognition quality and noise immunity of convolutional neural networks.

Asunto(s)

RESUMEN

Supercontinuum (SC) sources offer high illumination power from a single mode fiber with large spectral bandwidth including the visible spectrum, a growing application area for Optical Coherence Tomography (OCT). However, SC spectra suffer from pulse-to-pulse variations, increasing noise in the resulting images. By simultaneously collecting a normalization spectrum, OCT image noise can be reduced by more than half (7 dB) for single pulses without any pulse averaging using only simple optical components.

RESUMEN

In this paper, we have applied a recently developed complex-domain hyperspectral denoiser for the object recognition task, which is performed by the correlation analysis of investigated objects' spectra with the fingerprint spectra from the same object. Extensive experiments carried out on noisy data from digital hyperspectral holography demonstrate a significant enhancement of the recognition accuracy of signals masked by noise, when the advanced noise suppression is applied.

RESUMEN

We present a novel tomographic non-local-means based despeckling technique, TNode, for optical coherence tomography. TNode is built upon a weighting similarity criterion derived for speckle in a three-dimensional similarity window. We present an implementation using a two-dimensional search window, enabling the despeckling of volumes in the presence of motion artifacts, and an implementation using a three-dimensional window with improved performance in motion-free volumes. We show that our technique provides effective speckle reduction, comparable with B-scan compounding or out-of-plane averaging, while preserving isotropic resolution, even to the level of speckle-sized structures. We demonstrate its superior despeckling performance in a phantom data set, and in an ophthalmic data set we show that small, speckle-sized retinal vessels are clearly preserved in intensity images en-face and in two orthogonal, cross-sectional views. TNode does not rely on dictionaries or segmentation and therefore can readily be applied to arbitrary optical coherence tomography volumes. We show that despeckled esophageal volumes exhibit improved image quality and detail, even in the presence of significant motion artifacts.

RESUMEN

Optical coherence tomography (OCT) is a non-invasive technique with a large array of applications in clinical imaging and biological tissue visualization. However, the presence of speckle noise affects the analysis of OCT images and their diagnostic utility. In this article, we introduce a new OCT denoising algorithm. The proposed method is founded on a numerical optimization framework based on maximum-a-posteriori estimate of the noise-free OCT image. It combines a novel speckle noise model, derived from local statistics of empirical spectral domain OCT (SD-OCT) data, with a Huber variant of total variation regularization for edge preservation. The proposed approach exhibits satisfying results in terms of speckle noise reduction as well as edge preservation, at reduced computational cost.

RESUMEN

Optical coherence tomography (OCT) is based on coherence detection of interferometric signals and hence inevitably suffers from speckle noise. To remove speckle noise in OCT images, wavelet domain thresholding has demonstrated significant advantages in suppressing noise magnitude while preserving image sharpness. However, speckle noise in OCT images has different characteristics in different spatial scales, which has not been considered in previous applications of wavelet domain thresholding. In this study, we demonstrate a noise adaptive wavelet thresholding (NAWT) algorithm that exploits the difference of noise characteristics in different wavelet sub-bands. The algorithm is simple, fast, effective and is closely related to the physical origin of speckle noise in OCT image. Our results demonstrate that NAWT outperforms conventional wavelet thresholding.

RESUMEN

We propose using maximum a-posteriori (MAP) estimation to improve the image signal-to-noise ratio (SNR) in polarization diversity (PD) optical coherence tomography. PD-detection removes polarization artifacts, which are common when imaging highly birefringent tissue or when using a flexible fiber catheter. However, dividing the probe power to two polarization detection channels inevitably reduces the SNR. Applying MAP estimation to PD-OCT allows for the removal of polarization artifacts while maintaining and improving image SNR. The effectiveness of the MAP-PD method is evaluated by comparing it with MAP-non-PD, intensity averaged PD, and intensity averaged non-PD methods. Evaluation was conducted in vivo with human eyes. The MAP-PD method is found to be optimal, demonstrating high SNR and artifact suppression, especially for highly birefringent tissue, such as the peripapillary sclera. The MAP-PD based attenuation coefficient image also shows better differentiation of attenuation levels than non-MAP attenuation images.

RESUMEN

Optical coherence tomography (OCT) has become one of the most successful optical technologies implemented in medicine and clinical practice mostly due to the possibility of non-invasive and non-contact imaging by detecting back-scattered light. OCT has gone through a tremendous development over the past 25 years. From its initial inception in 1991 [Science254, 1178 (1991)] it has become an indispensable medical imaging technology in ophthalmology. Also in fields like cardiology and gastro-enterology the technology is envisioned to become a standard of care. A key contributor to the success of OCT has been the sensitivity and speed advantage offered by Fourier domain OCT. In this review paper the development of FD-OCT will be revisited, providing a single comprehensive framework to derive the sensitivity advantage of both SD- and SS-OCT. We point out the key aspects of the physics and the technology that has enabled a more than 2 orders of magnitude increase in sensitivity, and as a consequence an increase in the imaging speed without loss of image quality. This speed increase provided a paradigm shift from point sampling to comprehensive 3D in vivo imaging, whose clinical impact is still actively explored by a large number of researchers worldwide.

