NLMap-ATVR: A novel combination of nonlinear mapping network and adaptive total variation regularization for MRI denoising.

Hyperspectral sparse unmixing based on a novel adaptive total variation regularization

Electrical capacitance tomography (ECT) is a nonintrusive imaging technique for monitoring dynamics within a closed container. In this paper, an image reconstruction algorithm for ECT, termed Adaptive Total Variation Regularization (ATVR), is introduced. The ATVR algorithm reduces total image error while preventing loss of features of interest in fluid-air interface monitoring. Unlike traditional total variation regularization for ECT, ATVR allows the regularization factor to change during image reconstruction allowing the algorithm to achieve higher accuracy at discontinuous boundaries (such as fluid interfaces) for the same overall image error. ATVR doesn't require additional iteration steps compared to traditional TVR and so is equally well-suited for the online monitoring of rapidly flowing fluids. Experimental analysis has been conducted on a custom-built ECT test bed and confirms the simulation results and demonstrates the application of the algorithm to the case where a dielectric working fluid (oil) is undergoing steady pumping.

Hyperspectral sparse fusion using adaptive total variation regularization and superpixel-based weighted nuclear norm

Denoising of MR images with Rician noise using a wider neural network and noise range division

The last quarter century has witnessed major advancements that have brought biomedical imaging to a paramount status in life sciences. Generally speaking, the scope of biomedical imaging covers data acquisition, image reconstruction, and image analysis, involving theories, methods, systems, and applications. While many kinds of imaging modalities, such as X-ray computed tomography (CT) and magnetic resonance imaging (MRI), become increasingly sophisticated, the mathematical methods involved in these modalities play more and more critical roles in further improving their performance in anatomical, functional, cellular, and molecular applications. The overall goal of this issue is to promote research and development of biomedical imaging by publishing high-quality research articles in this rapidly growing interdisciplinary field. Due to the time limit, this special issue mainly focused on 4 kinds of biomedical imaging modalities: CT, MRI, ultrasound, and fluorescence imaging; several biomedical image processing methods were also involved. Each paper published in this special issue was reviewed by at least two reviewers and revised according to reviewer's comments. For CT imaging, A. Cai et al. developed an efficient iterative image reconstruction (IIR) algorithm, using cone beam CT reconstruction that is based on total-variation (TV) minimization to overcome the computational complexity of IIR scheme in cone beam CT reconstruction; L.-z. Deng et al. proposed a hybrid reconstruction method combining TV and nonaliasing contourlet transform (NACT) and using the Split-Bregman method to solve the optimization problem. This algorithm utilized the geometrical information of CT image and got a sparser representation compared with wavelet and gradient operator. For MRI imaging, in order to reduce time consuming in MRI image reconstruction, Q. Li et al. proposed a parallel computing method which was based on a novel patch-based nonlocal operator (PANO). Simulation results demonstrated that this method can accelerate PANO-based MRI reconstruction several times compared with original one. W. He et al. introduced a direct nonconvex Lp norm algorithm for MRI phase unwrapping which leaded to faithful phase correction. Also analytical high order tensor decomposition was introduced into crossing fibers detection in diffusion MRI by T. Megherbi et al., which provided a better angular resolution and accuracy than the classical maxima localization method. For biomedical image processing, R. Jaramillo et al. used a wavelet domain filter to improve the performance of the Prony method. In this work, MRI images were considered to be affected only by Rician noise, and a new wavelet domain bilateral filtering process was implemented to reduce the noise in the images. X. Wang et al. proposed a model that allows robustness to noise as well for handling the intensity inhomogeneity and weak boundary problems in medical image segmentation using region-scalable discriminant and fitting energy for image segmentation. Also, a region-based active contour model which introduced a total energy as a penalty function in medical image segmentation is proposed by T. Liu et al. Another active contours based segmentation method was proposed by F. Akram et al. for intensity inhomogeneous MRI images to enable boundaries detection of the homogenous regions. Not only CT and MRI but also other medical imaging modalities, such as ultrasound and fluorescence imaging, are included in this special issue. S.-K. Ueng et al. designed a special filter aiming at suppressing speckles and enhancing features in the ultrasound images. In this method, diffusion tensor of intensity at each pixel was represented in the form of a Hessian matrix which was used to compute eigen values at each pixel. The eigen values were used in detection and classifying the underlying structure and a refinement strategy was followed to improve the classification. Then, based on the computed structure types, feasible filters were selected from a filter pool to suppress speckles and enhance features. X. Lin et al. used three-dimensional fluorescent spectra imaging to investigate whether and how Tubeimoside 1 (TBMS 1) can affect HepG2 cells, which indicated that fluorescent spectra method is a promising substitute for flow cytometry in cancer research. A computational model to estimate fluence rate for a biological medium with inclusion was developed by M. Gantri. In this work, the entire setting of the medium was treated to have spatially and stochastically varying refractive index to match practical applications and Legendre integral transform technique is incorporated to solve the radiative transfer equation. These papers represent an insightful observation into the state of the art, as well as future topics in this biomedical imaging field. We hope that this special issue would attract a wide attention of the peers and provide a chance to share the latest research work. Peng Feng Kumar Durai Fenglin Liu Xiaobo Qu

