Medical Image Reconstruction Research Articles

Reconstruction of Medical Images Using Artificial Bee Colony Algorithm

The goal of this study is to assess the efficiency of Artificial Bee Colony (ABC) algorithm in finding the optimal solution of curve fitting problem specifically for medical images. Data of Computed Tomography (CT) images from two different patients were collected. The procedure of curve fitting for medical images include conversion of Digital and Communications in Medicine (DICOM) images to binary images, boundary and corner point detection, parameterization, and curve reconstruction by using ABC algorithm. Then, Sum Square Error (SSE) was used to calculate the distance of the fitted Cubic Bezier curve with the boundary of the original images. Based on the calculation and parameter tuning that had been done, the smallest error of both skulls is57.5754and28.8628, respectively. The finding of this study illustrated that the proposed method had efficiently produced fitted Bezier curve that resemble the original medical images. In addition, the used of Douglas Peucker algorithm helps to improve the performance of the proposed method since computational time can be minimized. This study had shown that the proposed method can be used as an alternative method in order to reconstruct or redesigned the medical images since it produces a small error. For future work, we are planning to explore and applied the ABC algorithm to reconstruct the missing part of the skull since it can reduce the time taken to produce the skull implant as well as reducing the cost of producing it.

Quantitative susceptibility mapping using deep neural network: QSMnet

Deep neural networks have demonstrated promising potential for the field of medical image reconstruction, successfully generating high quality images for CT, PET and MRI. In this work, an MRI reconstruction algorithm, which is referred to as quantitative susceptibility mapping (QSM), has been developed using a deep neural network in order to perform dipole deconvolution, which restores magnetic susceptibility source from an MRI field map. Previous approaches of QSM require multiple orientation data (e.g. Calculation of Susceptibility through Multiple Orientation Sampling or COSMOS) or regularization terms (e.g. Truncated K-space Division or TKD; Morphology Enabled Dipole Inversion or MEDI) to solve an ill-conditioned dipole deconvolution problem. Unfortunately, they either entail challenges in data acquisition (i.e. long scan time and multiple head orientations) or suffer from image artifacts. To overcome these shortcomings, a deep neural network, which is referred to as QSMnet, is constructed to generate a high quality susceptibility source map from single orientation data. The network has a modified U-net structure and is trained using COSMOS QSM maps, which are considered as gold standard. Five head orientation datasets from five subjects were employed for patch-wise network training after doubling the training data using a model-based data augmentation. Seven additional datasets of five head orientation images (i.e. total 35 images) were used for validation (one dataset) and test (six datasets). The QSMnet maps of the test dataset were compared with the maps from TKD and MEDI for their image quality and consistency with respect to multiple head orientations. Quantitative and qualitative image quality comparisons demonstrate that the QSMnet results have superior image quality to those of TKD or MEDI results and have comparable image quality to those of COSMOS. Additionally, QSMnet maps reveal substantially better consistency across the multiple head orientation data than those from TKD or MEDI. As a preliminary application, the network was further tested for three patients, one with microbleed, another with multiple sclerosis lesions, and the third with hemorrhage. The QSMnet maps showed similar lesion contrasts with those from MEDI, demonstrating potential for future applications.

Leveraging similarity joins for signal reconstruction

Signal reconstruction problem (SRP) is an important optimization problem where the objective is to identify a solution to an underdetermined system of linear equations that is closest to a given prior. It has a substantial number of applications in diverse areas including network traffic engineering, medical image reconstruction, acoustics, astronomy and many more. Most common approaches for SRP do not scale to large problem sizes. In this paper, we propose a dual formulation of this problem and show how adapting database techniques developed for scalable similarity joins provides a significant speedup. Extensive experiments on real-world and synthetic data show that our approach produces a significant speedup of up to 20x over competing approaches.