RESUMEN

Singular value decomposition (SVD) was used to identify and remove laser-induced noise in photoacoustic images acquired with a clinical ultrasound scanner. This noise, which was prominent in the radiofrequency data acquired in parallel from multiple transducer elements, was induced by the excitation light source. It was modelled by truncating the SVD matrices so that only the first few largest singular value components were retained, and subtracted prior to image reconstruction. The dependency of the signal amplitude and the number of the largest singular value components used for noise modeling was investigated for different photoacoustic source geometries. Validation was performed with simulated data and measured noise, and with photoacoustic images acquired from the human forearm and finger in vivo using L14-5/38 and L40-8/12 linear array clinical imaging probes. The use of only one singular value component was found to be sufficient to achieve near-complete removal of laser-induced noise from reconstructed images. This method has strong potential to increase image quality for a wide range of photoacoustic imaging systems with parallel data acquisition.

RESUMEN

Assessment of the cardiovascular parameters using noncontact video-based or imaging photoplethysmography (IPPG) is usually considered as inaccurate because of strong influence of motion artefacts. To optimize this technique we performed a simultaneous recording of electrocardiogram and video frames of the face for 36 healthy volunteers. We found that signal disturbances originate mainly from the stochastically enhanced dichroic notch caused by endogenous cardiovascular mechanisms, with smaller contribution of the motion artefacts. Our properly designed algorithm allowed us to increase accuracy of the pulse-transit-time measurement and visualize propagation of the pulse wave in the facial region. Thus, the accurate measurement of the pulse wave parameters with this technique suggests a sensitive approach to assess local regulation of microcirculation in various physiological and pathological states.

RESUMEN

Phase image in Fourier domain Doppler optical coherence tomography offers additional flow information of investigated samples, which provides valuable evidence towards accurate medical diagnosis. High quality phase images are thus desirable. We propose a noise reduction method for phase images by combining a synthetic noise estimation criteria based on local noise estimator (LNE) and distance median value (DMV) with anisotropic diffusion model. By identifying noise and signal pixels accurately and diffusing them with different coefficients respectively and adaptive iteration steps, we demonstrated the effectiveness of our proposed method in both phantom and mouse artery images. Comparison with other methods such as filtering method (mean, median filtering), wavelet method, probabilistic method and partial differential equation based methods in terms of peak signal-to-noise ratio (PSNR), equivalent number of looks (ENL) and contrast-to-noise ratio (CNR) showed the advantages of our method in reserving image energy and removing noise.

RESUMEN

Time-correlated single photon counting (TCSPC) enables acquisition of fluorescence lifetime decays with high temporal resolution within the fluorescence decay. However, many thousands of photons per pixel are required for accurate lifetime decay curve representation, instrument response deconvolution, and lifetime estimation, particularly for two-component lifetimes. TCSPC imaging speed is inherently limited due to the single photon per laser pulse nature and low fluorescence event efficiencies (<10%) required to reduce bias towards short lifetimes. Here, simulated fluorescence lifetime decays are analyzed by SPCImage and SLIM Curve software to determine the limiting lifetime parameters and photon requirements of fluorescence lifetime decays that can be accurately fit. Data analysis techniques to improve fitting accuracy for low photon count data were evaluated. Temporal binning of the decays from 256 time bins to 42 time bins significantly (p<0.0001) improved fit accuracy in SPCImage and enabled accurate fits with low photon counts (as low as 700 photons/decay), a 6-fold reduction in required photons and therefore improvement in imaging speed. Additionally, reducing the number of free parameters in the fitting algorithm by fixing the lifetimes to known values significantly reduced the lifetime component error from 27.3% to 3.2% in SPCImage (p<0.0001) and from 50.6% to 4.2% in SLIM Curve (p<0.0001). Analysis of nicotinamide adenine dinucleotide-lactate dehydrogenase (NADH-LDH) solutions confirmed temporal binning of TCSPC data and a reduced number of free parameters improves exponential decay fit accuracy in SPCImage. Altogether, temporal binning (in SPCImage) and reduced free parameters are data analysis techniques that enable accurate lifetime estimation from low photon count data and enable TCSPC imaging speeds up to 6x and 300x faster, respectively, than traditional TCSPC analysis.

RESUMEN

Images rendered by uncooled microbolometer-based infrared (IR) cameras are severely degraded by the spatial non-uniformity (NU) noise. The NU noise imposes a fixed-pattern over the true images, and the intensity of the pattern changes with time due to the temperature instability of such cameras. In this paper, we present a novel model and a compensation algorithm for the spatial NU noise and its temperature-dependent variations. The model separates the NU noise into two components: a constant term, which corresponds to a set of NU parameters determining the spatial structure of the noise, and a dynamic term, which scales linearly with the fluctuations of the temperature surrounding the array of microbolometers. We use a black-body radiator and samples of the temperature surrounding the IR array to offline characterize both the constant and the temperature-dependent NU noise parameters. Next, the temperature-dependent variations are estimated online using both a spatially uniform Hammerstein-Wiener estimator and a pixelwise least mean squares (LMS) estimator. We compensate for the NU noise in IR images from two long-wave IR cameras. Results show an excellent NU correction performance and a root mean square error of less than 0.25 ∘ C, when the array's temperature varies by approximately 15 ∘ C.

