Progressive Dual Priori Network for Generalized Breast Tumor Segmentation.

To promote the generalization ability of breast tumor segmentation models, as well as to improve the segmentation performance for breast tumors with smaller size, low-contrast and irregular shape, we propose a progressive dual priori network (PDPNet) to segment breast tumors from dynamic enhanced magnetic resonance images (DCE-MRI) acquired at different centers. The PDPNet first cropped tumor regions with a coarse-segmentation based localization module, then the breast tumor mask was progressively refined by using the weak semantic priori and cross-scale correlation prior knowledge. To validate the effectiveness of PDPNet, we compared it with several state-of-the-art methods on multi-center datasets. The results showed that, comparing against the suboptimal method, the DSC and HD95 of PDPNet were improved at least by 5.13% and 7.58% respectively on multi-center test sets. In addition, through ablations, we demonstrated that the proposed localization module can decrease the influence of normal tissues and therefore improve the generalization ability of the model. The weak semantic priors allow focusing on tumor regions to avoid missing small tumors and low-contrast tumors. The cross-scale correlation priors are beneficial for promoting the shape-aware ability for irregular tumors. Thus integrating them in a unified framework improved the multi-center breast tumor segmentation performance. The source code and open data can be accessed at https://github.com/wangli100209/PDPNet.

Lightweight semantic segmentation network for tumor cell nuclei and skin lesion.

In the field of medical image segmentation, achieving fast and accurate semantic segmentation of tumor cell nuclei and skin lesions is of significant importance. However, the considerable variations in skin lesion forms and cell types pose challenges to attaining high network accuracy and robustness. Additionally, as network depth increases, the growing parameter size and computational complexity make practical implementation difficult. To address these issues, this paper proposes MD-UNet, a fast cell nucleus segmentation network that integrates Tokenized Multi-Layer Perceptron modules, attention mechanisms, and Inception structures. Firstly, tokenized MLP modules are employed to label and project convolutional features, reducing computational complexity. Secondly, the paper introduces Depthwise Attention blocks and Multi-layer Feature Extraction modules. The Depthwise Attention blocks eliminate irrelevant and noisy responses from coarse-scale extracted information, serving as alternatives to skip connections in the UNet architecture. The Multi-layer Feature Extraction modules capture a wider range of high-level and low-level semantic features during decoding and facilitate feature fusion. The proposed MD-UNet approach is evaluated on two datasets: the International Skin Imaging Collaboration (ISIC2018) dataset and the PanNuke dataset. The experimental results demonstrate that MD-UNet achieves the best performance on both datasets.

Glioma segmentation based on dense contrastive learning and multimodal features recalibration.

Accurate segmentation of different regions of gliomas from multimodal magnetic resonance (MR) images is crucial for glioma grading and precise diagnosis, but many existing segmentation methods are difficult to effectively utilize multimodal MR image information to recognize accurately the lesion regions with small size, low contrast and irregular shape. To address this issue, this work proposes a novel 3D glioma segmentation model DCL-MANet. DCL-MANet has an architecture of multiple encoders and one single decoder. Each encoder is used to extract MR image features of a given modality. To overcome the entangle problems of multimodal semantic features, a dense contrastive learning (DCL) strategy is presented to extract the modality-specific and common features. Following that, feature recalibration block (RFB) based on modality-wise attention is used to recalibrate the semantic features of each modality, enabling the model to focus on the features that are beneficial for glioma segmentation. These recalibrated features are input into the decoder to obtain the segmentation results. To verify the superiority of the proposed method, we compare it with several state-of-the-art (SOTA) methods in terms of Dice, average symmetric surface distance (ASSD), HD95 and volumetric similarity (Vs). The comparison results show that the average Dice, ASSD, HD95 and Vs of DCL-MANet on all tumor regions are improved at least by 0.66%, 3.47%, 8.94% and 1.07% respectively. For small enhance tumor (ET) region, the corresponding improvement can be up to 0.37%, 7.83%, 11.32%, and 1.35%, respectively. In addition, the ablation results demonstrate the effectiveness of the proposed DCL and RFB, and combining them can significantly increase Dice (1.59%) and Vs (1.54%) while decreasing ASSD (40.51%) and HD95 (45.16%) on ET region. The proposed DCL-MANet could disentangle multimodal features and enhance the semantics of modality-dependent features, providing a potential means to accurately segment small lesion regions in gliomas.

Q-space imaging based on Gaussian radial basis function with Laplace regularization.

