Ultrasound Speckle Reduction Using Wavelet-Based Generative Adversarial Network.

Speckle noise removal plays a crucial role in ultrasound (US) image diagnosis, since the visual quality of the US images are largely corrupted by speckle noise. Numerous speckle noise removal techniques have been proposed in the literature based on anisotropic filtering, wavelets and morphology; however they have some major problems like loss of edge information, texture information and inability to remove low frequency noise. Despeckling of US images is usually carried out using conventional anisotropic diffusion or speckle reducing anisotropic diffusion. However, despeckling US images may not be able to preserve the edges which comprises of important clinical information. To overcome the issues in speckle noise removal (despeckling) of US images, contourlet transform based anisotropic nonlinear diffusion filtering is proposed in this paper. Contourlet transform improves some important features like multiscale and directionality. Adaptive nonlinear diffusion has been incorporated in anisotropic filtering to improve the filtering performance. The comparison performance of the proposed method with other despeckling techniques indicates that it has better noise removal performance for medical US images.

Ultrasound (US) imaging is considered as one of the most advanced diagnostic tools in medical use. However, a drawback of the medical US imaging is its poor quality of the image, which is affected by speckle noise. In this paper, a MultiLayer Back-Propagation Neural Networks (MLBPNNs) with the Epanechnikov fuzzy function is proposed to reduce the speckle, and while at the same time, enhance the lesion boundaries of the US image. The main goal of the proposed method is to improve the quality of US image so as to improve the quality of the humans interpretation and the computer systems autoedge detection. In order to automatically detect the lesion boundary by a computer system, an edge enhancement is required. Evaluating the simulation results by Peak Signal to Noise Ratio (PSNR), Normalized Mean Square Error (NMSE), Detail Variance (DV), and Background Variance (BV), the proposed method demonstrates an increased performance of reducing the speckle and enhancing the edge. The proposed method has higher PSNR than conventional methods and can remove the speckle sufficiently, so that tumor boundaries of real US breast tumor image could be preserved and detected.

In this paper, we propose to apply generative adversarial neural networks trained with a cycle consistency loss, or CycleGANs, to improve realism in ultrasound (US) simulation from computed tomography (CT) scans. A ray-casting US simulation approach is used to generate intermediate synthetic images from abdominal CT scans. Then, an unpaired set of these synthetic and real US images is used to train CycleGANs with two alternative architectures for the generator, a U-Net and a ResNet. These networks are finally used to translate ray-casting based simulations into more realistic synthetic US images. Our approach was evaluated both qualitatively and quantitatively. A user study performed by 21 experts in US imaging shows that both networks significantly improve realism with respect to the original ray-casting algorithm ([Formula: see text]), with the ResNet model performing better than the U-Net ([Formula: see text]). Applying CycleGANs allows to obtain better synthetic US images of the abdomen. These results can contribute to reduce the gap between artificially generated and real US scans, which might positively impact in applications such as semi-supervised training of machine learning algorithms and low-cost training of medical doctors and radiologists in US image interpretation.

Ultrasound (US) images are useful in medical diagnosis. US is preferred over other medical diagnosis technique because it is non-invasive in nature and has low cost. The presence of speckle noise in US images degrades its usefulness. A method that reduces the speckle noise in US images can help in correct diagnosis. This method also should preserve the important structural information in US images while removing the speckle noise. In this paper, a method for removing speckle noise using a combination of wavelet, total variation (TV) and morphological operations has been proposed. The proposed method achieves denoising by combining the advantages of the wavelet, TV and morphological operations along with the utilization of adaptive regularization parameter which controls the amount of smoothing during denoising. The work in this paper has the capability of reducing speckle noise while preserving the structural information in the denoised image. The proposed method demonstrates strong denoising for synthetic and real ultrasound images, which is also supported by the results of various quantitative measures and visual inspection.