In this paper, we propose the local complexity estimation based filtering method in wavelet domain for MRI (magnetic resonance imaging) denoising. A threshold selection methodology is proposed in which the edge and detail preservation properties for each pixel are determined by the local complexity of the input image. In the proposed filtering method, the current wavelet kernel is compared with a threshold to identify the signal- or noise-dominant pixels in a scale providing a good visual quality avoiding blurred and over smoothened processed images. We present a comparative performance analysis with different wavelets to find the optimal wavelet for MRI denoising. Numerical experiments and visual results in simulated MR images degraded with Rician noise demonstrate that the proposed algorithm consistently outperforms other denoising methods by balancing the tradeoff between noise suppression and fine detail preservation. The proposed algorithm can enhance the contrast between regions allowing the delineation of the regions of interest between different textures or tissues in the processed images. The proposed approach produces a satisfactory result in the case of real MRI denoising by balancing the detail preservation and noise removal, by enhancing the contrast between the regions of the image. Additionally, the proposed algorithm is compared with other approaches in the case of Additive White Gaussian Noise (AWGN) using standard images to demonstrate that the proposed approach does not need to be adapted specifically to Rician or AWGN noise; it is an advantage of the proposed approach in comparison with other methods. Finally, the proposed scheme is simple, efficient and feasible for MRI denoising.

Vessel segmentation is important for many clinical applications, such as the diagnosis of vascular diseases, the planning of surgery, or the monitoring of the progress of disease. Although various approaches have been proposed to segment vessel structures from 3-dimensional medical images, to the best of our knowledge, there has been no known technique that uses magnetic resonance imaging (MRI) as prior information within the vessel segmentation of magnetic resonance angiography (MRA) or magnetic resonance venography (MRV) images. In this study, we propose a novel method that uses MRI images as an atlas, assuming that the patient has an MRI image in addition to MRA/MRV images. The proposed approach intends to increase vessel segmentation accuracy by using the available MRI image as prior information. We use a rigid mutual information registration of the MRA/MRV to the MRI, which provides subvoxel accurate multimodal image registration. On the other hand, vessel segmentation methods tend to mostly suer from imaging artifacts, such as Rician noise, radio frequency (RF) inhomogeneity, or partial volume eects that are generated by imaging devices. Therefore, this proposed method aims to extract all of the vascular structures from MRA/MRI or MRV/MRI pairs at the same time, while minimizing the combined eects of noise and RF inhomogeneity. Our method is validated both quantitatively and visually using BrainWeb phantom images and clinical MRI, MRA, and MRV images. Comparison and observer studies are also realized using the BrainWeb database and clinical images. The computation time is markedly reduced by developing a parallel implementation using the Nvidia compute unied device architecture and OpenMP frameworks in order to allow the use of the method in clinical settings.

