Search

Vocal tract (VT) contour detection in real time MRI is a pre-stage to many speech production related applications such as articulatory analysis and synthesis. In this work, we present an algorithm for robust detection of keypoints on the vocal tract in rtMRI sequences using fully convolutional networks (FCN) via a heatmap regression approach. We as well introduce a spatio-temporal stabilization scheme based on a combination of Principal Component Analysis (PCA) and Kalman filter (KF) to extract stable landmarks in space and time. The proposed VT landmark detection algorithm generalizes well across subjects and demonstrates significant improvement over the state of the art baselines, in terms of spatial and temporal errors., NA

The goal of this study is to establish the analytical method of articulatory movement by real-time magnetic resonance images (rtMRI) without estimation of articulatory contours. We present the result of motion detection using a background subtraction method. As a result of applying the background subtraction method to the rtMRI data of one speaker, some motions were detected in tongue, lip, and lower jaw which are important places for speech generation. By the experiments with the movies of multiple speakers, we confirmed that it is possible to detect the motion of the basic articulatory movement, and some movements were different for each speaker. Furthermore, we adapted the proposed method for motion detection to the utterances aiming at the transmission of paralinguistic information. As a result, some similar movements to the previous research were observed.

Objective: We investigate the basic feasibility of estimating the brachial artery area-pressure relationship from MRI data obtained during pressure cuff inflations in-vivo. Methods: We acquired cross-sectional real-time MR images and cardiac-gated CINE MR images from the upper arm of a single male subject at rest during supra-systolic pressure cuff inflations and deflations. We estimate from the MR images the lumen area changes of the brachial artery, and, simultaneously, from the cuff pressure the systemic blood pressure of the subject. We reconstruct the area-pressure curve from two real-time and three CINE independent measurements. Results: The area-pressure curve can be reconstructed, and it is plausible and appears largely consistent with the literature using other methods. Conclusion: MR imaging during pressure cuff inflations is an easy to use, non-invasive candidate method to estimate the brachial artery pressure-area curve.

Spoken language training benefits from showing a video of native speakers' articulatory movements to train the second language learners. Typically, the articulatory video is prepared in conjunction with the audio which is collected simultaneously with the articulatory recording. Articulatory video recording requires specialized equipment and, hence, is expensive and time consuming. In this work, we propose a concatenative synthesis approach to obtain articulatory videos for an audio, which may not have a simultaneous articulatory recording. In the training stage of the proposed approach, we make a repository for phoneme specific articulatory image sequence from the available articulatory video. During testing, image sequences are selected from this repository to ensure a smooth transition across phonetic events. The selected image sequences are finally stitched to synthesize the articulatory video for the test audio. Articulatory videos are synthesized for 50 words randomly selected from the MRI-TIMIT database, not seen in the training data. Subjective evaluation on the quality of the synthesized videos using twelve subjects suggests that the videos are close to the original ones with a rating of 3.78 out of 5, where a score of 5 (1) indicates that there is no (great) difference in quality between the original and the synthesized videos.

In this paper, we propose a method that reconstruct the dynamic MRI from partial measurements by using a low-rank plus sparse regularization prior. It separated the dynamic MRI as a static background and dynamic organizations by robust principal component analysis. The minimization problem is solved by the alternating direction methods of multipliers (ADMM). The method is validated with numerical phantom simulations and cardiac MRI data against a direct Fourier inversion and k-t SLR algorithm. Results indicate this method improved the imaging quality with low computational time.

Magnetic resonance imaging (MRI) has been extensively used in clinical practice but suffers from long data acquisition time. Following the success of compressed sensing (CS) theory, many efforts have been made to accurately reconstruct MR images from undersampled k-space measurements and therefore dramatically reduce MRI scan time. To further improve image quality, we formulate undersampled MRI reconstruction as a least-squares optimization problem regularized by shearlet transform and overlapping-group sparsity-promoting total variation (OSTV). Shearlet transform, a directional representation system, is capable of capturing the optimal sparse representation for images with plentiful geometrical information. OSTV performs well in suppressing staircase-like artifacts often arising in traditional TV-based reconstructed images. To guarantee solution stability and efficiency, the resulting optimization problem is solved using an alternating direction methods of multipliers (ADMM)-based numerical algorithm. Extensive experimental results on both phantom and in vivo MRI datasets have demonstrated the superior performance of our proposed method in terms of both quantitative evaluation and visual quality.

