Extracting diffusion tensor fractional anisotropy and mean diffusivity from 3-direction DWI scans using deep learning

ssVERDICT: Self-supervised VERDICT-MRI for enhanced prostate tumor characterization

PURPOSE

Demonstrating and assessing self-supervised machine-learning fitting of the VERDICT (vascular, extracellular and restricted diffusion for cytometry in tumors) model for prostate cancer.

METHODS

We derive a self-supervised neural network for fitting VERDICT (ssVERDICT) that estimates parameter maps without training data. We compare the performance of ssVERDICT to two established baseline methods for fitting diffusion MRI models: conventional nonlinear least squares and supervised deep learning. We do this quantitatively on simulated data by comparing the Pearson's correlation coefficient, mean-squared error, bias, and variance with respect to the simulated ground truth. We also calculate in vivo parameter maps on a cohort of 20 prostate cancer patients and compare the methods' performance in discriminating benign from cancerous tissue via Wilcoxon's signed-rank test.

RESULTS

In simulations, ssVERDICT outperforms the baseline methods (nonlinear least squares and supervised deep learning) in estimating all the parameters from the VERDICT prostate model in terms of Pearson's correlation coefficient, bias, and mean-squared error. In vivo, ssVERDICT shows stronger lesion conspicuity across all parameter maps, and improves discrimination between benign and cancerous tissue over the baseline methods.

CONCLUSION

ssVERDICT significantly outperforms state-of-the-art methods for VERDICT model fitting and shows, for the first time, fitting of a detailed multicompartment biophysical diffusion MRI model with machine learning without the requirement of explicit training labels.

Finding the limits of deep learning clinical sensitivity with fractional anisotropy (FA) microstructure maps

Background

Quantitative maps obtained with diffusion weighted (DW) imaging, such as fractional anisotropy (FA) -calculated by fitting the diffusion tensor (DT) model to the data,-are very useful to study neurological diseases. To fit this map accurately, acquisition times of the order of several minutes are needed because many noncollinear DW volumes must be acquired to reduce directional biases. Deep learning (DL) can be used to reduce acquisition times by reducing the number of DW volumes. We already developed a DL network named "one-minute FA," which uses 10 DW volumes to obtain FA maps, maintaining the same characteristics and clinical sensitivity of the FA maps calculated with the standard method using more volumes. Recent publications have indicated that it is possible to train DL networks and obtain FA maps even with 4 DW input volumes, far less than the minimum number of directions for the mathematical estimation of the DT.

Methods

Here we investigated the impact of reducing the number of DW input volumes to 4 or 7, and evaluated the performance and clinical sensitivity of the corresponding DL networks trained to calculate FA, while comparing results also with those using our one-minute FA. Each network training was performed on the human connectome project open-access dataset that has a high resolution and many DW volumes, used to fit a ground truth FA. To evaluate the generalizability of each network, they were tested on two external clinical datasets, not seen during training, and acquired on different scanners with different protocols, as previously done.

Results

Using 4 or 7 DW volumes, it was possible to train DL networks to obtain FA maps with the same range of values as ground truth - map, only when using HCP test data; pathological sensitivity was lost when tested using the external clinical datasets: indeed in both cases, no consistent differences were found between patient groups. On the contrary, our "one-minute FA" did not suffer from the same problem.

Conclusion

When developing DL networks for reduced acquisition times, the ability to generalize and to generate quantitative biomarkers that provide clinical sensitivity must be addressed.