RESUMEN

Photoplethysmography (PPG)-imaging is an emerging noninvasive technique that maps spatial blood-volume variations in living tissue with a video camera. In this paper, we clarify how cardiac-related (i.e., ballistocardiographic; BCG) artifacts occur in this imaging modality and address these using algorithms from the remote-PPG literature. Performance is assessed under stationary conditions at the immobilized hand. Our proposal outperforms the state-of-the-art, blood pulsation imaging [Biomed. Opt. Express5, 3123 (2014). ], even in our best attempt to create diffused illumination. BCG-artifacts are suppressed to an order of magnitude below PPG-signal strength, which is sufficient to prevent interpretation errors.

RESUMEN

One of the main drawbacks of Digital Holography (DH) is the coherent nature of the light source, which severely corrupts the quality of holographic reconstructions. Although numerous techniques to reduce noise in DH have provided good results, holographic noise suppression remains a challenging task. We propose a novel framework that combines the concepts of encoding multiple uncorrelated digital holograms, block grouping and collaborative filtering to achieve quasi noise-free DH reconstructions. The optimized joint action of these different image-denoising methods permits the removal of up to 98% of the noise while preserving the image contrast. The resulting quality of the hologram reconstructions is comparable to the quality achievable with non-coherent techniques and far beyond the current state of art in DH. Experimental validation is provided for both single-wavelength and multi-wavelength DH, and a comparison with the most used holographic denoising methods is performed.

RESUMEN

A description of the rocking curve in diffraction enhanced imaging (DEI) is presented in terms of the angular signal response function and a simple multi-information retrieval algorithm based on the cosine function fitting. A comprehensive analysis of noise properties of DEI is also given considering the noise transfer characteristic of the X-ray source. The validation has been performed with synchrotron radiation experimental data and Monte Carlo simulations based on the Geant4 toolkit combined with the refractive process of X-rays, which show good agreement with each other. Moreover, results indicate that the signal-to-noise ratios of the refraction and scattering images are about one order of magnitude better than that of the absorption image at the edges of low-Z samples. The noise penalty is drastically reduced with the increasing photon flux and visibility. Finally, this work demonstrates that the analytical method can build an interesting connection between DEI and GDPCI (grating-based differential phase contrast imaging) and is widely suitable for a variety of measurement noise in the angular signal response imaging prototype. The analysis significantly contributes to the understanding of noise characteristics of DEI images and may allow improvements to the signal-to-noise ratio in biomedical and material science imaging.

RESUMEN

Over the years, many computed optical interferometric techniques have been developed to perform high-resolution volumetric tomography. By utilizing the phase and amplitude information provided with interferometric detection, post-acquisition corrections for defocus and optical aberrations can be performed. The introduction of the phase, though, can dramatically increase the sensitivity to motion (most prominently along the optical axis). In this paper, we present two algorithms which, together, can correct for motion in all three dimensions with enough accuracy for defocus and aberration correction in computed optical interferometric tomography. The first algorithm utilizes phase differences within the acquired data to correct for motion along the optical axis. The second algorithm utilizes the addition of a speckle tracking system using temporally- and spatially-coherent illumination to measure motion orthogonal to the optical axis. The use of coherent illumination allows for high-contrast speckle patterns even when imaging apparently uniform samples or when highly aberrated beams cannot be avoided.

RESUMEN

In fluorescence fluctuation polarization sensitive experiments, the limitations associated with detecting the rotational timescale are usually eliminated by applying fluorescence correlation spectroscopy analysis. In this paper, the variance of the time-averaged fluorescence intensity extracted from the second moment of the measured fluorescence intensity is analyzed in the short time limit, before fluctuations resulting from rotational diffusion average out. Since rotational correlation times of fluorescence molecules are typically much lower than the temporal resolution of the system, independently of the time bins used, averaging over an ensemble of time-averaged trajectories was performed in order to construct the time-averaged intensity distribution, thus improving the signal-to-noise ratio. Rotational correlation times of fluorescein molecules in different viscosities of the medium within the range of the anti-bunching time (1-10 ns) were then extracted using this method.

RESUMEN

In this paper, we make contact with the field of compressive sensing and present a development and generalization of tools and results for reconstructing irregularly sampled tomographic data. In particular, we focus on denoising Spectral-Domain Optical Coherence Tomography (SDOCT) volumetric data. We take advantage of customized scanning patterns, in which, a selected number of B-scans are imaged at higher signal-to-noise ratio (SNR). We learn a sparse representation dictionary for each of these high-SNR images, and utilize such dictionaries to denoise the low-SNR B-scans. We name this method multiscale sparsity based tomographic denoising (MSBTD). We show the qualitative and quantitative superiority of the MSBTD algorithm compared to popular denoising algorithms on images from normal and age-related macular degeneration eyes of a multi-center clinical trial. We have made the corresponding data set and software freely available online.