PURPOSE: To introduce the diffusion signal characteristics presented by spherical harmonics (SH) basis into the q-space imaging method based on Gaussian radial basis function (GRBF) to robustly reconstruct ensemble average diffusion propagator (EAP) in diffusion MRI (dMRI). METHODS: We introduced the Laplacian regularization of the signal into the dMRI imaging method based on GRBF, and derived the relevant indicators of microstructure imaging and the orientation distribution function (ODF) providing fiber bundle direction information based on EAP. In addition, this method is combined with a multi-compartment model to calculate the diameter of fiber bundle axons. The evaluation of the results included qualitative comparisons and quantitative assessments of the signal fitting. RESULTS: The results show that the proposed method achieves the more significant accuracy improvement in reconstructing signal. Meanwhile, ODFs estimated by the proposed method show the sharper profiles and less spurious peaks, even under the sparse and noisy conditions. In the 36 sets of axon diameter estimation experiments, 34 and 30 sets of results showed that the proposed method reduced the mean and SD of axon diameter estimates, respectively. Moreover, compared with the current state-of-the-art method, the mean and SD of axon diameter estimated by the proposed method are mostly lower, with 32 and 29 of 36 groups. CONCLUSION: The proposed method outperforms the GRBF regarding signal fitting and the estimation of the EAP and ODF with multi-shell sparse samples. Moreover, it shows the potential to recover important features of microstructures with less uncertainty by using proposed method together with multi-compartment models.

Brain fiber structure estimation based on principal component analysis and RINLM filter.

Diffusion magnetic resonance imaging is a technique for non-invasive detection of microstructure in the white matter of the human brain, which is widely used in neuroscience research of the brain. However, diffusion-weighted images(DWI) are sensitive to noise, which affects the subsequent reconstruction of fiber orientation direction, microstructural parameter estimation and fiber tracking. In order to better eliminate the noise in diffusion-weighted images, this study proposes a noise reduction method combining Marchenko-Pastur principal component analysis(MPPCA) and rotation-invariant non-local means filter(RINLM) to further remove residual noise and preserve the image texture detail information. In this study, the algorithm is applied to the fiber structure and the prevailing microstructural models within the human brain voxels based on simulated and real human brain datasets. Experimental comparisons between the proposed method and the state-of-the-art methods are performed in single-fiber, multi-fiber, crossed and curved-fiber regions as well as in different microstructure estimation models. Results demonstrated the superior performance of the proposed method in denoising DWI data, which can reduce the angular error in fiber orientation reconstruction to obtain more valid fiber structure estimation and enable more complete fiber tracking trajectories with higher coverage. Meanwhile, the method reduces the estimation errors of various white matter microstructural parameters and verifies the performance of the method in white matter microstructure estimation.

A Channel-Dimensional Feature-Reconstructed Deep Learning Model for Predicting Breast Cancer Molecular Subtypes on Overall b-Value Diffusion-Weighted MRI.

BACKGROUND: Dynamic contrast-enhanced (DCE) MRI commonly outperforms diffusion-weighted (DW) MRI in breast cancer discrimination. However, the side effects of contrast agents limit the use of DCE-MRI, particularly in patients with chronic kidney disease. PURPOSE: To develop a novel deep learning model to fully exploit the potential of overall b-value DW-MRI without the need for a contrast agent in predicting breast cancer molecular subtypes and to evaluate its performance in comparison with DCE-MRI. STUDY TYPE: Prospective. SUBJECTS: 486 female breast cancer patients (training/validation/test: 64%/16%/20%). FIELD STRENGTH/SEQUENCE: 3.0 T/DW-MRI (13 b-values) and DCE-MRI (one precontrast and five postcontrast phases). ASSESSMENT: The breast cancers were divided into four subtypes: luminal A, luminal B, HER2+, and triple negative. A channel-dimensional feature-reconstructed (CDFR) deep neural network (DNN) was proposed to predict these subtypes using pathological diagnosis as the reference standard. Additionally, a non-CDFR DNN (NCDFR-DNN) was built for comparative purposes. A mixture ensemble DNN (ME-DNN) integrating two CDFR-DNNs was constructed to identify subtypes on multiparametric MRI (MP-MRI) combing DW-MRI and DCE-MRI. STATISTICAL TESTS: Model performance was evaluated using accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). Model comparisons were performed using the one-way analysis of variance with least significant difference post hoc test and the DeLong test. P < 0.05 was considered significant. RESULTS: The CDFR-DNN (accuracies, 0.79 ~ 0.80; AUCs, 0.93 ~ 0.94) demonstrated significantly improved predictive performance than the NCDFR-DNN (accuracies, 0.76 ~ 0.78; AUCs, 0.92 ~ 0.93) on DW-MRI. Utilizing the CDFR-DNN, DW-MRI attained the predictive performance equal (P = 0.065 ~ 1.000) to DCE-MRI (accuracies, 0.79 ~ 0.80; AUCs, 0.93 ~ 0.95). The predictive performance of the ME-DNN on MP-MRI (accuracies, 0.85 ~ 0.87; AUCs, 0.96 ~ 0.97) was superior to those of both the CDFR-DNN and NCDFR-DNN on either DW-MRI or DCE-MRI. DATA CONCLUSION: The CDFR-DNN enabled overall b-value DW-MRI to achieve the predictive performance comparable to DCE-MRI. MP-MRI outperformed DW-MRI and DCE-MRI in subtype prediction. LEVEL OF EVIDENCE: 2 TECHNICAL EFFICACY STAGE: 1.