A versatile technique for visual enhancement of medical ultrasound images

Segmenting the lesion region from 2D ultrasound (US) images is an importance step in defining the intra-operative planning of computer-assisted therapy. However, it still remains a challenge owing to two reasons: (1) The signal-to-noise ratio (SNR) of the US images is originally low, and the inhomogeneous distribution of intensity causes local fuzzy boundaries. (2) The deformation of soft tissue causes makes the geometry contour of the segmentation area complex. Because of these factors, the classical segmentation methods do not yield the desired results in US images. To solve these problems, a novel segmentation method based on a dynamics statistical shape model is proposed. In this study, first, a radial time series model was defined for use as a parametric shape model of the lesion region, and a stochastic difference equation (SDE) was used as the contour dynamics generating model; then, the coefficients of the dynamics generating model and the shape prior probability were deduced by utilizing the property of Fokker-Planck equation (FPE) and training the samples. This statistical shape model (SSM) was referred to as the SF-SSM. Furthermore, a radial profile feature model (RPFM) was developed based on the statistical features of the intensity distribution from the inside to outside of the lesion region and based on the radial time series model. The RPFM provided a likelihood probability for the prior shape. Finally, through the segmentation optimization process, the maximum posteriori process was determined. In order to verify the performance of our method, we used real clinical ultrasound images as the training and test sets. We compared the proposed model with three well-known segmentation methods for the same test set. The experimental results showed that the proposed method provided more accurate results than the other methods for different quality US images. Furthermore, the friendly initialization of the proposed method improves the efficiency of interactive segmentation, thus benefitting computer-assisted therapy.

Despeckling of ultrasound images, as a very active topic research in medical image processing, plays an important or even indispensable role in subsequent ultrasound image processing. The non-local total variation (NLTV) method has been widely applied to denoising images corrupted by Gaussian noise, but it cannot provide satisfactory restoration results for ultrasound images corrupted by speckle noise. To address this problem, a novel non-local total variation despeckling method is proposed for speckle reduction. In the proposed method, the non-local gradient is computed on the images restored by the optimized Bayesian non-local means (OBNLM) method and it is introduced into the total variation method to suppress speckle in ultrasound images. Comparisons of the restoration performance are made among the proposed method and such state-of-the-art despeckling methods as the squeeze box filter (SBF), the non-local means (NLM) method and the OBNLM method. The quantitative comparisons based on synthetic speckled images show that the proposed method can provide higher Peak signal-to-noise ratio (PSNR) and structure similarity (SSIM) than compared despeckling methods. The subjective visual comparisons based on synthetic and real ultrasound images demonstrate that the proposed method outperforms other compared algorithms in that it can achieve better performance of noise reduction, artifact avoidance, edge and texture preservation.

Studies of medical flow imaging have technical limitations for accurate analysis of blood flow dynamics and vessel wall interaction at arteries. We propose a new deep learning-based boundary detection and compensation (DL-BDC) technique in ultrasound (US) imaging. It can segment vessel boundaries by harnessing the convolutional neural network and wall motion compensation in the analysis of near-wall flow dynamics. The network enables training from real and synthetic US images together. The performance of the technique is validated through synthetic US images and tissue-mimicking phantom experiments. The neural network performs well with high Dice coefficients of over 0.94 and 0.9 for lumens and walls, outperforming previous segmentation techniques. Then, the performance of the wall motion compensation is examined for compliant phantoms. When DL-BDC is applied to flow influenced by wall motion, root-mean-square errors are less than 0.07%. The technique is utilized to analyze flow dynamics and wall interaction with varying elastic moduli of the phantoms. The results show that the flow dynamics and wall shear stress values are consistent with the expected values of the compliant phantoms, and their wall motion behavior is observed with pulse wave propagation. This strategy makes US imaging capable of simultaneous measurement of blood flow and vessel dynamics in human arteries for their accurate interaction analysis. DL-BDC can segment vessel walls fast, accurately, and robustly. It enables to measure the near-wall flow precisely by determining the vessel boundary dynamics. This approach can be beneficial in flow dynamics and wall interaction analyses in various biomedical applications.

Ultrasound (US) imaging is a widely used clinical diagnostic tool in medical imaging techniques. It is a comparatively safe, economical, painless, portable, and noninvasive real-time tool compared to the other imaging modalities. However, the image quality of US imaging is severely affected by the presence of speckle noise and blur during the acquisition process. In order to ensure a high-quality clinical diagnosis, US images must be restored by reducing their speckle noise and blur. In general, speckle noise is modeled as a multiplicative noise following a Rayleigh distribution and blur as a Gaussian function. Hereto, we propose an intelligent estimator based on artificial neural networks (ANNs) to estimate the variances of noise and blur, which, in turn, are used to obtain an image without discernible distortions. A set of statistical features computed from the image and its complex wavelet sub-bands are used as input to the ANN. In the proposed method, we solve the inverse Rayleigh function numerically for speckle reduction and use the Richardson-Lucy algorithm for de-blurring. The performance of this method is compared with that of the traditional methods by applying them to a synthetic, physical phantom and clinical data, which confirms better restoration results by the proposed method.

