Cardiac right ventricle (RV) segmentation plays an essential role in the functional analysis of heart diseases, such as pulmonary hypertension. The myocardium of RV is thin and irregular-shaped, making the traditional segmentation methods less effective. To improve RV segmentation, a COLLATE (Consensus Level, Labeler Accuracy and Truth Estimation) fusion-based multi-atlas method was developed. The preprocessed target image was first registered to atlas images with a B-spline algorithm optimizing normalized mutual information. The registration coefficients obtained were then used to get a rough RV segmentation for COLLATE fusion. Shape-constrained region growing algorithm was used to correct the segmentation errors. Ten cardiac magnetic resonance datasets were blindly selected to compare the performance of RV segmentation between the method developed and a method based on deep learning. The results of manual segmentation were used as the golden standard. Ejection fraction (EF) calculated with the proposed segmentation method showed better correlation and consistency with the golden standard, relative to the results calculated with the deep learning method.

Proton density fat fraction (PDFF) is emerging as a useful quantitative magnetic resonance imaging (MRI) biomarker for the diagnosis of hepatic steatosis. In this work, a hybrid method combining complex-based and magnitude-based water-fat separation was proposed to provide accurate measurements of liver PDFF. Clinical experiments were performed to evaluate the feasibility and accuracy of the method proposed. The results demonstrated that PDFF results obtained with the proposed method had excellent correlation and agreement with those calculated from magnetic resonance spectroscopy (MRS).

Magnetic resonance spectroscopy (MRS) is useful for rapid and robust liver fat quantification. MR spectra of liver can be acquired in a single breath-hold with the high-speed multi-echo stimulated echo acquisition mode, from which T₂-corrected liver proton density fat fraction (PDFF) can be derived. However, the widespread application of this method is hampered by the cumbersome post-processing procedures for the spectroscopy data. In this work, an automated algorithm for liver fat quantification from multi-echo spectroscopy data was developed. The feasibility and accuracy of the algorithm were evaluated with fat-water phantoms experiments and human liver experiments. The results demonstrated excellent correlation and agreement between the PDFF measurements with the proposed algorithm and those obtained with the jMRUI software.

Real-time monitoring of tissue temperature is required during high intensity focused ultrasound (HIFU) tumor treatment to ensure safety and effectiveness. Magnetic resonance imaging (MRI) can be used to measure tissue temperature non-invasively during HIFU treatment. This paper examined the effects of coagulation necrosis-induced tissue phase transition on magnetic resonance thermometry (MRT) during HIFU tumor treatment. With a two-state rapid exchange model, the relationship between tissue longitudinal relaxation time (T₁) and temperature before and after HIFU radiation-induced coagulation necrosis/tissue phase transition were analyzed theoretically. Taking the effects of tissue phase transition into account, the experimental scheme and data processing procedures for MRT were optimized, and better temperature measurements were obtained. The work demonstrated the importance of considering the effects of tissue phase transition in real-time MRT during HIFU treatment.

Training data enrichment is a key factor in artificial intelligence (AI) technology development. At present, the bottleneck problem is that the quantity and type of labeled training data in valid samples are unable to meet the requirements of AI+MRI aided diagnosis. In this paper, an effective approach to solve the problem was presented. High resolution isotropic multi-dimensional data of regions of interests from patients or healthy volunteers were first acquired via a series of scanning on clinical MRI scanners, including quantitative T₁, T₂, proton density (Pd) and apparent diffusion coefficient (ADC) measurements. These data were then used as the ground truth, from which different types of images associated with different imaging sequences and parameters were obtained with a virtual MRI technology. The type of the images with the best boundary resolution were then selected manually by experienced doctors, on which three-dimensional mask matrix was obtained by manual contouring and labeling, serving as the template for other types of images. This enrichment method was developed as a software platform, which could provide sufficient quantity of image data from a small number of positive cases, thus meeting the data training enrichment requirement of AI+MRI diagnosis at low cost and with high efficiency.

Image registration is often used to transform images from different subjects into the same spatial space to ensure more accurate comparisons. With the traditional image registration methods, one image is usually specified as the reference image, while the other images are transformed into the space of the reference image. However, the image randomly selected as the reference image may have significant deviations from the group mean in datasets with large individual differences, causing registration biases and affecting the final results of groupwise comparisons. In this paper, a groupwise registration method based on the topology center of the image set was proposed. With the same open datasets, the performance of the proposed method was compared to that of two widely used traditional methods. The experimental results demonstrated that the proposed method had smaller groupwise registration bias and better registration results. The paper also proposed a simple measure for evaluating groupwise registration bias.