State-of-the-Art Technologies in Nuclear Medicine Imaging

Nuclear medicine imaging is an important tool for cancer diagnosis, brain research, molecular imaging research and so on. Therefore, various imaging techniques and methods are being developed and investigated in nuclear medicine physics. In this report, we introduce state-of-the-art techniques, such as Compton camera imaging, time-of-flight positron emission tomography, semiconductor detectors for medical applications, image reconstruction and deep learning, which were reported in the 2017 IEEE Nuclear Science Symposium & Medical Imaging Conference.

Parallel Multi-Core CPU and GPU for Fast and Robust Medical Image Watermarking

Securing medical images are a very essential process in medical image authentication. Medical image watermarking is a very popular tool to achieve this goal. In this paper, an extremely fast, highly accurate, and robust algorithm is proposed for watermarking both gray-level and color medical images. In the proposed method, a scrambled binary watermark is embedded in the host medical image. Simplified exact kernels are used to compute the moments of the polar complex exponential transform (PCET) for the host gray-level images and the moments of the quaternion PCET for the host color images without approximation errors. The stability of the computed moments enables us to use higher order moments in a perfect reconstruction of the watermarked medical images. The accurate moment invariant to rotation, scaling, and translation ensures the robustness of the proposed watermarking algorithm against geometric attacks. Performed experiments clearly show very high visual imperceptibility and robustness to different levels of geometric distortions and common signal processing attacks. The implementation of parallel multi-core CPU and GPU result in a tremendous reduction of the overall watermarking times. For a color image of size $256\times256$ , the watermarking time is accelerated by $20\times $ and $11\times $ using a GPU and a CPU with 16 cores, respectively.

Scalable pCT Image Reconstruction Delivered as a Cloud Service

We describe a cloud-based medical image reconstruction service designed to meet a real-time and daily demand to reconstruct thousands of images from proton cancer treatment facilities worldwide. Rapid reconstruction of a three-dimensional Proton Computed Tomography (pCT) image can require the transfer of 100 GB of data and use of approximately 120 GPU-enabled compute nodes. The nature of proton therapy means that demand for such a service is sporadic and comes from potentially hundreds of clients worldwide. We thus explore the use of a commercial cloud as a scalable and cost-efficient platform for pCT reconstruction. To address the high performance requirements of this application we leverage Amazon Web Services’ GPU-enabled cluster resources that are provisioned with high performance networks between nodes. To support episodic demand, we develop an on-demand multi-user provisioning service that can dynamically provision and resize clusters based on image reconstruction requirements, priorities, and wait times. We compare the performance of our pCT reconstruction service running on commercial cloud resources with that of the same application on dedicated local high performance computing resources. We show that we can achieve scalable and on-demand reconstruction of large scale pCT images for simultaneous multi-client requests, processing images in less than 10 minutes for less than $10 per image.

Learning-based 3D surface optimization from medical image reconstruction

Mesh optimization has been studied from the graphical point of view: It often focuses on 3D surfaces obtained by optical and laser scanners. This is despite the fact that isosurfaced meshes of medical image reconstruction suffer from both staircases and noise: Isotropic filters lead to shape distortion, while anisotropic ones maintain pseudo-features. We present a data-driven method for automatically removing these medical artifacts while not introducing additional ones. We consider mesh optimization as a combination of vertex filtering and facet filtering in two stages: Offline training and runtime optimization. In specific, we first detect staircases based on the scanning direction of CT/MRI scanners, and design a staircase-sensitive Laplacian filter (vertex-based) to remove them; and then design a unilateral filtered facet normal descriptor (uFND) for measuring the geometry features around each facet of a given mesh, and learn the regression functions from a set of medical meshes and their high-resolution reference counterparts for mapping the uFNDs to the facet normals of the reference meshes (facet-based). At runtime, we first perform staircase-sensitive Laplacian filter on an input MC (Marching Cubes) mesh, and then filter the mesh facet normal field using the learned regression functions, and finally deform it to match the new normal field for obtaining a compact approximation of the high-resolution reference model. Tests show that our algorithm achieves higher quality results than previous approaches regarding surface smoothness and surface accuracy.