Deep learning-based diffusion tensor cardiac magnetic resonance reconstruction: a comparison study

In vivo cardiac diffusion tensor imaging (cDTI) is a promising Magnetic Resonance Imaging (MRI) technique for evaluating the microstructure of myocardial tissue in living hearts, providing insights into cardiac function and enabling the development of innovative therapeutic strategies. However, the integration of cDTI into routine clinical practice poses challenging due to the technical obstacles involved in the acquisition, such as low signal-to-noise ratio and prolonged scanning times. In this study, we investigated and implemented three different types of deep learning-based MRI reconstruction models for cDTI reconstruction. We evaluated the performance of these models based on the reconstruction quality assessment, the diffusion tensor parameter assessment as well as the computational cost assessment. Our results indicate that the models discussed in this study can be applied for clinical use at an acceleration factor (AF) of × 2 and × 4 , with the D5C5 model showing superior fidelity for reconstruction and the SwinMR model providing higher perceptual scores. There is no statistical difference from the reference for all diffusion tensor parameters at AF × 2 or most DT parameters at AF × 4 , and the quality of most diffusion tensor parameter maps is visually acceptable. SwinMR is recommended as the optimal approach for reconstruction at AF × 2 and AF × 4 . However, we believe that the models discussed in this study are not yet ready for clinical use at a higher AF. At AF × 8 , the performance of all models discussed remains limited, with only half of the diffusion tensor parameters being recovered to a level with no statistical difference from the reference. Some diffusion tensor parameter maps even provide wrong and misleading information.

Ultrafast diffusion tensor imaging based on deep learning and multi-slice information sharing

Objective. Diffusion tensor imaging (DTI) is excellent for non-invasively quantifying tissue microstructure. Theoretically DTI can be achieved with six different diffusion weighted images and one reference image, but the tensor estimation accuracy is poor in this case. Increasing the number of diffusion directions has benefits for the tensor estimation accuracy, which results in long scan time and makes DTI sensitive to motion. It would be beneficial to decrease the scan time of DTI by using fewer diffusion-weighted images without compromising reconstruction quality.Approach. A novel DTI scan scheme was proposed to achieve fast DTI, where only three diffusion directions per slice was required under a specific direction switching manner, and a deep-learning based reconstruction method was utilized using multi-slice information sharing and correspondingT1-weighted image for high-quality DTI reconstruction. A network with two encoders developed from U-Net was implemented for better utilizing the diffusion data redundancy between neighboring slices. The method performed direct nonlinear mapping from diffusion-weighted images to diffusion tensor.Main results. The performance of the proposed method was verified on the Human Connectome Project public data and clinical patient data. High-quality mean diffusivity, fractional anisotropy, and directionally encoded colormap can be achieved with only three diffusion directions per slice.Significance. High-quality DTI-derived maps can be achieved in less than one minute of scan time. The great reduction of scan time will help push the wider application of DTI in clinical practice.

Evaluation of brain nerve function in ICU patients with Delirium by deep learning algorithm-based resting state MRI

The purpose of this study was to explore the value of resting-state magnetic resonance imaging (MRI) based on the brain extraction tool (BET) algorithm in evaluating the cranial nerve function of patients with delirium in intensive care unit (ICU). A total of 100 patients with delirium in hospital were studied, and 20 healthy volunteers were used as control. All the subjects were examined by MRI, and the images were analyzed by the BET algorithm, and the convolution neural network (CNN) algorithm was introduced for comparison. The application effects of the two algorithms were analyzed, and the differences of brain nerve function between delirium patients and normal people were explored. The results showed that the root mean square error, high frequency error norm, and structural similarity of the BET algorithm were 70.4%, 71.5%, and 0.92, respectively, which were significantly higher than those of the CNN algorithm (P < 0.05). Compared with normal people, the ReHo values of pontine, hippocampus (right), cerebellum (left), midbrain, and basal ganglia in delirium patients were significantly higher. ReHo values of frontal gyrus, middle frontal gyrus, left inferior frontal gyrus, parietal lobe, and temporal lobe and anisotropy scores (FA) of cerebellums (left), frontal lobe, temporal lobe (left), corpus callosum, and hippocampus (left) decreased significantly. The average diffusivity (MD) of medial frontal lobe, superior temporal gyrus (right), the first half of cingulate gyrus, bilateral insula, and caudate nucleus (left) increased significantly (P < 0.05). MRI based on the deep learning algorithm can effectively improve the image quality, which is valuable in evaluating the brain nerve function of delirium patients. Abnormal brain structure damage and abnormal function can be used to help diagnose delirium.