nnUnetFormer: an automatic method based on nnUnet and transformer for brain tumor segmentation with multimodal MR images.

Objective. Both local and global context information is crucial semantic features for brain tumor segmentation, while almost all the CNN-based methods cannot learn global spatial dependencies very well due to the limitation of convolution operations. The purpose of this paper is to build a new framework to make full use of local and global features from multimodal MR images for improving the performance of brain tumor segmentation.Approach. A new automated segmentation method named nnUnetFormer was proposed based on nnUnet and transformer. It fused transformer modules into the deeper layers of the nnUnet framework to efficiently obtain both local and global features of lesion regions from multimodal MR images.Main results.We evaluated our method on BraTS 2021 dataset by 5-fold cross-validation and achieved excellent performance with Dice similarity coefficient (DSC) 0.936, 0.921 and 0.872, and 95th percentile of Hausdorff distance (HD95) 3.96, 4.57 and 10.45 for the regions of whole tumor (WT), tumor core (TC), and enhancing tumor (ET), respectively, which outperformed recent state-of-the-art methods in terms of both average DSC and average HD95. Besides, ablation experiments showed that fusing transformer into our modified nnUnet framework improves the performance of brain tumor segmentation, especially for the TC region. Moreover, for validating the generalization capacity of our method, we further conducted experiments on FeTS 2021 dataset and achieved satisfactory segmentation performance on 11 unseen institutions with DSC 0.912, 0.872 and 0.759, and HD95 6.16, 8.81 and 38.50 for the regions of WT, TC, and ET, respectively.Significance. Extensive qualitative and quantitative experimental results demonstrated that the proposed method has competitive performance against the state-of-the-art methods, indicating its interest for clinical applications.

Deep wavelet scattering orthogonal fusion network for glioma IDH mutation status prediction.

Accurately predicting the isocitrate dehydrogenase (IDH) mutation status of gliomas is greatly significant for formulating appropriate treatment plans and evaluating the prognoses of gliomas. Although existing studies can accurately predict the IDH mutation status of gliomas based on multimodal magnetic resonance (MR) images and machine learning methods, most of these methods cannot fully explore multimodal information and effectively predict IDH status for datasets acquired from multiple centers. To address this issue, a novel wavelet scattering (WS)-based orthogonal fusion network (WSOFNet) was proposed in this work to predict the IDH mutation status of gliomas from multiple centers. First, transformation-invariant features were extracted from multimodal MR images with a WS network, and then the multimodal WS features were used instead of the original images as the inputs of WSOFNet and were fully fused through an adaptive multimodal feature fusion module (AMF2M) and an orthogonal projection module (OPM). Finally, the fused features were input into a fully connected classifier to predict IDH mutation status. In addition, to achieve improved prediction accuracy, four auxiliary losses were also used in the feature extraction modules. The comparison results showed that the prediction area under the curve (AUC) of WSOFNet on a single-center dataset was 0.9966 and that on a multicenter dataset was approximately 0.9655, which was at least 3.9% higher than that of state-of-the-art methods. Moreover, the ablation experimental results also proved that the adaptive multimodal feature fusion strategy based on orthogonal projection could effectively improve the prediction performance of the model, especially for an external validation dataset.

MD-SGT: Multi-dilation spherical graph transformer for unsupervised medical image registration.

Deformable medical image registration is an essential preprocess step for several clinical applications. Even though the existing convolutional neural network and transformer based methods achieved the promising results, the limited long-range spatial dependence and non-uniform attention span of these models prohibit further improving the registration performance. To deal with this issue, we proposed a multi-dilation spherical graph transformer (MD-SGT), in which the encoder combined the advantages of convolutional and graph transformer blocks to distinguish effectively the differences between the reference and the template images at various scales. Specifically, the features of each voxel were obtained by aggregating the information from its neighbors sampled from different spherical regions with different dilation rates. The implicit convolution inductive bias and long-range uniform attention span induced by such information aggregation manner made the features more representative for registration. Through the qualitative and quantitative comparisons with state-of-the-art methods on two datasets, we demonstrated that combining long-range uniform attention span and inductive bias are beneficial for promoting the image registration performance, with the Dice score, ASD and HD95 being improved at least by 0.5%, 2.2% and 1.1%, respectively.