Implementation of 2-Dimensional Medical Image Reconstruction Algorithm for Photoacoustic Imaging

In this paper, we review and discuss our proposed image reconstruction algorithm for two dimensional stereo photo-acoustic medical imaging system. The proposed algorithms aim to evaluate the existence possibility of acoustic source within a search area by using the geometric information which is the distance between each sensor element of ultrasound probe and corresponding searching point. Besides, we implemented the hardware and software aspects of an photo-acoustic imaging system that is described in hardware implementation section of this paper. At the end of this paper, some simulation results are presented which utilized the proposed algorithm while compared the image quality of its and conventional k-wave algorithm in which FFT should be used. The simulation results proved the effectiveness of the proposed reconstruction algorithm.

Variational Bayesian blind restoration reconstruction based on shear wave transform for low-dose medical CT image

A variational Bayesian blind restoration reconstruction based on shear wave transform for low-dose medical computed tomography (CT) image is proposed. The proposed algorithm eliminates the effects of the point spread function in the process of low-dose medical CT image reconstruction and improves the reconstructed image quality. The shear wave transform is used to sparsely represent the CT image, which can speed up the efficiency of image processing. In the Bayesian framework, a posteriori probability objective function with unknown parameters is constructed. These unknown parameters include the shear wave coefficients, the point spread function, and the hyper parameters. It specified a Laplacian distribution model for the prior probability distribution of the shear wave coefficients. The autoregressive model is used as the prior model of the point spread function. All of hyper parameters follow the gamma distribution. The variational Bayesian method is used to estimate all of unknown parameters and solve the above posteriori probability objective function. These generalized parameter estimators are used to realize the low-dose CT image blind restoration reconstruction. Computer simulation results indicate that a good performance-reconstructed image can be obtained and some metrics such as peak signal-to-noise ratio (PSNR), universal image quality index (UIQI), structural similarity index metric (SSIM), and sum of square differences error (SSDE) are improved.

Convolutional Neural Networks for Inverse Problems in Imaging: A Review

In this survey paper, we review recent uses of convolution neural networks (CNNs) to solve inverse problems in imaging. It has recently become feasible to train deep CNNs on large databases of images, and they have shown outstanding performance on object classification and segmentation tasks. Motivated by these successes, researchers have begun to apply CNNs to the resolution of inverse problems such as denoising, deconvolution, super-resolution, and medical image reconstruction, and they have started to report improvements over state-of-the-art methods, including sparsity-based techniques such as compressed sensing. Here, we review the recent experimental work in these areas, with a focus on the critical design decisions: Where does the training data come from? What is the architecture of the CNN? and How is the learning problem formulated and solved? We also bring together a few key theoretical papers that offer perspective on why CNNs are appropriate for inverse problems and point to some next steps in the field.

A generalized entropy-based two-phase threshold algorithm for noisy medical image edge detection

Edge detection in medical imaging is a significant task for object recognition of human organs and it is considered a pre-processing step in medical image segmentation and reconstruction. This article proposes an efficient approach based on generalized Hill entropy to find a good solution for detecting edges under noisy conditions in medical images. The proposed algorithm uses a two-phase thresholding: firstly, a global threshold is calculated by means of generalized Hill entropy and used to separate the image into object and background. Afterwards, a local threshold value is determined for each part of the image. The final edge map image is a combination of these two separate images based on the three calculated thresholds. The performance of the proposed algorithm is compared against Canny and Tsallis entropy by using sets of medical images corrupted with various types of noise. We used Pratt’s Figure of Merit (PFOM) as a quantitative measure for an objective comparison. Experimental results indicate that the proposed algorithm displayed superior noise resilience and better edge detection than Canny and Tsallis entropy methods for the four different types of noise analyzed, and thus it can be considered as a very interesting edge detection algorithm on noisy medical images.