dtiRIM: A generalisable deep learning method for diffusion tensor imaging

Diffusion weighted MRI is an indispensable tool for routine patient screening and diagnostics of pathology. Recently, several deep learning methods have been proposed to quantify diffusion parameters, but poor generalisation to new data prevents broader use of these methods, as they require retraining of the neural network for each new scan protocol. In this work, we present the dtiRIM, a new deep learning method for Diffusion Tensor Imaging (DTI) based on the Recurrent Inference Machines. Thanks to its ability to learn how to solve inverse problems and to use the diffusion tensor model to promote data consistency, the dtiRIM can generalise to variations in the acquisition settings. This enables a single trained network to produce high quality tensor estimates for a variety of cases. We performed extensive validation of our method using simulation and in vivo data, and compared it to the Iterated Weighted Linear Least Squares (IWLLS), the approach of the state-of-the-art MRTrix3 software, and to an implementation of the Maximum Likelihood Estimator (MLE). Our results show that dtiRIM predictions present low dependency on tissue properties, anatomy and scanning parameters, with results comparable to or better than both IWLLS and MLE. Further, we demonstrate that a single dtiRIM model can be used for a diversity of data sets without significant loss in quality, representing, to our knowledge, the first generalisable deep learning based solver for DTI.

Undersampled single-shell to MSMT fODF reconstruction using CNN-based ODE solver

BACKGROUND AND OBJECTIVE

Diffusion MRI (dMRI) has been considered one of the most popular non-invasive techniques for studying the human brain's white matter (WM). dMRI is used to delineate the brain's microstructure by approximating the WM region's fiber tracts. The achieved fiber tracts can be utilized to assess mental diseases like Multiple sclerosis, ADHD, Seizures, Intellectual disability, and others. New techniques such as high angular resolution diffusion-weighted imaging (HARDI) have been developed, providing precise fiber directions, and overcoming the limitation of traditional DTI. Unlike Single-shell, Multi-shell HARDI provides tissue fractions for white matter, gray matter, and cerebrospinal fluid, resulting in a Multi-shell Multi-tissue fiber orientation distribution function (MSMT fODF). This MSMT fODF comes up with more precise fiber directions than a Single-shell, which helps to get correct fiber tracts. In addition, various multi-compartment diffusion models, including as CHARMED and NODDI, have been developed to describe the brain tissue microstructural information. This type of model requires multi-shell data to obtain more specific tissue microstructural information. However, a major concern with multi-shell is that it takes a longer scanning time restricting its use in clinical applications. In addition, most of the existing dMRI scanners with low gradient strengths commonly acquire a single b-value (shell) upto b=1000s/mm² due to SNR (Signal-to-noise ratio) reasons and severe imaging artifacts.

METHODS

To address this issue, we propose a CNN-based ordinary differential equations solver for the reconstruction of MSMT fODF from under-sampled and fully sampled Single-shell (b=1000s/mm²) dMRI. The proposed architecture consists of CNN-based Adams-Bash-forth and Runge-Kutta modules along with two loss functions, including L₁ and total variation.

RESULTS

We have shown quantitative results and visualization of fODF, fiber tracts, and structural connectivity for several brain regions on the publicly available HCP dataset. In addition, the obtained angular correlation coefficients for white matter and full brain are high, showing the proposed network's utility.Finally, we have also demonstrated the effect of noise by adjusting SNR from 5 to 50 and observed the network robustness.

CONCLUSION

We can conclude that our model can accurately predict MSMT fODF from under-sampled or fully sampled Single-shell dMRI volumes.