A Novel Longitudinal Phenotype-Genotype Association Study Based on Deep Feature Extraction and Hypergraph Models for Alzheimer's Disease.

Traditional image genetics primarily uses linear models to investigate the relationship between brain image data and genetic data for Alzheimer's disease (AD) and does not take into account the dynamic changes in brain phenotype and connectivity data across time between different brain areas. In this work, we proposed a novel method that combined Deep Subspace reconstruction with Hypergraph-Based Temporally-constrained Group Sparse Canonical Correlation Analysis (DS-HBTGSCCA) to discover the deep association between longitudinal phenotypes and genotypes. The proposed method made full use of dynamic high-order correlation between brain regions. In this method, the deep subspace reconstruction technique was applied to retrieve the nonlinear properties of the original data, and hypergraphs were used to mine the high-order correlation between two types of rebuilt data. The molecular biological analysis of the experimental findings demonstrated that our algorithm was capable of extracting more valuable time series correlation from the real data obtained by the AD neuroimaging program and finding AD biomarkers across multiple time points. Additionally, we used regression analysis to verify the close relationship between the extracted top brain areas and top genes and found the deep subspace reconstruction approach with a multi-layer neural network was helpful in enhancing clustering performance.

msQSM: Morphology-based self-supervised deep learning for quantitative susceptibility mapping.

Quantitative susceptibility mapping (QSM) has been applied to the measurement of iron deposition and the auxiliary diagnosis of neurodegenerative disease. There still exists a dipole inversion problem in QSM reconstruction. Recently, deep learning approaches have been proposed to resolve this problem. However, most of these approaches are supervised methods that need pairs of the input phase and ground-truth. It remains a challenge to train a model for all resolutions without using the ground-truth and only using one resolution data. To address this, we proposed a self-supervised QSM deep learning method based on morphology. It consists of a morphological QSM builder to decouple the dependency of the QSM on acquisition resolution, and a morphological loss to reduce artifacts effectively and save training time efficiently. The proposed method can reconstruct arbitrary resolution QSM on both human data and animal data, regardless of whether the resolution is higher or lower than that of the training set. Our method outperforms the previous best unsupervised method with a 3.6% higher peak signal-to-noise ratio, 16.2% lower normalized root mean square error, and 22.1% lower high-frequency error norm. The morphological loss reduces training time by 22.1% with respect to the cycle gradient loss used in the previous unsupervised methods. Experimental results show that the proposed method accurately measures QSM with arbitrary resolutions, and achieves state-of-the-art results among unsupervised deep learning methods. Research on applications in neurodegenerative diseases found that our method is robust enough to measure significant increase in striatal magnetic susceptibility in patients during Alzheimer's disease progression, as well as significant increase in substantia nigra susceptibility in Parkinson's disease patients, and can be used as an auxiliary differential diagnosis tool for Alzheimer's disease and Parkinson's disease.

Semi-supervised super-resolution of diffusion-weighted images based on multiple references.

Spatial resolution of diffusion tensor images is usually compromised to accelerate the acquisitions, and the state-of-the-art (SOTA) image super-resolution (SR) reconstruction methods are commonly based on supervised learning models. Considering that matched low-resolution (LR) and high-resolution (HR) diffusion-weighted (DW) image pairs are not readily available, we propose a semi-supervised DW image SR reconstruction method based on multiple references (MRSR) extracted from other subjects. In MRSR, the prior information of multiple HR reference images is migrated into a residual-like network to assist SR reconstruction of DW images, and a CycleGAN-based semi-supervised strategy is used to train the network with 30% matched and 70% unmatched LR-HR image pairs. We evaluate the performance of the MRSR by comparing against SOTA methods on an HCP dataset in terms of the quality of reconstructed DW images and diffusion metrics. MRSR achieves the best performance, with the mean PSNR/SSIM of DW images being improved by at least 14.3%/28.8% and 1%/1.4% respectively relative to SOTA unsupervised and supervised learning methods, and with the fiber orientations deviating from the ground truth by about 6.28° on average, the RMSEs of fractional anisotropy, mean diffusivity, axial diffusivity and radial diffusivity being 3.0%, 4.6%, 5.7% and 4.5% respectively relative to the ground truth. We validate the effectiveness of the proposed network structure, multiple-reference and CycleGAN-based semi-supervised learning strategies for SR reconstruction of diffusion tensor images through the ablation studies. The proposed method allows us to achieve SR reconstruction for diffusion tensor images with a limited number of matched image pairs.

