Magnetic Resonance Imaging Algorithms Research Articles

Alzheimer image registration using hybrid random forest and deep regression network algorithm

<span>Image registration involves superimposing images (two or more) of similar background obtained at various periods of time, at different angles, and/or with various detectors. Geometrical alignment of two scans, reference image as well as capture image. The current dissimilarity between images is because of distinct image conditions. Image registration is difficult step in image analysis works on change detection, image fusion as well as <br /> multi-channel images recovery to obtain concluded data from integration of different sources. In this analysis image registration using hybrid random forest (RF) and deep regression network algorithm for magnetic resonance imaging (MRI) applications is implemented. The Alzheimer’s disease neuroimaging initiative (ADNI) database provided by the dataset utilised in this implementation. From results it can observe that compared with individual random of forest, Hybrid RF and deep regression network algorithm improves the accuracy, precision and F1-score in effective way.</span>

K-strip: A novel segmentation algorithm in k-space for the application of skull stripping

Background and objectiveWe present a novel deep learning-based skull stripping algorithm for magnetic resonance imaging (MRI) that works directly in the information rich complex valued k-space. MethodsUsing four datasets from different institutions with a total of around 200,000 MRI slices, we show that our network can perform skull-stripping on the raw data of MRIs while preserving the phase information which no other skull stripping algorithm is able to work with. For two of the datasets, skull stripping performed by HD-BET (Brain Extraction Tool) in the image domain is used as the ground truth, whereas the third and fourth dataset comes with per-hand annotated brain segmentations. ResultsAll four datasets were very similar to the ground truth (DICE scores of 92 %–99 % and Hausdorff distances of under 5.5 pixel). Results on slices above the eye-region reach DICE scores of up to 99 %, whereas the accuracy drops in regions around the eyes and below, with partially blurred output. The output of k-Strip often has smoothed edges at the demarcation to the skull. Binary masks are created with an appropriate threshold. ConclusionWith this proof-of-concept study, we were able to show the feasibility of working in the k-space frequency domain, preserving phase information, with consistent results. Besides preserving valuable information for further diagnostics, this approach makes an immediate anonymization of patient data possible, already before being transformed into the image domain. Future research should be dedicated to discovering additional ways the k-space can be used for innovative image analysis and further workflows.

Magnetic resonance imaging relevance in diagnosis and prognosis of early postoperative period following pancreatic cancer surgical treatment

Purpose of the study. Improvement of the prognostic outcomes for the pancreatic fistula development in surgical treatment for pancreatic cancer by implementation of new diagnostic algorithms for magnetic resonance imaging (MRI) assessment of supposed pancreatic stump.Materials and methods. We performed a retrospective analysis of MRI results of 1136 patients from the medical data base of National Medical Research Centre for Oncology for 2009–2020. An original scanning technique, trans- and cross-pancreatic imaging and MR spectroscopy of the pancreas in patients with pancreatic cancer, was developed and applied. Preoperative examinations were compared retrospectively: a standard MRI protocol without anatomical orientation of the series; MRI protocol using the developed technology.Results. 717 patients were diagnosed with advanced pancreatic cancer. Lymph nodes were affected in 302 patients among 419 patients with radical surgeries. All cases were confirmed morphologically. In the group of 419 patients with pancreatic resections, based on the analysis of morphological and clinical data and preoperative MRI data, we developed an original preoperative scale for assessing the risk of pancreatic fistula development and compared its accuracy with other intraoperative scales, FRS and modified FRS. Lactate and lipid complex were selected from a wide range of metabolites. The surgical protocols and results of histological examination of the surgical material were used to prove the accuracy of the study. The overall accuracy of the technique in predicting fistula development was 97.5 %. New visual predictors («domino» and «white on white» symptoms) based on MRI data were used to improve the scale accuracy.Conclusion. The developed method of using a modified scale for risk assessment of pancreatic fistula development allows predicting the onset of early postoperative complications already at the preoperative stage. When comparing the calculated risks of pancreatic fistula according to the developed scale with the results by the known scales (FRS and modified FRS), statistical analysis showed a significant difference for the better when compared with FRS (p = 0.0477), and a tendency when compared with modified FRS (p = 0.0544).