Conventional phase-contrast (PC) MRI relies on electrocardiogram (ECG)-synchronized cine acquisition and respiration control. It often results in relatively low data acquisition efficiency, and is unable to assess blood flow variabilities. Real-time imaging is a promising technique to overcome these limitations; however, it results in a challenging image reconstruction problem with highly-undersampled (k; t)-space data. This paper presents a novel model-based imaging method, which integrates low-rank modeling with parallel imaging, to enable 4D real-time PC MRI without ECG gating and respiration control. The proposed method achieves an isotropic spatial resolution of 2.4 mm and temporal resolution of 35.2 ms, with three directional flow encodings. Moreover, it is able to resolve beat-by-beat flow variations, which cannot be achieved by the conventional cine-based approach. The proposed method was evaluated with in vivo experiments with one healthy subject and one arrhythmic patient. For the first time, we demonstrate the feasibility of 4D real-time PC MRI.

During a human brain MRI acquisition the resulting image is formed out of 2D slices. The slices must then be aligned and reconstructed to provide a 3-dimensional (3D) visualization of the brain volume. We propose in this work, an integrated system for the register ion and 3D reconstruction of DICOM MRI images and lesions of the brain acquired from multiple sclerosis (MS) subjects at two different time intervals (time 0 (T0) and time 1 (T1)). The system facilitates the follow up of the MS disease development and will aid the doctor to accurately manage the follow up of the disease. It involves a 6-stage analysis (preprocessing, lesion segmentation, registration, 3D reconstruction, volume estimation and method evaluation), as well as module quantitative evaluation of the method. The system was evaluated based on one MRI phantom and one DICOM MRI image of the brain. The accuracy of the proposed registration and reconstruction (- / -) method was 78.5%/97.2% and 95.4%/95.8% for the phantom and the MRI images respectively. These preliminary results provide evidence that the proposed system could be applied in future in the clinical practice.

Accurate 3D reconstruction of human tissue is a challenge problem in medical imaging. In this paper, a novel 3D reconstruction method of human brain MRI images is proposed based on the segmentation of human tissue. First, we propose a novel region-based growing algorithm to get points of an MRI image. Then, the moving cubes algorithm is used to reconstruct the accurate 3D object model. Further, in order to display well, a multiple angle observation is provided in our experiment. Experimental results show that the proposed method is better than the traditional methods.

International audience; In this paper we address the problem of bone segmentation in MRIimages of children, in the region of the pelvis. To cope with the complex structure of the bones in this region and their changing topologyduring growth, we propose a method relying on 3D bone templates.These models are built from 3D CT images. For a given MRI volume, the closest template is chosen and registered on the MRI data.This leads to an initial segmentation which is then refined using a deformable model approach, where the regularization parameters depend on the local curvature, and the landmarks used during the registration are fixed anchors during the deformation. This approach wassuccessfully applied to 15 MRI volumes of children between 1 and18 years old, with an average accuracy in terms of medium distanceof M D = 1.17 ± 0.29 mm and Dice Index of DC = 0.81 ± 0.04.

International audience; Example-based single image super-resolution (SR) has recently shown outcomes with high reconstruction performance. Several methods based on neural networks have successfully introduced techniques into SR problem. In this paper, we propose a three-dimensional (3D) convolutional neural network to generate high-resolution (HR) brain image from its input low-resolution (LR) with the help of patches of other HR brain images. Our work demonstrates the need of fitting data and network parameters for 3D brain MRI.

Automatic MR whole prostate segmentation is a challenging task. Recent approaches have attempted to harness the capabilities of deep learning for MR prostate segmentation to tackle pixel-level labeling tasks. Patch-based and hierarchical features-based deep CNN models were used to delineate the prostate boundary. To further investigate this problem, we introduce a Holistically-Nested Edge Detector (HED) MRI prostate deep learning segmentation and 3D surface reconstruction model that facilitate the registration of multi-parametric MRI with histopathology slides from radical prostatectomy specimens and targeted biopsy specimens. Application of this technique combines deep learning and computer aided design to provide a generalized solution to construct a high-resolution 3D prostate surface from MRI images in three orthogonal views. The performance of the segmentation is evaluated with MRI scans of 100 patients in 4-fold cross-validation. We achieve a mean Dice Similarity of 88.6%.

This talk describes a new adaptive variant of K-SVD dictionary learning that is suitable for the highly varied features present in MRI images. This variant is used to regularize SPIRiT parallel MRI reconstruction to produce higher quality images from incomplete measurements. The approach described involves modeling the dictionary approximation error as nearly sparse across the patches of the reconstructed image. This presentation will feature experiments reconstructing MRI data from multiple types of scans, including neuroimaging and cardiac imaging. These results demonstrate that the proposed method yields better reconstruction quality in both these settings. This talk will conclude with a discussion of related challenges in MRI reconstruction.