Self-supervised structural similarity-based convolutional neural network for cardiac diffusion tensor image denoising.

BACKGROUND: Diffusion tensor imaging (DTI) is a promising technique for non-invasively investigating the myocardial fiber structures of human heart. However, low signal-to-noise ratio (SNR) has been a major limit of cardiac DTI to prevent us from detecting myocardium structure accurately. Therefore, it is important to remove the effect of noise on diffusion weighted (DW) images. PURPOSE: Although the conventional and deep learning-based denoising methods have shown the potential to deal with effectively the noise in DW images, most of them are redundant information dependent or require the noise-free images as golden standard. In addition, the existed DW image denoising methods often suffer from problems of over-smoothing. To address these issues, we propose a self-supervised learning model, structural similarity based convolutional neural network with edge-weighted loss (SSECNN), to remove the noise effectively in cardiac DTI. METHODS: Considering that the DW images acquired along different diffusion directions have structural similarity, and the noise in these DW images is independent and identically distributed, the structural similarity-based matching algorithm is proposed to search for the most similar DW images. Such similar noisy DW image pairs are then used as the input and target of the denoising network SSECNN, which consists of several convolutional and residual blocks. Through the self-supervised training with these image pairs, the network can restore the clean DW images and retain the correlations between the denoised DW images along different directions. To avoid the over-smoothing problem, we design a novel edge-weighted loss which enables the network to adaptively adjust the loss weights with iterations and therefore to improve the detail preserve ability of the model. To verify the superiority of the proposed method, comparisons with state-of-the-art (SOTA) denoising methods are performed on both synthetic and real acquired DTI datasets. RESULTS: Experimental results show that SSECNN can effectively reduce the noise in the DW images while preserving detailed texture and edge information and therefore achieve better performance in DTI reconstruction. For synthetic dataset, compared to the SOTA method, the root mean square error (RMSE), peak signal-to-noise ratio (PSNR), and structure similarity index measure (SSIM) between the denoised DW images obtained with SSECNN and noise-free DW images are improved by 6.94%, 1.98%, and 0.76% respectively when the noise level is 10%. As for the acquired cardiac DTI dataset, the SSECNN method could significantly improve SNR and contrast to noise ratio (CNR) of cardiac DW images and achieve more regular helix angle (HA) and transverse angle (TA) maps. The ablation experimental results validate that using the structure similarity-based method to search the similar DW image pairs yield the smallest loss, and with the help of the edge-weighted loss, the denoised DW images and diffusion metric maps can preserve more details. CONCLUSIONS: The proposed SSECNN method can fully explore the similarity between the DW images along different diffusion directions. Using such similarity and an edge-weighted loss enable us to denoise cardiac DTI effectively in a self-supervised manner. Our method can overcome the redundancy information dependence and over-smoothing problem of the SOTA methods.

Comparison of Dynamic Contrast-Enhanced MRI and Non-Mono-Exponential Model-Based Diffusion-Weighted Imaging for the Prediction of Prognostic Biomarkers and Molecular Subtypes of Breast Cancer Based on Radiomics.

BACKGROUND: Dynamic contrast-enhanced (DCE) MRI and non-mono-exponential model-based diffusion-weighted imaging (NME-DWI) that does not require contrast agent can both characterize breast cancer. However, which technique is superior remains unclear. PURPOSE: To compare the performances of DCE-MRI, NME-DWI and their combination as multiparametric MRI (MP-MRI) in the prediction of breast cancer prognostic biomarkers and molecular subtypes based on radiomics. STUDY TYPE: Prospective. POPULATION: A total of 477 female patients with 483 breast cancers (5-fold cross-validation: training/validation, 80%/20%). FIELD STRENGTH/SEQUENCE: A 3.0 T/DCE-MRI (6 dynamic frames) and NME-DWI (13 b values). ASSESSMENT: After data preprocessing, high-throughput features were extracted from each tumor volume of interest, and optimal features were selected using recursive feature elimination method. To identify ER+ vs. ER-, PR+ vs. PR-, HER2+ vs. HER2-, Ki-67+ vs. Ki-67-, luminal A/B vs. nonluminal A/B, and triple negative (TN) vs. non-TN, the following models were implemented: random forest, adaptive boosting, support vector machine, linear discriminant analysis, and logistic regression. STATISTICAL TESTS: Student's t, chi-square, and Fisher's exact tests were applied on clinical characteristics to confirm whether significant differences exist between different statuses (±) of prognostic biomarkers or molecular subtypes. The model performances were compared between the DCE-MRI, NME-DWI, and MP-MRI datasets using the area under the receiver-operating characteristic curve (AUC) and the DeLong test. P < 0.05 was considered significant. RESULTS: With few exceptions, no significant differences (P = 0.062-0.984) were observed in the AUCs of models for six classification tasks between the DCE-MRI (AUC = 0.62-0.87) and NME-DWI (AUC = 0.62-0.91) datasets, while the model performances on the two imaging datasets were significantly poorer than on the MP-MRI dataset (AUC = 0.68-0.93). Additionally, the random forest and adaptive boosting models (AUC = 0.62-0.93) outperformed other three models (AUC = 0.62-0.90). DATA CONCLUSION: NME-DWI was comparable with DCE-MRI in predictive performance and could be used as an alternative technique. Besides, MP-MRI demonstrated significantly higher AUCs than either DCE-MRI or NME-DWI. EVIDENCE LEVEL: 2. TECHNICAL EFFICACY: Stage 2.