Diagnostic evaluation of low‐rank matrix denoising algorithm‐based magnetic resonance imaging on tibial plateau fracture complicated with meniscus injury

AbstractIt was to analyse the denoising ability of low‐rank matrix‐based denoising algorithms for magnetic resonance imaging (MRI) images, and explore the diagnostic value of MRI for tibial plateau fracture (TPF) with meniscus injury. The Gaussian mixture algorithm was introduced based on low‐rank matrix algorithm, and a new algorithm Gaussian mixture model‐based low‐rank matrix factorization (GLRMF) was established. Then, the root‐mean‐squared error (RMSE) of the GLRMF algorithm under different noise intensities was calculated, and the structural similarity (SSIM) of the GLRMF algorithm, anisotropic diffusion filter (ADF), and nonlocal means (NLM) algorithm was compared. Sixty cases of TPF patients diagnosed in the Department of Orthopaedics and Traumatology of X Hospital from March 2018 to October 2019 were recruited as the research objects. The results of Schatzker classification of patients were counted, and the results of intraoperative exploration or arthroscopy were used as the standard to analyse the accuracy of MRI in diagnosing TPFs with meniscus injuries. It was found that with the increase of noise intensity, the RMSE of the GLRMF algorithm showed an increasing trend. When the K value was 100, the denoising effect was relatively good. Under the same noise intensity, the SSIM of the GLRMF algorithm was higher than the other two algorithms. Moreover, type VI patients accounted for 55% at most. MRI diagnosed 56 cases (93.33%) of TPF patients with meniscus injury, and 52 cases (86.67%) of meniscus injury were found in intraoperative exploration. The sensitivity of MRI in diagnosing meniscus injury was 94.76%, the accuracy was 93.33%, and the specificity was 89%. It showed that the new GLRMF algorithm based on the low‐rank matrix algorithm had a good noise reduction effect and significantly improved the image quality. The accuracy of knee MRI scanning was significantly improved, it was safe, and it was worthy of clinical application.

Automatic segmentation algorithm for magnetic resonance imaging in prediction of breast tumour histological grading

AbstractTo analyse the application value of dynamic contrast enhanced magnetic resonance imaging (DCE‐MRI) based on computer semi‐automatic segmentation (CSS) algorithm in tumour histological grading of breast cancer patients, the CSS algorithm of DCE‐MRI breast image was established based on Canny edge detection operator and dynamic binarization (DB) algorithm to compare with the fuzzy‐c‐means (FCM) algorithm based on FCM clustering and wavelet transform (WT) algorithm. Besides, CSS was applied to DCE‐MRI image diagnosis in 121 breast cancer patients, who were then classified as grade I, II, and III according to the 2019 edition of the World Health Organization Histological Grading Criteria for Breast Tumors. The results showed that the false positive rate (FPR) of CSS was sharply lower than that of FCM and WT, while the true positive rate (TPR) of CSS increased greatly in contrast to FCM and WT, indicating that there were statistically significant (p < 0.05). The area under the curve (AUC) predicted by FCM, WT, and CSS for histological classification were 0.818, 0.801, and 0.924, respectively. The fractional anisotropy (FA) value of patients with grade III was less than that of patients with grade I and II, while the apparent diffusion coefficient (ADC) value of patients with grade III was greater than that of patients with grade I and II (p < 0.05). The tumour diameter, number of lesions, and number of lymph node metastasis of patients with grade III increased markedly compared to patients with grade I and II (p < 0.05). To sum up, DCE‐MRI image segmentation results based on CSS were superior to FCM and WT in the evaluation of breast tumour histological grading. What's more, there were substantial differences in the obtained tumour diameter, FA value, ADC value, and number of lymph node metastasis based on DCE‐MRI of patients with grade I, II, and III, which could be applied to predict the histological grading of patients with breast cancer.

Effect of MRI-based semiautomatic size-assessment in cerebral metastases on the RANO-BM classification.

Evaluation of asemiautomatic software algorithm for magnetic resonance imaging (MRI)-based assessment of cerebral metastases in cancer patients. Brain metastases (n = 131) in 38patients, assessed by contrast-enhanced MRI, were retrospectively evaluated at two timepoints (baseline, follow-up) by two experienced neuroradiologists in ablinded manner. The response assessment in neuro-oncology (RANO) criteria for brain metastases (RANO-BM) were applied by means of asoftware (autoRANO-BM) as well as manually (manRANO-BM) at an interval of 3weeks. The average diameter of metastases was 12.03 mm (SD ± 6.66 mm) for manRANO-BM and 13.97 mm (SD ± 7.76 mm) for autoRANO-BM. Diameter figures were higher when using semiautomatic measurements (median = 11.8 mm) as compared to the manual ones (median = 10.2 mm; p = 0.000). Correlation coefficients for intra-observer variability were 0.993 (autoRANO-BM) and 0.979 (manRANO-BM). The interobserver variability (R1/R2) was 0.936/0.965 for manRANO-BM and 0.989/0.998 for autoRANO-BM. A total of 19 lesions (15%) were classified differently when using semiautomatic measurements. In 14cases with suspected disease progression by manRANO-BM astable course was found according to autoRANO-BM. Computerized measuring techniques can aid in the assessment of cerebral metastases by reducing examiner-dependent effects and may consequently result in adifferent classification according to RANO-BM criteria.