Accelerating Cardiac Diffusion Tensor Imaging With a U-Net Based Model: Toward Single Breath-Hold

BACKGROUND

In vivo cardiac diffusion tensor imaging (cDTI) characterizes myocardial microstructure. Despite its potential clinical impact, considerable technical challenges exist due to the inherent low signal-to-noise ratio.

PURPOSE

To reduce scan time toward one breath-hold by reconstructing diffusion tensors for in vivo cDTI with a fitting-free deep learning approach.

STUDY TYPE

Retrospective.

POPULATION

A total of 197 healthy controls, 547 cardiac patients.

FIELD STRENGTH/SEQUENCE

A 3 T, diffusion-weighted stimulated echo acquisition mode single-shot echo-planar imaging sequence.

ASSESSMENT

A U-Net was trained to reconstruct the diffusion tensor elements of the reference results from reduced datasets that could be acquired in 5, 3 or 1 breath-hold(s) (BH) per slice. Fractional anisotropy (FA), mean diffusivity (MD), helix angle (HA), and sheetlet angle (E2A) were calculated and compared to the same measures when using a conventional linear-least-square (LLS) tensor fit with the same reduced datasets. A conventional LLS tensor fit with all available data (12 ± 2.0 [mean ± sd] breath-holds) was used as the reference baseline.

STATISTICAL TESTS

Wilcoxon signed rank/rank sum and Kruskal-Wallis tests. Statistical significance threshold was set at P = 0.05. Intersubject measures are quoted as median [interquartile range].

RESULTS

For global mean or median results, both the LLS and U-Net methods with reduced datasets present a bias for some of the results. For both LLS and U-Net, there is a small but significant difference from the reference results except for LLS: MD 5BH (P = 0.38) and MD 3BH (P = 0.09). When considering direct pixel-wise errors the U-Net model outperformed significantly the LLS tensor fit for reduced datasets that can be acquired in three or just one breath-hold for all parameters.

DATA CONCLUSION

Diffusion tensor prediction with a trained U-Net is a promising approach to minimize the number of breath-holds needed in clinical cDTI studies.

EVIDENCE LEVEL

4 TECHNICAL EFFICACY: Stage 1.

Hybrid Graph Transformer for Tissue Microstructure Estimation with Undersampled Diffusion MRI Data

Advanced contemporary diffusion models for tissue microstructure often require diffusion MRI (DMRI) data with sufficiently dense sampling in the diffusion wavevector space for reliable model fitting, which might not always be feasible in practice. A potential remedy to this problem is by using deep learning techniques to predict high-quality diffusion microstructural indices from sparsely sampled data. However, existing methods are either agnostic to the data geometry in the diffusion wavevector space ( q -space) or limited to leveraging information from only local neighborhoods in the physical coordinate space ( x -space). Here, we propose a hybrid graph transformer (HGT) to explicitly consider the q -space geometric structure with a graph neural network (GNN) and make full use of spatial information with a novel residual dense transformer (RDT). The RDT consists of multiple densely connected transformer layers and a residual connection to facilitate model training. Extensive experiments on the data from the Human Connectome Project (HCP) demonstrate that our method significantly improves the quality of microstructural estimations over existing state-of-the-art methods.

A generalized deep learning network for fractional anisotropy reconstruction: Application to epilepsy and multiple sclerosis