Radiomic analysis based on multi-phase magnetic resonance imaging to predict preoperatively microvascular invasion in hepatocellular carcinoma.

BACKGROUND: The prognosis of hepatocellular carcinoma (HCC) remains poor and relapse occurs in more than half of patients within 2 years after hepatectomy. In terms of recent studies, microvascular invasion (MVI) is one of the potential predictors of recurrence. Accurate preoperative prediction of MVI is potentially beneficial to the optimization of treatment planning. AIM: To develop a radiomic analysis model based on pre-operative magnetic resonance imaging (MRI) data to predict MVI in HCC. METHODS: A total of 113 patients recruited to this study have been diagnosed as having HCC with histological confirmation, among whom 73 were found to have MVI and 40 were not. All the patients received preoperative examination by Gd-enhanced MRI and then curative hepatectomy. We manually delineated the tumor lesion on the largest cross-sectional area of the tumor and the adjacent two images on MRI, namely, the regions of interest. Quantitative analyses included most discriminant factors (MDFs) developed using linear discriminant analysis algorithm and histogram analysis with MaZda software. Independent significant variables of clinical and radiological features and MDFs for the prediction of MVI were estimated and a discriminant model was established by univariate and multivariate logistic regression analysis. Prediction ability of the above-mentioned parameters or model was then evaluated by receiver operating characteristic (ROC) curve analysis. Five-fold cross-validation was also applied via R software. RESULTS: The area under the ROC curve (AUC) of the MDF (0.77-0.85) outperformed that of histogram parameters (0.51-0.74). After multivariate analysis, MDF values of the arterial and portal venous phase, and peritumoral hypointensity in the hepatobiliary phase were identified to be independent predictors of MVI (P < 0.05). The AUC value of the model was 0.939 [95% confidence interval (CI): 0.893-0.984, standard error: 0.023]. The result of internal five-fold cross-validation (AUC: 0.912, 95%CI: 0.841-0.959, standard error: 0.0298) also showed favorable predictive efficacy. CONCLUSION: Noninvasive MRI radiomic model based on MDF values and imaging biomarkers may be useful to make preoperative prediction of MVI in patients with primary HCC.

Learn Less, Infer More: Learning in the Fourier Domain for Quantitative Susceptibility Mapping.

Quantitative susceptibility mapping (QSM) aims to evaluate the distribution of magnetic susceptibility from magnetic resonance phase measurements by solving the ill-conditioned dipole inversion problem. Removing the artifacts and preserving the anisotropy of tissue susceptibility simultaneously is still a challenge in QSM. To deal with this issue, a novel k-QSM network is proposed to resolve dipole inversion issues in QSM reconstruction. The k-QSM network converts the results obtained by truncated k-space division (TKD) into the Fourier domain as inputs. After passing through several convolutional and residual blocks, the ill-posed signals of TKD are corrected by making the network output close to the calculation of susceptibility through multiple orientation sampling (COSMOS)-labeled QSM. To evaluate the superiority of k-QSM, comparisons with several state-of-the-art methods are performed in terms of QSM artifacts removing, anisotropy preserving, generalization ability, and clinical applications. Compared to existing methods, the k-QSM achieves a 22.31% lower normalized root mean square error, 10.30% higher peak signal-to-noise ratio (PSNR), 33.10% lower high-frequency error norm, and 1.06% higher structural similarity. In addition, the orientation-dependent susceptibility variation obtained by k-QSM is significant, verifying that k-QSM has the ability to preserve susceptibility anisotropy. When the trained models are tested on the dataset from different centers, our k-QSM shows a strong generalization ability with the highest PSNR. Moreover, by comparing the susceptibility maps between healthy controls and drug addicts with different methods, we found the proposed k-QSM is more sensitive to the susceptibility abnormality in the patients. The proposed k-QSM method learns less-only to fix the ill-posed signals of TKD, but infers more-both COSMOS-like and anisotropy-preserving QSM results. Its generalization ability and great sensitivity to susceptibility changes can make it a potential method for distinguishing some diseases.