Adaptive wavelet tight frame construction for accelerating MRI reconstruction

The sparsity regularization approach, which assumes that the image of interest is likely to have sparse representation in some transform domain, has been an active research area in image processing and medical image reconstruction. Although various sparsifying transforms have been used in medical image reconstruction such as wavelet, contourlet, and total variation (TV) etc., the efficiency of these transforms typically rely on the special structure of the underlying image. A better way to address this issue is to develop an overcomplete dictionary from the input data in order to get a better sparsifying transform for the underlying image. However, the general overcomplete dictionaries do not satisfy the so-called perfect reconstruction property which ensures that the given signal can be perfectly represented by its canonical coefficients in a manner similar to orthonormal bases, resulting in time consuming in the iterative image reconstruction. This work is to develop an adaptive wavelet tight frame method for magnetic resonance image reconstruction. The proposed scheme incorporates the adaptive wavelet tight frame approach into the magnetic resonance image reconstruction by solving a l0-regularized minimization problem. Numerical results show that the proposed approach provides significant time savings as compared to the over-complete dictionary based methods with comparable performance in terms of both peak signal-to-noise ratio and subjective visual quality.

Super resolution reconstruction of medical image based on adaptive quad-tree decomposition

Super resolution reconstruction of medical image based on adaptive quad-tree decomposition

Relaxed two points projection method for solving the multiple-sets split equality problem

The multiple-sets split equality problem, a generalization and extension of the split feasibility problem, has a variety of specific applications in real world, such as medical care, image reconstruction, and signal processing. It can be a model for many inverse problems where constraints are imposed on the solutions in the domains of two linear operators as well as in the operators’ ranges simultaneously. Although, for the split equality problem, there exist many algorithms, there are but few algorithms for the multiple-sets split equality problem. Hence, in this paper, we present a relaxed two points projection method to solve the problem; under some suitable conditions, we show the weak convergence and give a remark for the strong convergence method in the Hilbert space. The interest of our algorithm is that we transfer the problem to an optimization problem, then, based on the model, we present a modified gradient projection algorithm by selecting two different initial points in different sets for the problem (we call the algorithm as two points algorithm). During the process of iteration, we employ subgradient projections, not use the orthogonal projection, which makes the method implementable. Numerical experiments manifest the algorithm is efficient.

An algorithm for learning sparsifying transforms of multidimensional signals

Multidimensional signals contain information of an object in more than one dimension, and usually their processing relies on complex methods in comparison with their unidimensional counterparts. In signal processing, finding a sparse representation of a signal is of great importance for compression purposes. Analytical multidimensional bases such as the Fourier, Cosine, or Wavelet Transform have been conventionally used. Recently, the use of learned dictionaries that directly adapt to the given signal are becoming popular in tasks such as image classification, image denoising, spectral unmixing, and medical image reconstruction. This paper presents an algorithm to learn transformation bases for the sparse representation of multidimensional signals. The proposed algorithm alternates between a sparse coding step solved by hard or soft thresholding strategies, and an updating dictionary step solved by a conjugate gradient method. Furthermore, the algorithm is tested using both: two-dimensional and three-dimensional patches, which are compared in terms of the sparsity performance for different types of multidimensional signals such as hyperspectral images, computerized axial tomography images and, magnetic resonance images. The attained results are compared against traditional analytical transforms and the state-of-the-art dictionary learning method: K-SVD.

Strong convergence theorems for split inclusion problems in Hilbert spaces

The split feasibility problem is an inverse problem which arises in signal processing and medical image reconstruction. So there is practical value in studying it. While both the split equality problem and the split variational inclusion problem are generalized form of the split feasibility problem which are more meaningful than the split feasibility problem. In this paper, fusing the two problems, we research a split inclusion problem and propose relevant methods for solving it. What counts is that not only the proposed algorithms have strong convergence, but also the limit points of the algorithms are the minimal norm solution of the split inclusion problem.