The possibility to perform an early and repeatable assessment of imaging performance is fundamental in the design and development process of new ultrasound (US) probes. Particularly, a more realistic analysis with application-specific imaging targets can be extremely valuable to assess the expected performance of US probes in their potential clinical field of application. The experimental protocol presented in this work was purposely designed to provide an application-specific assessment procedure for newly-developed US probe prototypes based on Capacitive Micromachined Ultrasonic Transducer (CMUT) technology in relation to brain imaging. The protocol combines the use of a bovine brain fixed in formalin as the imaging target, which ensures both realism and repeatability of the described procedures, and of neuronavigation techniques borrowed from neurosurgery. The US probe is in fact connected to a motion tracking system which acquires position data and enables the superposition of US images to reference Magnetic Resonance (MR) images of the brain. This provides a means for human experts to perform a visual qualitative assessment of the US probe imaging performance and to compare acquisitions made with different probes. Moreover, the protocol relies on the use of a complete and open research and development system for US image acquisition, i.e. the Ultrasound Advanced Open Platform (ULA-OP) scanner. The manuscript describes in detail the instruments and procedures involved in the protocol, in particular forthe calibration, image acquisition and registration of US and MR images. The obtained results prove the effectiveness of the overall protocol presented, which is entirely open (within the limits of the instrumentation involved), repeatable, and covers the entire set of acquisition and processing activities for US images.

Cross-modal Transfer Learning Based on an Improved CycleGAN Model for Accurate Kidney Segmentation in Ultrasound Images

Histology-Based Simulations of Ultrasound Imaging: Methodology

Medical ultrasound imaging has transformed the disease identification in the human body in the last few decades. The major setback for ultrasound medical images is speckle noise. Speckle noise is created in ultrasound images due to numerous reflections of ultrasound signals from hard tissues of human body. Speckle noise corrupts the medical ultrasound images dropping the detectable quality of the image. An endeavor is made to recover the image quality of ultrasound medical images by using block based hard and soft thresholding of wavelet coefficients. Medical ultrasound image is transformed to wavelet domain using debauchee's mother wavelet. Divide the approximate and detailed coefficients into uniform blocks of size 8×8, 16×16, 32×32 and 64×64. Hard and soft thresholding on these blocks of approximate and detailed coefficients are applied. Inverse transformation to original spatial domain produces a noise reduced ultrasound image. Experiments were conducted on medical ultrasound images obtained from diagnostic centers in Vijayawada, India. Quality of improved images in measured using peak signal to noise ratio (PSNR), image quality index (IQI), structural similarity index (SSEVI).

Noninvasive ultrasound (US) brain therapies, including US brain stimulation, US drug delivery by BBB opening, or US thrombolysis, have great potential in clinical applications. But target positioning is a challenge. Dual-mode US system provides a promising solution, which can carry out US therapy function under the guidance of US imaging. However, the brain skull usually deteriorates the brain image seriously, because of the great amplitude of US echoes reflected from the skull. In order to improve the guidance function of US image, we investigated the relationship of skull structures and US image in rodents. Computerized tomography (CT) images of the heads of four rats were scanned by a micro-CT system. US skull imaging was carried out by a custom-built US imaging and stimulation system with a dual-mode 2D matrix array with 1 MHz center frequency and 256 elements. The features of the CT images and the US images of rat skulls were carefully recognized, which were used to identify the coordinates of the rat brain from US images. After the US imaging and brain position was determined, multi-target BBB opening was carried out based on the identified coordinate. The result indicated that the deviation of the coordinate was less than 1 mm. US skull imaging will be a useful method for image-guided noninvasive therapy applications.

Chapter 14 - Despeckling of ultrasound images based on the multiresolution approach and Gaussianization transform