Multimodal MRI Image Decision Fusion-Based Network for Glioma Classification.

Purpose: Glioma is the most common primary brain tumor, with varying degrees of aggressiveness and prognosis. Accurate glioma classification is very important for treatment planning and prognosis prediction. The main purpose of this study is to design a novel effective algorithm for further improving the performance of glioma subtype classification using multimodal MRI images. Method: MRI images of four modalities for 221 glioma patients were collected from Computational Precision Medicine: Radiology-Pathology 2020 challenge, including T1, T2, T1ce, and fluid-attenuated inversion recovery (FLAIR) MRI images, to classify astrocytoma, oligodendroglioma, and glioblastoma. We proposed a multimodal MRI image decision fusion-based network for improving the glioma classification accuracy. First, the MRI images of each modality were input into a pre-trained tumor segmentation model to delineate the regions of tumor lesions. Then, the whole tumor regions were centrally clipped from original MRI images followed by max-min normalization. Subsequently, a deep learning-based network was designed based on a unified DenseNet structure, which extracts features through a series of dense blocks. After that, two fully connected layers were used to map the features into three glioma subtypes. During the training stage, we used the images of each modality after tumor segmentation to train the network to obtain its best accuracy on our testing set. During the inferring stage, a linear weighted module based on a decision fusion strategy was applied to assemble the predicted probabilities of the pre-trained models obtained in the training stage. Finally, the performance of our method was evaluated in terms of accuracy, area under the curve (AUC), sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), etc. Results: The proposed method achieved an accuracy of 0.878, an AUC of 0.902, a sensitivity of 0.772, a specificity of 0.930, a PPV of 0.862, an NPV of 0.949, and a Cohen's Kappa of 0.773, which showed a significantly higher performance than existing state-of-the-art methods. Conclusion: Compared with current studies, this study demonstrated the effectiveness and superiority in the overall performance of our proposed multimodal MRI image decision fusion-based network method for glioma subtype classification, which would be of enormous potential value in clinical practice.

Connecting macroscopic diffusion metrics of cardiac diffusion tensor imaging and microscopic myocardial structures based on simulation.

To investigate the relationship between microscopic myocardial structures and macroscopic measurements of diffusion tensor imaging (DTI), we proposed a cardiac DTI simulation method using the Bloch equation and the Monte Carlo random walk in a realistic myocardium model reconstructed from polarized light imaging (PLI) data of the entire human heart. To obtain a realistic simulation, with the constraints of prior knowledge pertaining to the maturational change of the myocardium structure, appropriate microstructure modeling parameters were iteratively determined by matching DTI simulations and real acquisitions of the same hearts in terms of helix angle, fractional anisotropy (FA) and mean diffusivity (MD) maps. Once a realistic simulation was obtained, we varied the extra-cellular volume (ECV) ratio, myocyte orientation heterogeneity and myocyte size, and explored the effects of microscopic changes in tissue structure on macroscopic diffusion metrics. The experimental results demonstrated the feasibility of simulating the DTI of the whole heart using PLI measurements. When varying ECV from 15% to 55%, mean FA decreased from 0.55 to 0.26, axial diffusivity increased by 0.6 µm2/ms, and radial diffusivity increased by 0.7 µm2/ms. When orientation heterogeneity was varied from 0 to 20∘, mean FA decreased from 0.4 to 0.3, axial diffusivity decreased by 0.08 µm2/ms, and radial diffusivity increased by 0.03 µm2/ms. When mean diameter of myocytes was varied from 6 µm to 10 µm, FA decreased from 0.67 to 0.46, axial and radial diffusivities increased by 0.05 and 0.2 µm2/ms, respectively.

An Unsupervised Deep Learning Approach for Dynamic-Exponential Intravoxel Incoherent Motion MRI Modeling and Parameter Estimation in the Liver.