Voxelisation in the 3-D Fly Algorithm for PET

Abstract The Fly Algorithm was initially developed for 3-D robot vision applications. It consists in solving the inverse problem of shape reconstruction from projections by evolving a population of 3-D points in space (the ‘flies’), using an evolutionary optimisation strategy. Here, in its version dedicated to tomographic reconstruction in medical imaging, the flies are mimicking radioactive photon sources. Evolution is controlled using a fitness function based on the discrepancy of the projections simulated by the flies with the actual pattern received by the sensors. The reconstructed radioactive concentration is derived from the population of flies, i.e. a collection of points in the 3-D Euclidean space, after convergence. ‘Good’ flies were previously binned into voxels. In this paper, we study which flies to include in the final solution and how this information can be sampled to provide more accurate datasets in a reduced computation time. We investigate the use of density fields, based on Metaballs and on Gaussian functions respectively, to obtain a realistic output. The spread of each Gaussian kernel is modulated in function of the corresponding fly fitness. The resulting volumes are compared with previous work in terms of normalised-cross correlation. In our test-cases, data fidelity increases by more than 10% when density fields are used instead of binning. Our method also provides reconstructions comparable to those obtained using well-established techniques used in medicine (filtered back-projection and ordered subset expectation-maximisation).

A combined scheme of pixel and block level splitting for medical image compression and reconstruction

Medical images are very important source of information of human body diagnosis or therapy. These images contain huge amount of information which needs to be compressed to reduce the storage capacity issues. These images require lossless compression techniques to save the information. Another crucial issue in medical imaging is to reconstruct image efficiently without losing information. In order to achieve the reconstruction of medical image we present a new approach by introducing convex-smoothing problems. This approach is implemented by dividing the input image into sub-problems. According to this approach we divide the input image into sub-problems which are solved by using compressed sensing method. After achieving the output of compressed sensed image, this problem is passed to the iterative process of problem solver to reduce the time, computational complexity and reconstruction error. Proposed model is implemented using MATLAB tool and tested with medical images. Results are carried in terms of the PSNR, reconstruction error and time.

Modeling and analysis of visual digital impact model for a Chinese human thorax.

To establish a three-dimensional finite element model of the human chest for engineering research on individual protection. Computed tomography (CT) scanning data were used for three-dimensional reconstruction with the medical image reconstruction software Mimics. The finite element method (FEM) preprocessing software ANSYS ICEM CFD was used for cell mesh generation, and the relevant material behavior parameters of all of the model's parts were specified. The finite element model was constructed with the FEM software, and the model availability was verified based on previous cadaver experimental data. A finite element model approximating the anatomical structure of the human chest was established, and the model's simulation results conformed to the results of the cadaver experiment overall. Segment data of the human body and specialized software can be utilized for FEM model reconstruction to satisfy the need for numerical analysis of shocks to the human chest in engineering research on body mechanics.

Analysis of stability of tomographic reconstruction of X-ray medical images

Slice reconstruction in X-ray computed tomography is reduced to the solution of integral equations, or a system of algebraic equations in discrete case. It is considered to be an ill-posed problem due to the inconsistencies in the number of equations and variables and due to errors in the experimental data. Therefore, determination of the best method of the slice reconstruction is of great interest. Furthermore, all available methods give approximate results. The aim of this article was two-fold: i) to compare two methods of image reconstruction, viz. inverse projection and variation, using the numerical experiment; ii) to obtain the relationship between image accuracy and experimental error. It appeared that the image obtained by inverse projection is unstable: there was no convergence of the approximate image to the accurate one, when the experimental error reached zero. In turn, the image obtained by variational method was accurate at zero experimental error. Finally, the latter showed better slice reconstruction, despite the low number of projections and the experimental errors.