MRI reconstruction based on three regularizations: Total variation and two wavelets

Abstract This paper proposes an efficient algorithm for magnetic resonance imaging (MRI) reconstruction based on compressed sensing (CS) to further improve the accuracy of the reconstructed image and simultaneously take into account the reconstruction speed. The proposed algorithm makes full use of prior knowledge of the sparsity of the MR image in different transform domains. It uses the total variation (TV), the orthogonal Haar wavelet, and the orthogonal Daubechies20 wavelet as regularizations, on top the split Bregman iteration algorithm. Experimental results show that when the sample ratio is 20%, SNR is 30 dB, the average reconstruction PSNR of the proposed algorithm is 41.09 dB. The proposed method is compared to nonlinear conjugate gradients combined with the backtracking line-search method, split Bregman algorithm based on TV regularization method and split Bregman algorithm based on TV + orthogonal Haar wavelet regularizations method, and the average reconstruction peak signal to noise ratio (PSNR) increases 3.71 dB, 3.44 dB and 1.14 dB, respectively.

Task-Space Motion Planning of MRI-Actuated Catheters for Catheter Ablation of Atrial Fibrillation.

This paper presents a motion planning algorithm for Magnetic Resonance Imaging (MRI) actuated catheters for catheter ablation of atrial fibrillation. The MRI-actuated catheters is a new robotic catheter concept which utilizes MRI for remote steering and guidance. Magnetic moments generated by a set of coils wound near the tip are used to steer the catheter under MRI scanner magnetic field. The catheter during an ablation procedure is modeled as a constrained robotic manipulator with flexible joints, and the proposed motion-planning algorithm calculates a sequence of magnetic moments based on the manipulator model to move the tip of the catheter along a predefined trajectory on the surface of the left atrium. The difficulties in motion planning of the catheter are due to kinematic redundancy and underactuation. The proposed motion planning algorithm overcomes the challenges by operating in the task space instead of the configuration space. The catheter is then regulated around this nominal trajectory using feedback control to reduce the effect of uncertainties.

Tumor Extraction by Level Set Method using Thershold Algorithm

The level set method may be used as a strongest pawn for segmentation of a tumor to achieve an accurate estimation of its volume. The level set method is a numeric technique for tracking interfaces and shapes. In this method, equation parameters are being set or speed function is being set. A threshold based method is introduced for tumor segmentation. In this paper, tumor segmentation and its extraction is achieved by a threshold based scheme and by utilizing a global threshold, the level set speed function is designed. This threshold based scheme provides better flexibility and it is updated through the whole process. Search based and adaptive bases threshold can be used here for better efficiency through segmentation. Tumor segmentation does not need any vast knowledge about the tumor and non-tumor density function. Depending upon the tumor shape and size, it may be implemented in an automatic or semi-automatic form. Here we use this algorithm for magnetic resonance images (MRI). We see that the performance can be evaluated accurately for quantitatively images. The results from this experiment provide better efficiency and high performance.

The clique potential of Markov random field in a random experiment for estimation of noise levels in 2D brain MRI

ABSTRACTEffective performance of many image processing and image analysis algorithms is strongly dependent on accurate estimation of noise level. We exploit the simplicity and similarity of statistics of human anatomy among different subjects to develop new noise level estimation algorithm for magnetic resonance images of brain. Objects of the experiment are noise‐free 3D brain MRI of 422 subjects. There are 21 slices for each subject. For each slice, total clique potential (TCP) of Markov random field, computed from local clique potential, is indexed by 200 different levels of noise. The sample space is the set of TCP‐noise level data of each slice. The random variable is the set of indices of noise level of TCP in each element of sample space that is closest in numerical value to TCP measured from a test MRI slice. Noise level is estimated from the mean and variance of the random variable. We also report the formulation of a generalized mathematical model describing relationship between TCP and Rician noise level in brain MRI images. Our proposal can operate in the absence of signals in the background and significantly reduce modeling errors inherent in strong parametric assumptions adopted by some of the current algorithms. © 2013 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 23, 304–413, 2013

Accuracy of device-specific 2D and 3D image distortion correction algorithms for magnetic resonance imaging of the head provided by a manufacturer

For the application of magnetic resonance imaging (MRI) in precision radiotherapy, image distortions must be reduced to a minimum to maintain geometrical accuracy. Recently, two-dimensional (2D) and three-dimensional (3D) algorithms for MRI-device-specific distortion corrections were developed by the manufacturers of MRI devices. A previously developed phantom (Karger C P et al 2003 Phys. Med. Biol. 48 211–21) was used to quantify and assess the size of geometrical image distortions before and after application of the 2D and 3D correction algorithm in the head region. Four different types of MRI devices with different gradient systems were measured. For comparison, measurements were also performed with two computed tomography (CT) devices. Mean distortions of up to 4.6±1.4 mm (maximum: 5.8 mm) were found prior to the correction. After the correction, the mean distortions were well below 2.0 mm in most cases. Distortions in the CT images were below or equal to 1.0 mm on average. Generally, the 3D algorithm produced comparable or better results than the 2D algorithm. The remaining distortions after the correction appear to be acceptable for fractionated radiotherapy.