Denoising of MR (Magnetic Resonance) images is one of the crucial tasks in the field of medicine as it is often corrupted by Rician noise which exhibits a signal dependent behaviour at different SNR. Due to the biased nature of Rician noise, it is a challenging task to attenuate Rician noise from MR images. In this paper, we analyse the basic and effective MRI denoising techniques. Also, we have proposed a two-stage denoising procedure which has a strong potential to remove Rician noise from MR images effectively. The efficacy of proposed two-stage MRI denoising method is analysed through qualitative and quantitative measures and it is observed that the performance of the proposed two-stage denoising procedure is superior to single-stage MRI denoising methods.

CNN-DMRI: A Convolutional Neural Network for Denoising of Magnetic Resonance Images

Objective To investigate the way to accurately delineate gross tumor volume (GTV) of high grade gliomas(HGG) for intensity modulated radiation therapy (IMRT) by using computed tomography (CT) and magnetic resonance imaging (MRI) image fusion technique. Methods CT and MRI images were fused from 19 patients. The GTV of each patient were independently delineated by one chief doctor and one resident doctor on CT and MRI image. The GTV contoured on CT (GTVCT), MRI (GTVMRI) were measured, and composite volumes (GTVCT+MRI) were the sum of CT-defined GTV and MRI-defined GTV. The differences of these volumes were compared. Results Whether chief or resident doctors delineated, all were GTVMRI ＞GTVCT(P ＜0.050). The percentages of GTVMRI on GTVCT+MRI were (98.57±7.00)% by chief doctors, and (97.84±10.00)% by resident doctors. Compared the difference between GTVCT and GTVMRI in postoperative patients and preoperative patients, P =0.046, and the difference between chief doctors and resident doctors was statistically significant for GTV defined by CT (P =0.020), but not by MRI and composite image (P ＞0.050).Conclusion The GTV of HGG patients must be delineated on both CT image and MRI image, including using CT and MRI image fusion. But the composite volumes(GTVCT+MRI) should be the sum of CT-defined GTV and MRI-defined GTV. Especially for the postoperative patients,delineating GTV should be taken more attention. And the GTV should be delineated by doctors with full experiences. Key words: Glioma; Tomography, X-ray computed; Magnetic resonance imaging; Radiotherapy,intensity-modulated; Gross tumor volume

Malignant Mimics of Trigeminal Schwannoma

A critical step in predicting and avoiding radiation injury of organs at risk in radiation therapy of nasopharyngeal carcinoma is to carry out an accurate dose evaluation in planning design. In the present study, we investigated the dose evaluation feature of organs at risk on magnetic resonance imaging (MRI) images in intensity‐modulated radiation therapy of nasopharyngeal carcinoma compared with computed tomography (CT) images. A total of 35 nasopharyngeal carcinoma patients were selected for this trial. CT simulation with non‐contrast and contrast‐enhanced scan, and MRI simulation with non‐contrast and contrast‐enhanced T1, T2, and diffusion weighted imaging were obtained sequentially. The organs at risk were contoured on the CT and MRI images after rigid registration, respectively. Nine‐beam intensity‐modulated radiation therapy plans with equal division angles were designed for every patient, and the prescription dose for the tumor target was set as 72 Gy (2.4Gy/fraction). The boundary display, volume, and dosimetric indices of each organ were compared between MRI and CT images. We found that MRI showed clearer boundary of the brainstem, spinal cord, deep lobe of the parotid gland, and the optical nerve in the canal compared with CT. MRI images increased the volume of the lens and optic nerve, while slightly reducing the volume of eye; the maximum dose of the lens, and the mean dose of the eyes and optic nerve increased to different extents, though no statistical differences were found. The left and right parotid gland volume on MRI increased by 7.07% and 8.13%, and the mean dose increased by 14.95% (4.01 Gy) and 18.76% (4.95 Gy), with a statistically significant difference (P < 0.05). The brainstem volume reduced by 9.33% (P < 0.05), and the dose of 0.1 cm3 volume reduced by a mean 8.46% (4.32 Gy), whereas the dose of 0.1 cm3 of the spinal cord increased by 1.5 Gy on MRI. The maximum dose region of the spinal cord was very close on CT and MRI images, and was similar to the brainstem. In conclusion, it is credible to evaluate the radiation dose of the lens, eye, brainstem, and the spinal cord by applying simulation CT; whereas MRI images are sometimes necessary to evaluate the dose of the parotid glands and the optical nerve.