Fractional anisotropy (FA) is a quantitative map sensitive to microstructural properties of tissues in vivo and it is extensively used to study the healthy and pathological brain. This map is classically calculated by model fitting (standard method) and requires many diffusion weighted (DW) images for data quality and unbiased readings, hence needing the acquisition time of several minutes. Here, we adapted the U-net architecture to be generalized and to obtain good quality FA from DW volumes acquired in 1 minute. Our network requires 10 input DW volumes (hence fast acquisition), is robust to the direction of application of the diffusion gradients (hence generalized), and preserves/improves map quality (hence good quality maps). We trained the network on the human connectome project (HCP) data using the standard model-fitting method on the entire set of DW directions to extract FA (ground truth). We addressed the generalization problem, i.e., we trained the network to be applicable, without retraining, to clinical datasets acquired on different scanners with different DW imaging protocols. The network was applied to two different clinical datasets to assess FA quality and sensitivity to pathology in temporal lobe epilepsy and multiple sclerosis, respectively. For HCP data, when compared to the ground truth FA, the FA obtained from 10 DW volumes using the network was significantly better (p <10-4) than the FA obtained using the standard pipeline. For the clinical datasets, the network FA retained the same microstructural characteristics as the FA calculated with all DW volumes using the standard method. At the subject level, the comparison between white matter (WM) ground truth FA values and network FA showed the same distribution; at the group level, statistical differences of WM values detected in the clinical datasets with the ground truth FA were reproduced when using values from the network FA, i.e., the network retained sensitivity to pathology. In conclusion, the proposed network provides a clinically available method to obtain FA from a generic set of 10 DW volumes acquirable in 1 minute, augmenting data quality compared to direct model fitting, reducing the possibility of bias from sub-sampled data, and retaining FA pathological sensitivity, which is very attractive for clinical applications.

Diffusion tensor estimation with transformer neural networks

Diffusion tensor imaging (DTI) is a widely used method for studying brain white matter development and degeneration. However, standard DTI estimation methods depend on a large number of high-quality measurements. This would require long scan times and can be particularly difficult to achieve with certain patient populations such as neonates. Here, we propose a method that can accurately estimate the diffusion tensor from only six diffusion-weighted measurements. Our method achieves this by learning to exploit the relationships between the diffusion signals and tensors in neighboring voxels. Our model is based on transformer networks, which represent the state of the art in modeling the relationship between signals in a sequence. In particular, our model consists of two such networks. The first network estimates the diffusion tensor based on the diffusion signals in a neighborhood of voxels. The second network provides more accurate tensor estimations by learning the relationships between the diffusion signals as well as the tensors estimated by the first network in neighboring voxels. Our experiments with three datasets show that our proposed method achieves highly accurate estimations of the diffusion tensor and is significantly superior to three competing methods. Estimations produced by our method with six diffusion-weighted measurements are comparable with those of standard estimation methods with 30-88 diffusion-weighted measurements. Hence, our method promises shorter scan times and more reliable assessment of brain white matter, particularly in non-cooperative patients such as neonates and infants.

VRfRNet: Volumetric ROI fODF reconstruction network for estimation of multi-tissue constrained spherical deconvolution with only single shell dMRI

Diffusion MRI (dMRI) is one of the most popular techniques for studying the brain structure, mainly the white matter region. Among several sampling methods in dMRI, the high angular resolution diffusion imaging (HARDI) technique has attracted researchers due to its more accurate fiber orientation estimation. However, the current single-shell HARDI makes the intravoxel structure challenging to estimate accurately. While multi-shell acquisition can address this problem, it takes a longer scanning time, restricting its use in clinical applications. In addition, most existing dMRI scanners with low gradient-strengths often acquire single-shell up to b=1000s/mm² because of signal-to-noise ratio issues and severe image artefacts. Hence, we propose a novel generative adversarial network, VRfRNet, for the reconstruction of multi-shell multi-tissue fiber orientation distribution function from single-shell HARDI volumes. Such a transformation learning is performed in the spherical harmonics (SH) space, as raw input HARDI volume is transformed to SH coefficients to soften gradient directions. The proposed VRfRNet consists of several modules, such as multi-context feature enrichment module, feature level attention, and softmax level attention. In addition, three loss functions have been used to optimize network learning, including L₁, adversarial, and total variation. The network is trained and tested using standard qualitative and quantitative performance metrics on the publicly available HCP data-set.