Noise is major concern in MRI images. There is always a demand for noise free and enhanced MRI images for accurate diagnosis of brain diseases. Thermal noise, Gaussian noise and Rician noise are some of the noises present in MRI images. The causes for these noises are patient condition, imaging system and unskilled manpower. This leads to poor quality of image, blurring, distortion and difficulty in detection and diagnosis of brain disease. Denoising is a process of removing the noise. The aim of this work is to understand the noise models and non-linear denoising techniques. Anisotropic Filter, Bilateral Filter and Non Local means filter are the non-linear filters used for denoising of MRI images. The performance of the filters is evaluated in terms of PSNR, Entropy, MSE, SSIM and average execution time. The Non-local Means filters gives better PSNR, MSE and SSI for Gaussian denoised MRI images.

To utilize the synergy between computed tomography (CT) and magnetic resonance imaging (MRI) data sets from an object at the same time, an edge-guided dual-modality image reconstruction approach is proposed. The key is to establish a knowledge-based connection between these two data sets for the tight fusion of different imaging modalities. Our scheme consists of four inter-related elements: 1) segmentation; 2) initial guess generation; 3) CT image reconstruction; and 4) MRI image reconstruction. Our experiments show that, aided by the image obtained from one imaging modality, even with highly under-sampled data, we can better reconstruct the image of the other modality. This approach can be potentially useful for a simultaneous CT-MRI system.

Segmentation of brain magnetic resonance imaging (MRI) data plays an important role in the computer-aided diagnosis and neuroscience research. Fuzzy c-means (FCM) clustering algorithm is one of the most usually used techniques for brain MRI image segmentation because of its fuzzy nature. However, the conventional FCM method fails to carry out segmentation well enough due to intensity inhomogeneity in MRI data. To overcome this issue, we propose an improved algorithm based on FCM clustering for segmentation of brain MRI data. Specifically, we modify the conventional FCM algorithm to allow for intensity inhomogeneity by introducing the regularization of the neighborhood influence and bias field. Results show that our proposed algorithm obtains reasonable segmentation of white matter (WM), gray matter (GM), and cerebrospinal fluid (CSF) from MRI data, which is superior to the expectation-maximization (EM) and conventional FCM methods.

In medical image processing, the accurate understanding and analysis of medical images is progressively challenging and demanding in providing the clear detection and diagnosis of the diseases. A very important and primary step in automatic detection of abnormalities in brain MRI images is reliable elimination of presence of the noise in the acquired images. Evaluation of the various noise removal algorithms to estimate the noise parameters is widely recognized as a typical standards for medical image processing solicitations. Validation of such brain MRI images can be done based on the evaluation through various statistical noise parameters. But the enactment of noise removal procedure profoundly depend on superiority of input brain MRI images and image processing skills employed for restoring original quality of brain MRI image by eliminating noises present in it and make the input image free from noises. This paper projects the type of noise existing in the brain MRI images and also removal of noises from the images using different types of noise elimination filters. The work proposes the design and implementation of the simple and efficient filtering technique for removal of noises in medical images specifically for brain MRI images and to upsurges the performance capabilities of the method. The experimental outcomes of suggested filtering algorithm are compared with the additional existing filtering approaches. The simulated results of filtering process executed on a standard set of assessment images shows that proposed algorithm provides good results with high SNR and low MSE values for the noise level of 95% and outperforms the conventional steps for removal of noise present in brain MRI images.

In recent years, with the development of medical equipment technology, tomograms have become important. As for Magnetic Resonance Imaging (MRI) in particular, study of the device that provides a tomographic image has become active. One technique to visualize blood vessels using MRI is called Magnetic Resonance Angiography (MRA). Fast Fourier transform (FFT) is used for analysis of the MRI, and the resolution of the FFT depends on the magnetic field strength. We focused on Non-harmonic Analysis (NHA). NHA has a frequency resolution that is much higher than FFT. Therefore, there is possibility that NHA can be visualized microstructure measured by ultra-high field MRI. In the experiment, we investigated the resolution of NHA by computer simulation. From the results, the FFT could not to express an original phantom, owing to the lack of resolution. On the other hand, NHA was able to exactly express an observation object and greatly reduce the occurrence of side lobes.

Many studies have been made in order to propose automatic diagnostic in medical fields. This paper proposes a new approach to deal with the problem of spectral clustering for signal extracted from brain MRI images. The tool-chain developed during this study can be easily implemented for the extraction and the analysis of information from perfusion MRI. In order to prove this, a reliable program which can easily isolate healthy from any pathological tissues. Experimental results are shown and discussed.

Magnetic Resonance Imaging (MRI) has been widely used in medical diagnose because of its non-invasive manner and excellent depiction of soft-tissue changes. Recently, the compressive sensing (CS) theory has been applied to reconstruct the MR image from highly down-sampled k-space data, which can reduce the scanning duration. To obtain useful information as much as possible with the same sampling rate, a weighted sampling strategy is studied. Moreover, based on the advantage of CS, a Wavelet tree based reconstruction approach is proposed. The experimental results demonstrate that the proposed method is preferable to other methods.

