Filter Pruning Based on Information Capacity and Independence.

Filter pruning has gained widespread adoption for the purpose of compressing and speeding up convolutional neural networks (CNNs). However, the existing approaches are still far from practical applications due to biased filter selection and heavy computation cost. This article introduces a new filter pruning method that selects filters in an interpretable, multiperspective, and lightweight manner. Specifically, we evaluate the contributions of filters from both individual and overall perspectives. For the amount of information contained in each filter, a new metric called information capacity is proposed. Inspired by the information theory, we utilize the interpretable entropy to measure the information capacity and develop a feature-guided approximation process. For correlations among filters, another metric called information independence is designed. Since the aforementioned metrics are evaluated in a simple but effective way, we can identify and prune the least important filters with less computation cost. We conduct comprehensive experiments on benchmark datasets employing various widely used CNN architectures to evaluate the performance of our method. For instance, on ILSVRC-2012, our method outperforms state-of-the-art methods by reducing floating-point operations (FLOPs) by 77.4% and parameters by 69.3% for ResNet-50 with only a minor decrease in an accuracy of 2.64%.

Finding core labels for maximizing generalization of graph neural networks.

Graph neural networks (GNNs) have become a popular approach for semi-supervised graph representation learning. GNNs research has generally focused on improving methodological details, whereas less attention has been paid to exploring the importance of labeling the data. However, for semi-supervised learning, the quality of training data is vital. In this paper, we first introduce and elaborate on the problem of training data selection for GNNs. More specifically, focusing on node classification, we aim to select representative nodes from a graph used to train GNNs to achieve the best performance. To solve this problem, we are inspired by the popular lottery ticket hypothesis, typically used for sparse architectures, and we propose the following subset hypothesis for graph data: "There exists a core subset when selecting a fixed-size dataset from the dense training dataset, that can represent the properties of the dataset, and GNNs trained on this core subset can achieve a better graph representation". Equipped with this subset hypothesis, we present an efficient algorithm to identify the core data in the graph for GNNs. Extensive experiments demonstrate that the selected data (as a training set) can obtain performance improvements across various datasets and GNNs architectures.

Diffusion-tensor-imaging 1-year-old and 2-year-old infant brain atlases with comprehensive gray and white matter labels.

Human infancy is marked by fastest postnatal brain structural changes. It also coincides with the onset of many neurodevelopmental disorders. Atlas-based automated structure labeling has been widely used for analyzing various neuroimaging data. However, the relatively large and nonlinear neuroanatomical differences between infant and adult brains can lead to significant offsets of the labeled structures in infant brains when adult brain atlas is used. Age-specific 1- and 2-year-old brain atlases covering all major gray and white matter (GM and WM) structures with diffusion tensor imaging (DTI) and structural MRI are critical for precision medicine for infant population yet have not been established. In this study, high-quality DTI and structural MRI data were obtained from 50 healthy children to build up three-dimensional age-specific 1- and 2-year-old brain templates and atlases. Age-specific templates include a single-subject template as well as two population-averaged templates from linear and nonlinear transformation, respectively. Each age-specific atlas consists of 124 comprehensively labeled major GM and WM structures, including 52 cerebral cortical, 10 deep GM, 40 WM, and 22 brainstem and cerebellar structures. When combined with appropriate registration methods, the established atlases can be used for highly accurate automatic labeling of any given infant brain MRI. We demonstrated that one can automatically and effectively delineate deep WM microstructural development from 3 to 38 months by using these age-specific atlases. These established 1- and 2-year-old infant brain DTI atlases can advance our understanding of typical brain development and serve as clinical anatomical references for brain disorders during infancy.

Multilevel Contrastive Graph Masked Autoencoders for Unsupervised Graph-Structure Learning.