Single-shell to multi-shell dMRI transformation using spatial and volumetric multilevel hierarchical reconstruction framework

Single or Multi-shell high angular resolution diffusion imaging (HARDI) has become an important dMRI acquisition technique for studying brain white matter fibers. Existing single-shell HARDI makes it challenging to estimate the intravoxel structure up to the desired resolution. However, multi-shell acquisition (with multiple b-values) can provide higher resolution for the intravoxel structure, which further helps in getting accurate fiber tracts; But, this comes at the cost of larger acquisition time and larger setup. Hence, we propose a novel deep learning architecture for the reconstruction of diffusion MRI volumes for different b-values (degree of diffusion weighting) using acquisitions at a fixed b-value (termed as single-shell) acquisition. This reconstruction has been performed in the spherical harmonics space to better manage varying gradient directions. In this work, we have demonstrated such a reconstruction for b = 3000 s/mm² and b = 2000 s/mm² from b = 1000 s/mm². The proposed Multilevel Hierarchical Spherical Harmonics Coefficients Reconstruction (MHSH) framework takes advantage of contextual information within each slice as well as across the slices by involving Slice Level ReconNet (SLRNet) network and a Volumetric ROI Level ReconNet (VPLRNet) network, respectively. Three-loss functions have been used to optimize network learning, i.e., L₁, Adversarial, and Total Variation Loss. Finally, the network is trained and validated on the publicly available HCP data-set with standard qualitative and quantitative performance measures and achieves promising results.

Training data distribution significantly impacts the estimation of tissue microstructure with machine learning

PURPOSE

Supervised machine learning (ML) provides a compelling alternative to traditional model fitting for parameter mapping in quantitative MRI. The aim of this work is to demonstrate and quantify the effect of different training data distributions on the accuracy and precision of parameter estimates when supervised ML is used for fitting.

METHODS

We fit a two- and three-compartment biophysical model to diffusion measurements from in-vivo human brain, as well as simulated diffusion data, using both traditional model fitting and supervised ML. For supervised ML, we train several artificial neural networks, as well as random forest regressors, on different distributions of ground truth parameters. We compare the accuracy and precision of parameter estimates obtained from the different estimation approaches using synthetic test data.

RESULTS

When the distribution of parameter combinations in the training set matches those observed in healthy human data sets, we observe high precision, but inaccurate estimates for atypical parameter combinations. In contrast, when training data is sampled uniformly from the entire plausible parameter space, estimates tend to be more accurate for atypical parameter combinations but may have lower precision for typical parameter combinations.

CONCLUSION

This work highlights that estimation of model parameters using supervised ML depends strongly on the training-set distribution. We show that high precision obtained using ML may mask strong bias, and visual assessment of the parameter maps is not sufficient for evaluating the quality of the estimates.

Deep Learning and Its Application to Function Approximation for MR in Medicine: An Overview

This article presents an overview of deep learning (DL) and its applications to function approximation for MR in medicine. The aim of this article is to help readers develop various applications of DL. DL has made a large impact on the literature of many medical sciences, including MR. However, its technical details are not easily understandable for non-experts of machine learning (ML).The first part of this article presents an overview of DL and its related technologies, such as artificial intelligence (AI) and ML. AI is explained as a function that can receive many inputs and produce many outputs. ML is a process of fitting the function to training data. DL is a kind of ML, which uses a composite of many functions to approximate the function of interest. This composite function is called a deep neural network (DNN), and the functions composited into a DNN are called layers. This first part also covers the underlying technologies required for DL, such as loss functions, optimization, initialization, linear layers, non-linearities, normalization, recurrent neural networks, regularization, data augmentation, residual connections, autoencoders, generative adversarial networks, model and data sizes, and complex-valued neural networks.The second part of this article presents an overview of the applications of DL in MR and explains how functions represented as DNNs are applied to various applications, such as RF pulse, pulse sequence, reconstruction, motion correction, spectroscopy, parameter mapping, image synthesis, and segmentation.