The authors present a statistical evaluation of the MRI images of a plant tissue. Therefore, we compare the commonly used MRI contrast and looking for the best contrast to evaluation of MRI data of a plant tissue. The acquired images are classified in terms of the signal-to-noise ratio, intensity difference and steepness of the edges. We focused on the manual evaluation of the statistical data in the area of the plant tissue.

Magnetic Resonance Imaging is a medical imaging modality used to produce good quality images of soft tissue in ligaments and other internal body organs. MRI is non-invasive scanning technique based on the principle of Nuclear Magnetic Resonance. The MRI scan time depends on the size of the scanned area and the number of images being reconstructed. This scan time reduction may reduce the artifacts in the reconstruction by improving the patient comfort. Compressed sensing (CS) theory helps MRI to reduce the scan time by reconstructing MR images with fewer sampled measurements. Application of CS to MRI gives acceleration in MR image acquisition. This paper focuses on randomly under sampled k-space data and use of CS-MR image reconstruction. This work compares variable density mask and Poisson mask and show their usefulness in Compressed Sensing applied to MRI image reconstruction. Image reconstruction using Nonlinear conjugate gradient method has been performed on the cardiac dataset at different acceleration factors. Further in the paper, reconstructed images are quantified by Peak Signal To Noise Ratio (PSNR).

Using magnetic resonance imaging (MRI) for guiding minimally invasive interventions requires surgical devices which on one hand are visible in the MR image but on the other hand do not generate large artifacts, which distort the overall imaging process. Passive markers are one way to visualize devices such as catheters or biopsy needles in MRI. The evaluation of newly developed passive markers usually requires access to high-field MRI scanners (1.5 T and 3 T). This makes the practical evaluation time-consuming and expensive. Hence, we propose to use a high-resolution, low field (0.55 T) benchtop MRI system to quantify the size of an artifact and to make a prediction for its corresponding size in a clinical high-field system. For the evaluation of the proposed method, catheters coated with different passive marker materials in varying concentrations were imaged in the 0. 55 T benchtop MRI scanner as well as in clinical 3 T MRI system using FLASH sequences. The experimental results revealed that an artifact prediction based on measurements in the 0. 55 T is possible for the tested marker materials. Hence, the proposed approach has a high potential for testing newly developed medical devices at a low cost, in less time and during the development process for fast feedback.

MRI (Magnetic resonance imaging, MRI) is a very important contrast in the clinical diagnosis, using radio wave pulses, so that the body of water molecules in the resonance of the hydrogen atom, and change the magnetic field generated by the echo signals into images by a computer. After a long clinical trial, have proved to be harmless. MRI is not invasive to human body, does not produce ionizing radiation, direction scans, three dimensional images and high contrast resolution and many other advantages, so the message so that it provides a large and rich set of organizations, but for medical personnel, inadequate amount large or will affect the results of judgments, so issues classification algorithms, become the focus of this study. Physician interpretation of a large number of medical imaging in the past with medical knowledge, modified extension now through the optimization method for physicians as a reference, MRI does substantially reduce the burden of multiple large spectrum of information on physicians, make Diagnostics more efficient, precise focus lies. Finally, K-meansTools for automatic segmentation and brain (FMRIB's Automated Segmentation Tool, FAST), and the use of receiver operating characteristics (Receiver Operating Characteristic, ROC) to effectiveness evaluation, and proved its superiority.

Three-dimensional brain imaging through cutting-edge MRI technology allows assessment of physical and chemical tissue properties at sub-millimeter resolution. In order to improve brain understanding as part of diagnostic tasks using MRI images, other imaging modalities to obtain deep cerebral structures and cytoarchitectural boundaries have been investigated. Under availability of postmortem samples, the fusion of MRI to brain histology supports more accurate description of neuroanatomical structures since it preserves microscopic entities and reveal fine anatomical details, unavailable otherwise. Nonetheless, histological processing causes severe tissue deformation and loss of the brain original 3D conformation, preventing direct comparisons between MRI and histology. This paper proposes an interactive computational pipeline designed to register multimodal brain data and enable direct histology-MRI correlation. Our main contribution is to develop schemes for brain data fusion, distortion corrections, using appropriate diffeomorphic mappings to align the 3D histological and MRI volumes. We describe our pipeline and preliminary developments of scalable processing schemes for highresolution images. Tests consider a postmortem human brain, and include qualitatively and quantitatively results, such as 3D visualizations and the Dice coefficient (DC) between brain structures. Preliminary results show promising DC values when comparing our scheme results to manually labeled neuroanatomical regions defined by a neurosurgeon on MRI and histology data sets. DC was computed for the left caudade gyrus (LC), right hippocampus (RH) and lateral ventricles (LV).