BACKGROUND: Dynamic-exponential intravoxel incoherent motion (IVIM) imaging is a potential technique for prediction, monitoring, and differential diagnosis of hepatic diseases, especially liver tumors. However, the use of such technique at voxel level is still limited. PURPOSE: To develop an unsupervised deep learning approach for voxel-wise dynamic-exponential IVIM modeling and parameter estimation in the liver. STUDY TYPE: Prospective. POPULATION: Ten healthy subjects (4 males; age 28 ± 6 years). FIELD STRENGTH/SEQUENCE: Single-shot spin-echo echo planar imaging (SE-EPI) sequence with monopolar diffusion-encoding gradients (12 b-values, 0-800 seconds/mm2 ) at 3.0 T. ASSESSMENT: The proposed deep neural network (DNN) was separately trained on simulated and in vivo hepatic IVIM datasets. The trained networks were compared to the approach combining least squares with Akaike information criterion (LSQ-AIC) in terms of dynamic-exponential modeling accuracy, inter-subject coefficients of variation (CVs), and fitting residuals on the simulated subsets and regions of interest (ROIs) in the left and right liver lobes. The ROIs were delineated by a radiologist (H.-X.Z.) with 7 years of experience in MRI reading. STATISTICAL TESTS: Comparisons between approaches were performed with a paired t-test (normality) or a Wilcoxon rank-sum test (nonnormality). P < 0.05 was considered statistically significant. RESULTS: In simulations, DNN gave significantly higher accuracy (91.6%-95.5%) for identification of bi-exponential decays with respect to LSQ-AIC (79.7%-86.8%). For tri-exponential identification, DNN was also superior to LSQ-AIC despite not reaching a significant level (P = 0.08). Additionally, DNN always yielded comparatively low root-mean-square error for estimated parameters. For the in vivo IVIM measurements, inter-subject CVs (0.011-0.150) of DNN were significantly smaller than those (0.049-0.573) of LSQ-AIC. Concerning fitting residuals, there was no significant difference between the two approaches (P = 0.56 and 0.76) in both the simulated and in vivo studies. DATA CONCLUSION: The proposed DNN is recommended for accurate and robust dynamic-exponential modeling and parameter estimation in hepatic IVIM imaging. LEVEL OF EVIDENCE: 2 TECHNICAL EFFICACY: Stage 1.

Short-Term Repeatability of in Vivo Cardiac Intravoxel Incoherent Motion Tensor Imaging in Healthy Human Volunteers.

BACKGROUND: Intravoxel incoherent motion (IVIM) tensor imaging is a promising technique for diagnosis and monitoring of cardiovascular diseases. Knowledge about measurement repeatability, however, remains limited. PURPOSE: To evaluate short-term repeatability of IVIM tensor imaging in normal in vivo human hearts. STUDY TYPE: Prospective. POPULATION: Ten healthy subjects without history of heart diseases. FIELD STRENGTH/SEQUENCE: Balanced steady-state free-precession cine sequence and single-shot spin-echo echo planar IVIM tensor imaging sequence (9 b-values, 0-400 seconds/mm2 and six diffusion-encoding directions) at 3.0 T. ASSESSMENT: Subjects were scanned twice with an interval of 15 minutes, leaving the scanner between studies. The signal-to-noise ratio (SNR) was evaluated in anterior, lateral, septal, and inferior segments of the left ventricle wall. Fractional anisotropy (FA), mean diffusivity (MD), mean fraction (MF), and helix angle (HA) in the four segments were independently measured by five radiologists. STATISTICAL TESTS: IVIM tensor indexes were compared between observers using a one-way analysis of variance or between scans using a paired t-test (normal data) or a Wilcoxon rank-sum test (non-normal data). Interobserver agreement and test-retest repeatability were assessed using the intraclass correlation coefficient (ICC), within-subject coefficient of variation (WCV), and Bland-Altman limits of agreements. RESULTS: SNR of inferior segment was significantly lower than the other three segments, and inferior segment was therefore excluded from repeatability analysis. Interobserver repeatability was excellent for all IVIM tensor indexes (ICC: 0.886-0.972; WCV: 0.62%-4.22%). Test-retest repeatability was excellent for MD of the self-diffusion tensor (D) and MF of the perfusion fraction tensor (fp ) (ICC: 0.803-0.888; WCV: 1.42%-9.51%) and moderate for FA and MD of the pseudo-diffusion tensor (D* ) (ICC: 0.487-0.532; WCV: 6.98%-10.89%). FA of D and fp and HA of D presented good test-retest repeatability (ICC: 0.732-0.788; WCV: 3.28%-8.71%). DATA CONCLUSION: The D and fp indexes exhibited satisfactory repeatability, but further efforts were needed to improve repeatability of D* indexes. LEVEL OF EVIDENCE: 2 TECHNICAL EFFICACY: Stage 1.