In magnetic resonance imaging (MRI), a gradient preemphasis module with more compensation channels and adjustable parameters is often desired for better imaging quality and faster imaging speed. With the conventional design, however, a high performance preemphasis module often requires high resource consumption such that the field programmable gate array (FPGA) may encounter huge resource overhead. In this study, a new gradient preemphasis implementation scheme based on time-division multiplexing was proposed. With the new design, an 11 channels×4 sets of parameters implementation could be achieved with only 1/44 of the resources required by the conventional scheme. To test the performance of the new implementation, eddy current curves and magnetic resonance images before and after preemphasis compensation were measured on a 0.35 T scanner and compared. The results demonstrated that the module could effectively reduce eddy currents and related artifacts on the images.

This paper presents the design of a magnetic resonance signal receiving system based on all programmable SoC (System-on-a-Chip) and LabVIEW (Laboratory Virtual Instrument Engineering Workbench). Using the all programmable SoC that integrates ARM (Advanced RISC Machines) and FPGA (Field Programmable Gate Array) as the main chips of the receiver, the DDC (Digital Down Converter) algorithm was designed using System Generator, a DSP (Digital Signal Processing) development tool provided by Xilinx, and the receiver hardware circuit designed by the authors. The visual programming platform LabVIEW was used to design the magnetic resonance upper computer software. Display, storage and communication of the magnetic resonance signals were done with the receiver. Experimental results showed that the receiver designed could receive magnetic resonance echo signals correctly and provided a high signal-to-noise ratio. Using LabVIEW to design magnetic resonance software on personal computers significantly improved software development efficiency.

Compressed sensing is widely used in accelerated magnetic resonance imaging (MRI) to reduce scan time. With compressed sensing, high-quality MR images could be acquired and reconstructed with only a small amount of K space data. The compressed sensing algorithm models image reconstruction as a linear combination minimization problem that includes data fidelity terms, sparse priors, and total variation terms. Sparse representation is a key assumption of the compressed sensing theory, and the quality of reconstruction largely depends on sparse transformation. In this article, we proposed a compressed sensing low-field MRI reconstruction algorithm that combined dual-tree wavelet transform and wavelet tree sparsity. Experimental results demonstrated that the proposed algorithm had certain advantages over the conventional reconstruction algorithm, in terms of certain objective evaluation indicators.

Methods to map radiofrequency (RF) field generated by an RF coil/pulse have been developed. By measuring the changes of RF field in biological tissues when an RF pulse passing through, the electrical properties of the tissue could be retrieved and used for early diagnosis of diseases such as cancer. So far, RF field mapping is mainly done for birdcage coils at high fields. Less research has been performed on phased array coils at low fields. This work studied how loading would affect the distribution of RF field in low-field MRI with permanent magnet (i.e., 17.8 MHz). Both finite element simulation and experimental results demonstrated that coil loading changed the uniformity of RF field significantly, and the distortion of RF field under the loaded condition could reflect the electrical characteristics of the loaded biological tissues.

Chemical exchange saturation transfer (CEST) is a novel molecular magnetic resonance imaging (MRI) method that makes full use of the chemical exchange between water protons and exchangeable protons from solute molecules. Through chemical exchanges, the MRI signal of bulk water decreases when saturating at specific frequency of solute protons. Important biological information related to the chemical exchange processes could be extracted from the amplitude of water signal decrease. For example, the CEST signals have been used for in vivo pH imaging since the proton exchange rate is often pH-dependent. CEST signals originated from endogenous proteins/peptides and exogenous small molecule/metal chelate probes have been used for in vivo pH imaging. Using the ratiometric methods or the amine and amide concentration-independent detection (AACID) methods, in vivo pH maps have been acquired from kidneys, ischemic brain and tumors. In this paper, we thoroughly reviewed the progresses during the past twenty years in the field of in vivo pH imaging with CEST contrast. Principles, methods and applications were discussed, as well as development trends and future directions.

Diffusion-weighted magnetic resonance images contain rich information on brain white matter (WM), and have been used to construct brain templates/atlases. However, the accuracy of brain templates reconstructed with the diffusion tensor model is often limited in regions with complex WM configurations. To overcome this problem, high angular resolution diffusion imaging acquisition and reconstruction have been developed for building higher quality brain template. In this paper, the research progresses on the construction of brain template using diffusion tensor imaging (DTI) are reviewed. Firstly, the technical aspects and limitations in using DTI data to build brain templates were discussed. Secondly, diffusion spectrum imaging and high angular resolution diffusion imaging techniques were introduced, with their advantages over the DTI techniques explained. Finally, the existing problems and the perspectives of the field were discussed.