The brain tumor segmentation method of MRI images is of key importance for clinical analysis of glioma. The majority of existing methods are focused on structural MRI such as T1-weighted and T2-weighted. Additionally, functional MRI including Magnetic Resonance Spectroscopy (MRS), Diffusion Weighted Imaging (DWI), and Blood-Oxygen-Level Dependent (BOLD) can also contribute to increasing the validity and accuracy of the results. This paper proposes a framework of automatic brain tumor segmentation method based on information fusion of structural and functional signals. The method consists of four steps: intensity mapping for feature, region growing for tumor, region growing for edema and necrosis detection. The performance evaluation has been done by using some clinical MRI data with glioma. Comparing the segmentation results with the manual segmentation as “ground truth”, it has achieved average Dice score 83.7% in the tumor, and 88.5% in the whole lesion area, which indicated the validity and robustness of the proposed method.

This paper provides a brief survey of application of compressive sampling or compressive sensing (CS) in a major biomedical imaging application such as magnetic resonance imaging (MRI). MRI has a time-consuming data collection process. CS is an emerging mathematical technique which enables us to reconstruct the image or signal from very few samples. Recently severe efforts are made to apply CS to MRI to speed up the acquisition process. Reducing the amount of data reduces the scan time which in turn reduces the time of exposition of patient to the magnetic fields. Various optimization algorithms are proposed for the reconstruction of MR images from small set of available data without compromising with the image quality. Aim of this paper is to provide a brief analysis of current trends in CS applied to MRI.

Dynamic contrast enhanced (DCE) MRI is one of the important tools for diagnosis. In particular, time-resolved angiography with interleaved stochastic trajectories (TWIST) has been widely used thanks to the improved temporal and spatial resolution. However, the temporal resolution of TWIST is not a true one, because the periphery of k-space data from several time frames should be shared to produce one image. To address this issue, here we proposed a high performance low-rank k-space interpolation method that significantly improves the temporal resolution without sacrificing the image quality. The main idea of the proposed method is originated from the state-of-art MR reconstruction methods using annihilating filter-based low rank Hankel matrix approach (ALOHA). Experimental results using in-vivo data showed significantly improved temporal resolution compared with the standard reconstruction method.

Many limitations of diffusion MRI are due to the instability of the model fitting procedure. Major shortcomings of the model-based approach are a partial information loss due to model simplicity, long scan time requirements due to fitting instability, and the lack of knowledge about how the parameters of a given model would respond to previously unseen microstructural changes, possibly failing to detect certain previously unseen pathologies. Here we show that diffusion MRI pathology detection is feasible without any models and without any prior knowledge of specific pathological changes whatsoever. Instead, raw q-space measurements are used directly without a model, only healthy population data is used for reference, and any deviations in a patient dataset from the healthy reference database are detected using novelty detection methods. This is done in each voxel independently, i.e. without spatial bias.

Real-time MRI is a novel noninvasive imaging technique that allows the visualization of physiological processes with both good spatial and temporal resolutions. However, the reconstruction of images from highly undersampled data, needed to perform real-time imaging, remains challenging. Recently, the combination of Compressed Sensing theory with motion compensation techniques has shown to achieve better results than previous methods. In this work we describe a real-time MRI algorithm based on the acquisition of the k-space data following a Golden Radial trajectory, Compressed Sensing reconstruction and a groupwise temporal registration algorithm for the estimation and compensation of the motion in the image, all this embedded within a temporal multiresolution scheme. We have applied the proposed method to the reconstruction of free-breathing acquisition of short axis views of the heart, achieving a temporal resolution of 25ms, corresponding to an acceleration factor of 28 with respect to fully sampled Cartesian acquisitions.

Combining anatomical and functional information from Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET) scans brings great opportunities to improve diagnosis in oncology and treatment planning in radiation oncology. In this work, we propose a PET-guided MR tumor segmentation method minimizing a globally convex energy in a multiphase framework to account for the context variability of lesions. The method was evaluated in four patients with atypical meningiomas of different shapes, locations and metabolism, and the Dice index obtained by comparing with a manual tumor segmentation performed by an expert was 0.65 ± 0.13. The results demonstrated a good ability of the method to differentiate tumors from tissues with similar MRI intensity values.