Magnetic resonance images (MRI) plays a crucial role in neuroscience and medical diagnosis. Denoising MRI images is an important preprocessing step required in many of the automatic computed aided-diagnosis systems in neuroscience. Rician noise occurs in the MRI image during acquisition. Non local mean filter is used for denoising. But the parameter selection is not optimized. The proposed method removal of rician noise in MRI images using bilateral filter by fuzzy trapezoidal membership function improves the denoising efficiency at various noise variances, preserves the fine structures and edges. The fuzzy weights were obtained with the statistical features such as local mean (μ i ) and global mean (μ g ) by constructing trapezoidal membership function. Bilateral filter is used to preserve the edges by smoothening the noises in MRI image and preserves the structural information. Local filter preserves the edges. MRI images are restored by multiplying its corresponding fuzzy weight with the restored image of local order filter and bilateral filter. Experiments on simulated and real MRI data were done at different noise levels by the proposed method and the existing methods. The result shows that the proposed method restores the image in better visual quality and can be well utilized for diagnostic purpose at both low and high densities of rician noise.

Simple SummaryConventional magnetic resonance imaging (MRI) sequences have known limitations in target delineation for radiation treatment (RT) planning of cerebral gliomas. Advanced physiology-based MRI techniques and radionuclide imaging techniques, including positron emission tomography (PET) with amino acid radiopharmaceuticals, may increase the specificity for glioma tissue characterization. Our work aims to provide a comprehensive review of the advanced MRI and PET imaging modalities that can complement conventional MRI for RT planning of gliomas. A detailed overview of their basic principles and clinical results is given based on the most updated literature.The accuracy of target delineation in radiation treatment (RT) planning of cerebral gliomas is crucial to achieve high tumor control, while minimizing treatment-related toxicity. Conventional magnetic resonance imaging (MRI), including contrast-enhanced T1-weighted and fluid-attenuated inversion recovery (FLAIR) sequences, represents the current standard imaging modality for target volume delineation of gliomas. However, conventional sequences have limited capability to discriminate treatment-related changes from viable tumors, owing to the low specificity of increased blood-brain barrier permeability and peritumoral edema. Advanced physiology-based MRI techniques, such as MR spectroscopy, diffusion MRI and perfusion MRI, have been developed for the biological characterization of gliomas and may circumvent these limitations, providing additional metabolic, structural, and hemodynamic information for treatment planning and monitoring. Radionuclide imaging techniques, such as positron emission tomography (PET) with amino acid radiopharmaceuticals, are also increasingly used in the workup of primary brain tumors, and their integration in RT planning is being evaluated in specialized centers. This review focuses on the basic principles and clinical results of advanced MRI and PET imaging techniques that have promise as a complement to RT planning of gliomas.

In order to improve the quality of blind image restoration, we propose an algorithm which combines Non-negativity and Support constraint Recursive Inverse Filtering (NAS-RIF) and adaptive total variation regularization. In the proposed algorithm, the total variation regularization constraint term is added in the NAS-RIF algorithm cost function. The majorization-minimization approach and conjugate gradient iterative algorithm are adopted to improve the convergence speed. We do the simulation experiments for the blurred classic test image which is added additive random noise. Experimental results show that the restoration effect of our algorithm is better than the spatially adaptive Tikhonov regularization method and the NAS-RIF spatially adaptive regularization algorithm, while the value of improvement of signal to noise ratio (ISNR) has improved.