Post-processing noise removal algorithm for magnetic resonance imaging based on edge detection and wavelet analysis

A post-processing noise suppression technique for biomedical MRI images is presented. The described procedure recovers both sharp edges and smooth surfaces from a given noisy MRI image; it does not blur the edges and does not introduce spikes or other artefacts. The fine details of the image are also preserved. The proposed algorithm first extracts the edges from the original image and then performs noise reduction by using a wavelet de-noise method. After the application of the wavelet method, the edges are restored to the filtered image. The result is the original image with less noise, fine detail and sharp edges. Edge extraction is performed by using an algorithm based on Sobel operators. The wavelet de-noise method is based on the calculation of the correlation factor between wavelet coefficients belonging to different scales. The algorithm was tested on several MRI images and, as an example of its application, we report the results obtained from a spin echo (multi echo) MRI image of a human wrist collected with a low field experimental scanner (the signal-to-noise ratio, SNR, of the experimental image was 12). Other filtering operations have been performed after the addition of white noise on both channels of the experimental image, before the magnitude calculation. The results at SNR = 7, SNR = 5 and SNR = 3 are also reported. For SNR values between 5 and 12, the improvement in SNR was substantial and the fine details were preserved, the edges were not blurred and no spikes or other artefacts were evident, demonstrating the good performances of our method. At very low SNR (SNR = 3) our result is worse than that obtained by a simpler filtering procedure.

An iterative method for maximum entropy regularization reconstruction in MRI

Many problems in physics involve imaging objects with high spatial frequency content in a limited amount of time. The limitation of available experimental data leads to the infamous problem of diffraction limited data, which manifests itself by causing ringing in the image. This ringing is due to interference phenomena in optics and is known as the Gibbs phenomenon in engineering. In this paper, an iterative maximum entropy regularization (IMER) algorithm for magnetic resonance imaging (MRI) is developed, which produces a super-resolution and optimal signal-to-noise solution to the problem of reconstructing a source from partial Fourier transform data. This method is capable, in principle, of unlimited resolution and is robust with respect to Gaussian white noise perturbation. Comparisons of the IMER method with the conventional Fourier transform method are carried out with the real magnetic resonance data to illustrate the performance of the proposed method. © 2000 John Wiley & Sons, Inc. Int J Imaging Syst Technol, 10, 427–431, 1999

Radiologist evaluation of a multispectral image compression algorithm for magnetic resonance images.

With the advent of teleradiology and picture archiving and communication systems (PACS), the expense and time required for image transmission and long term image archiving become important. The use of validated image compression algorithms can greatly reduce these costs. A lossy, multispectral image compression scheme at compression ratios (CR) of 25:1 and 32:1 was used for a set of 26 different patient MR exams. The original and compressed/decompressed (CD) images sets were evaluated in a blinded fashion by four radiologists in two phases. The main objective was to determine whether radiologic interpretation would vary between the two types of CD image sets and the corresponding originals. In general, the compression algorithm caused a slight decrease in image quality; however, the interpretation of pathology did not change between the original and CD image sets. In only one case at the maximum CR = 32 did one of four radiologists change the interpretation of pathology after CD. In this study, lossy multispectral image compression of MR images at CR = 25 maintained diagnostic integrity. This could play a significant role in image storage and communications.

Reconstruction by weighted correlation for MRI with time-varying gradients

A general reconstruction algorithm for magnetic resonance imaging (MRI) with gradients having arbitrary time dependence is presented. This method estimates spin density by calculating the weighted correlation of the observed free induction decay signal and the phase modulation function at each point. A theorem which states that this method can be derived from the conditions of linearity and shift invariance is presented. Since these conditions are general, most of the MRI reconstruction algorithms proposed so far are equivalent to the weighted correlation method. An explicit representation of the point spread function (PSF) in the weighted correlation method is given. By using this representation, a method to control the PSF and the static field inhomogeneity effects is studied. A correction method for the inhomogeneity is proposed, and a limitation is clarified. Some simulation results are presented.