We present a novel pipeline for acquisition protocol independent spine labeling in volumetric Magnetic Resonance Imaging (MRI) data of the lumbar spine. Our learning-based system uses local Entropy-optimized Texture Models (ETMs) for reducing the intensity scale in clinical data to only a few gray levels. The task of intervertebral disc localization is then performed on the normalized data. The benefit of our method is, that we can deal with various MRI protocols, such as T1-weighted (T1w) and T2-weighted (T2w) scans. Using the entropy objective allows us furthermore to apply the algorithm to acquisition protocols which are not covered by the training set. We achieve high disc localization accuracies for both, MRI protocols which are covered and not covered by training. The approach can be easily extended to other modalities.

A single MRI scan generates a large number of images of various cross sections of the body. This large set of accumulated data makes manual analysis time consuming thus a smart tool for screening is vital. This paper presents a novel classification and segmentation method which has the ability to identify white matter in MRI mages. Based on those findings, a supporting web based tool for the MRI image analysis and screening is introduced.

Biomedical image acquisition mechanism endows vital information pertaining to anatomy and supports contemporary quantitative image analysis which eases the work in diagnosing and treating the patients effectively. One of the frequently used tools for such diagnosis and treatment assessment has been Magnetic Resonance Imaging (MRI). However, the images acquired through MRI scans can adversely affect quantitative image analysis due to the noise hindrances. To overcome these, comparative study on the filtering methods namely Median, Gaussian and Gabor have been discussed based on the Performance Evaluating Parameters -PSNR, Energy, Entropy, Variance, Correlation and Contrast. Finally, the appropriate filtering method is identified for the image correction to bring in the evaluation strategies and application.

Increasing the resolution of the MRI scans results in lowering the signal-to-noise ratio and/or increasing the scan time. In this paper, we improved the resolution of MRI images by obtaining a high resolution image from a sequence of low resolution images using the super-resolution reconstruction methodology. Image restoration is an important step in the reconstruction process, the final appearance and the quality of the reconstructed image depend greatly on the restoration method used. In this paper, we evaluated the performance of three restoration techniques used in the reconstruction, the first one is based on truncated singular value decomposition, the second one is based on Tikhonov regularization and the last one is based on total variation. Experiments are performed on synthetic images and real MRI brain images to verify the results obtained from the numerical simulations.

Date of Conference: 16-19 May 2016 Conference Name: IEEE 24th Signal Processing and Communications Applications Conference, SIU 2016 Diffusion-weighted imaging (DWI) suffers from low SNR when compared to regular MRI images. To attain reasonable SNR levels with high spatial resolution, multiple acquisitions of each slice have to be acquired. These repetitions are then averaged, with the goal of increasing the SNR. However, subject motion during diffusion-sensitizing gradients creates varying phases between repeated acquisitions. When direct complex averaging is performed, these phase offsets can cause unpredictable phase cancellations and local signal drops. Here, we propose a technique that simultaneously corrects phase and performs a non-local means filtering to reduce noise. The success of the proposed technique is demonstrated in vivo at 3T with DWI of the cervical spinal cord.

MRI inter-slice gap is required during the image acquisition process in order to prevent leakage protons which cause blurred images. The gap is one of the major obstacles in image analysis, particularly in various modern computer aided diagnosis applications because the data in those inter-slice gaps disappear. In this paper, an inter-slice interpolation algorithm based on the projection on convex curve set (POCS) is proposed to reconstruct MRI voxels and fill inter-slice gaps by using joint information provided by different MRI viewing planes including axial, coronal and sagittal planes. Difference of pixel spacing of these sequences has been transformed and matched to the reference frame. Pixels from different planes are then located together in a unit voxel space matrix. They are then interpolated to create gapless MRI data. The proposed algorithm has been tested with five MRI datasets of bone tumor. The interpolated results are then compared with that of frequently used interpolation techniques including nearest neighbor, linear, and cubic interpolation. Interpolation results from these algorithms have been measured in terms of signal to noise ratio which represents image similarity to the original one. The comparison results showed that the proposed algorithm is superior to the others in which average similarity between the interpolated images and the original MRI images is highest.

The resolution of MRI images is limited due to several factors such as imaging hardware or time constraints. However, high MRI image resolution is desired in many medical applications. Traditional Super Resolution (SR) algorithms are generally unable to recover the high frequency (HF) information of MRI images. Recently, spatial adaptive SR algorithms have utilized the combined edge preserving and smoothness constraint methods to improve the quality of the images. Segmenting the image into edge and smooth blocks is a common step which adds to the complexity and execution time of these methods. This paper presents a fast SR technique for MRI images that preserve the edges and improve the visual quality of MRI images without segmenting the images. Experimental results prove the ability of this proposed approach with respect to the traditional SR methods in terms of better visual quality and less execution time.

Compressive Sensing Magnetic Resonance Imaging (CSMRI) exploits sparsity of medical images to reconstruct them accurately from undersampled k-space data. In this paper, we propose a novel patch-based nonlocal MRI reconstruction algorithm with low-rank regularization to exploit the structural sparsity of the observed data. In the proposed algorithm, the low-rank regularization is transformed into the nuclear norm minimization problem then the problem is solved by the Singular Value Thresholding (SVT) method with adaptive thresholds estimation and the Alternative Direction Multiplier Method(ADMM). Experimental results show the proposed MRI reconstruction method outperforms many existing algorithms in CSMRI.

The process of medical images using Magnetic resonance imaging (MRI) is important in medical areas; such as the anatomy of human soft tissue, and diagnosis of stroke or cancer. Using mathematical and computational transformations, frequency is transferred to receive MRI images. From the process of an MRI, a large amount of frequency data are obtained. However, to determine the final image, the entire frequency information is not needed. To decrease the size of the frequency domain in a relatively large k-space, and to produce an MRI image of a brain, data was gathered in certain ways by utilizing various filters. Through utilizing MRI filters based on proposed Gaussian filters such as y = exp(−((m−L/2)2)/ck), where m, L, c and k are all variables, and by adjusting the two main variables, c and k; relatively clear images with smaller k-space are able to be created. The constants c and k are tested from 10 and 1, increasing by 5 and 1, respectively.

The main objective of this paper is to present an analytical method to detect lesions (cysts) in digitized MRI data. Segmentation techniques are applied on different sequences of MRI images (T1&T2) which helps to differentiate between malignant region from normal region in the given original image. The abnormal part is captured in the JPEG format. The segmentation of the image is then used to detect the part of the image which depicts abnormalities more accurately. The proposed algorithm helps the radiologists to take primitive measures for diagnosis. An efficient method by integrating thresholding and canny edge detector has been explained in this paper. This process requires very less time and hence the method can detect the cyst in the early stage more accurately. Time complexity of the advanced segmentations is also discussed in this paper.

Now-a-days Bio-medical image processing is the most challenging and emerging field in medical diagnosis. Processing of MRI images is one of the crucial parts of this field. This paper focuses on the proposed strategy to detect and extract brain tumor from patient's MRI scan image. Detection and extraction of tumor from MRI scan images of the brain can be done by threshold segmentation method and the tumor clarity can be improved by duly mitigating the noise present in the processed image using Wiener filter. The tumor with improved lucidity will be extracted from input MRI image for diagnosis. For the purpose of analysis and processing real time patient data has been taken.

In this paper we have proposed an effective fusion method to combine the information of brain MRI and PET images to study Alzheimer's disease of the patients. The proposed methodology is applied to the wavelet based multi-scale image fusion algorithm. In this method principal component analysis (PCA) approach for integrating each RGB component of approximation images decomposed by multiresolution analysis provides superior visual interpretation and high-quality synthesis of MRI and PET images. The proposed rule overcomes the instability and inconsistency of information occurring in the wavelet based decomposition of images. The experimental results show that this method is very much suitable for multimodal images, especially for integrating structural and functional information of brain MRI and PET data.

The tumor location, evaluation of its shape, its position and its interactions with close brain structures, etc., are medical acts very important and in great demand. Clinical Tools brain's observation areas are now numerous, like X-ray scanner, magnetic resonance imaging (MRI), computed tomography and scintigraphy. Among these techniques, MRI is the increasingly used. This non-invasive technique exploits the magnetic properties of tissues and allows the observation of brain anatomy. The resulting image with high resolution allows the precise location of the various structures. Actually MRI technology allows to have digital images with image restoration "of the material" necessary for the implementation of processing and analysis algorithms that helps in the diagnosis of brain tumors. The real challenge is to accurately detect the tumor area. In this article we present a new method for detecting and segmenting tumor regions. This method consists of two steps: a) study and histogram analysis of the MRI image to find a threshold of intensity which can help us to extract tumor; b) tumor detection and extraction. Our method will be completed by a characterization of the tumor area by determining its geometric properties.

Multi-modality medical imaging has found its increasing application for better healthcare. The main advantage of multi-modality imaging is that the weakness of each modality are offset by the others. MRI and X-ray CT simultaneous imaging is a possible multi-modality imaging mode in clinical applications of the further. Both MRI and CT are widely used imaging modalities. Though both of them both belong to structure imaging, MRI and CT can provide complementary information about patient anatomy. Usually, MRI may offer good soft tissue information while CT depicts better quality image on bones. Hence, the idea of CT-MRI hybrid imaging integrated in one machine was put forward, which is expected to provide high image quality in both soft tissue and bones. Moreover, because CT scanning is much fast than MRI, it is possible to greatly shorten the scanning time of MRI by reconstructing the MRI image from the highly under-sampled data and the information of the CT image at the same tissues. This paper focuses on the problem of multi-modalities medical image reconstruction problem (MRI-CT), and proposes a self-adaptive mask-enhanced dual-dictionary learning (DDL) for MRI image reconstruction with highly under-sampled measurement data and training data of well-registered MRI and CT images. Its main feature is to establish a one-to-one mapping relationship between MRI image and CT image of the same tissues, and obtain a good initial value of MRI imaging from the CT data. Numerical simulations were performed to evaluate and validate the proposed algorithm.

4D MRI is current gaining attention as an imaging modality which is able to capture inter-cycle variability of respiratory motion. Such information is beneficial for example in radiotherapy planning and delivery. However the motion extracted from 4D MRI is dependant on the method of processing the images. Typically a deformable registration algorithm is used so that the 3D motion of every point in the image can be found. We have used several popular registration algorithms which have been recently published with publicly available data. Our findings suggest that due to the low contrast of 4D MRI, the registration is only accurate near clear boundaries, and less accurate elsewhere, to different degrees depending on the algorithm used. We thus suggest modelling organ deformation as an affine transformation, which is then embedded into the deformation field for more accurate modelling of respiratory motion.

Simultaneous PET/MRI shows great promise for improved detection and diagnosis of disease. Commercially available whole body combined PET/MRI systems are currently prohibitively expensive, limiting access to the technology. MRI-compatible PET inserts present a lower-cost alternative, adding PET capabilities to existing MRI installations. A brain sized radio-frequency (RF)-penetrable PET insert has been designed for simultaneous operation within MRI systems. This insert takes advantage of electro-optical coupling and battery power to electrically float the PET insert relative to the MRI ground, permitting RF signals to be transmitted and received through small gaps between the PET modules. Non-magnetic silicon photomultipliers are used in conjunction with a compressed sensing signal multiplexing scheme, and optical fibers transmit analog PET detector signals out of the MRI room for decoding, processing, and image reconstruction. Results of the first simultaneous PET/MR imaging studies with the full RF-penetrable PET ring are presented in this work. The PET insert was placed within a 3T whole body MRI system, and tomographic images of a 3D-printed resolution phantom with hot and cold rods of varying sizes were acquired both with only the B 0 field present, and under continuous pulsing from three MR imaging sequences: gradient echo (GRE), fast spin echo (FSE), and echo planar imaging (EPI). The resulting PET images have comparable contrast-to-noise ratios (CNR) under all pulsing conditions, with an average (of all rod types) maximum CNR difference of 14.0% among images. MR images of the resolution phantom were successfully acquired through the RF-penetrable PET shielding using only the built in MR body coil, with mean CNR decreases of 19.5% and 29.8% for GRE and FSE, respectively. These results suggest that simultaneous imaging is possible with the RF-penetrable insert, and show promise for this technology as an alternative to costly integrated PET/MRI scanners.

In this work we proposed a new respiratory and cardiac motion correction method for cardiac PET/MR imaging by using 2D-MRI image navigator technique and novel dynamic MRI reconstruction method. In the PET/MR acquisition, ECG trigger is used. After trigger starts, 2D-MRI image navigator and golden angle cardiac imaging sequence are applied. After acquisition, cardiac MRI data and PET data are binned to different motion phases according to 2D-MRI image navigator and ECG trigger. We use the high down sampling acquisition and novel reconstruction method based on Low-rank and Sparsity decomposition to reduce MRI acquisition time and improve the time resolution. Finally, we estimate the motion field with both cardiac and respiratory motion based on the reconstructed MRI image series. And PET images are reconstructed by OSEM algorithm based on the binned data. After reconstruction, the PET images are deformed to the same motion phase by accurate motion field from MRI. Finally, sum up all the deformed images to get corrected PET image. PET data is simulated via analytical projection method using segmented MRI images. The results of the proposed method are compared to methods with only respiratory and without motion correction. The simulation shows that the proposed method could improve the image quality and the accuracy of lesion estimation especially on the ventricular wall with cardiac motion. In conclusion, the simulation result demonstrated the feasibility and improvement of the new method.

Compressed sensing utilizes the sparsity of Magnetic resonance (MR) images to obtain accurate reconstructions from undersampled k-space data. In this paper, a novel nonlocal dynamic MRI reconstruction method with low-rank regularization is developed to exploit the spatiotemporal structural sparsity of a MRI sequence. The nonlocal prior and low rank prior are combined organically by grouping similar patches in both spatial and temporal domain. The low-rank regularization can be approximated by nuclear norm minimization solved by a singular value thresholding (SVT) method with adaptive thresholds estimation. The objective function is divided into several sub-problems that are easier to solve by alternative direction multiplier method (ADMM). Extensive experiments show that the new method outperforms commonly used classical dynamic MRI reconstruction algorithms.