Unsupervised graph-structure learning (GSL) which aims to learn an effective graph structure applied to arbitrary downstream tasks by data itself without any labels' guidance, has recently received increasing attention in various real applications. Although several existing unsupervised GSL has achieved superior performance in different graph analytical tasks, how to utilize the popular graph masked autoencoder to sufficiently acquire effective supervision information from the data itself for improving the effectiveness of learned graph structure has been not effectively explored so far. To tackle the above issue, we present a multilevel contrastive graph masked autoencoder (MCGMAE) for unsupervised GSL. Specifically, we first introduce a graph masked autoencoder with the dual feature masking strategy to reconstruct the same input graph-structured data under the original structure generated by the data itself and learned graph-structure scenarios, respectively. And then, the inter-and intra-class contrastive loss is introduced to maximize the mutual information in feature and graph-structure reconstruction levels simultaneously. More importantly, the above inter-and intra-class contrastive loss is also applied to the graph encoder module for further strengthening their agreement at the feature-encoder level. In comparison to the existing unsupervised GSL, our proposed MCGMAE can effectively improve the training robustness of the unsupervised GSL via different-level supervision information from the data itself. Extensive experiments on three graph analytical tasks and eight datasets validate the effectiveness of the proposed MCGMAE.

Sparse Additive Machine With the Correntropy-Induced Loss.

Sparse additive machines (SAMs) have shown competitive performance on variable selection and classification in high-dimensional data due to their representation flexibility and interpretability. However, the existing methods often employ the unbounded or nonsmooth functions as the surrogates of 0-1 classification loss, which may encounter the degraded performance for data with outliers. To alleviate this problem, we propose a robust classification method, named SAM with the correntropy-induced loss (CSAM), by integrating the correntropy-induced loss (C-loss), the data-dependent hypothesis space, and the weighted lq,1 -norm regularizer ( q ≥ 1 ) into additive machines. In theory, the generalization error bound is estimated via a novel error decomposition and the concentration estimation techniques, which shows that the convergence rate O(n-1/4) can be achieved under proper parameter conditions. In addition, the theoretical guarantee on variable selection consistency is analyzed. Experimental evaluations on both synthetic and real-world datasets consistently validate the effectiveness and robustness of the proposed approach.

Region-related focal loss for 3D brain tumor MRI segmentation.

BACKGROUND: In the brain tumor magnetic resonance image (MRI) segmentation, although the 3D convolution networks (CNNs) has achieved state-of-the-art results, the class and hard-voxel imbalances in the 3D images have not been well addressed. Voxel independent losses are dependent on the setting of class weights for the class imbalance issue, and are hard to assign each class equally. Region-related losses cannot correctly focus on hard voxels dynamically and not be robust to misclassification of small structures. Meanwhile, repeatedly training on the additional hard samples augmented by existing methods would bring more class imbalance, overfitting and incorrect knowledge learning to the model. PURPOSE: A novel region-related loss with balanced dynamic weighting while alleviating the sensitivity to small structures is necessary. In addition, we need to increase the diversity of hard samples in the training to improve the performance of model. METHODS: The proposed Region-related Focal Loss (RFL) reshapes standard Dice Loss (DL) by up-weighting the loss assigned to hard-classified voxels. Compared to DL, RFL adaptively modulate its gradient with an invariant focalized point that voxels with lower-confidence than it would achieve a larger gradient, and higher-confidence voxels would get a smaller gradient. Meanwhile, RFL can adjust the parameters to set where and how much the network is focused. In addition, an Intra-classly Transformed Augmentation network (ITA-NET) is proposed to increase the diversity of hard samples, in which the 3D registration network and intra-class transfer layer are used to transform the shape and intensity respectively. A selective hard sample mining(SHSM) strategy is used to train the ITA-NET for avoiding excessive class imbalance. Source code (in Tensorflow) is available at: https://github.com/lb-whu/RFL_ITA. RESULTS: The experiments are carried out on public data set: Brain Tumor Segmentation Challenge 2020 (BratS2020). Experiments with BraTS2020 online validation set show that proposed methods achieve an average Dice scores of 0.905, 0.821, and 0.806 for whole tumor (WT), tumor core (TC) and enhancing tumor (ET), respectively. Compared with DL (baseline), the proposed RFL significantly improves the Dice scores by an average of 1%, and for the small region ET it can even increase by 3%. And the proposed method combined with ITA-NET improves the Dice scores of ET and TC by 5% and 3% respectively. CONCLUSIONS: The proposed RFL can converge with a invariant focalized point in the training of segmentation network, thus effectively alleviating the hard-voxel imbalance in brain tumor MRI segmentation. The negative region term of RFL can effectively reduce the sensitivity of the segmentation model to the misclassification of small structures. The proposed ITA-NET can increase the diversity of hard samples by transforming their shape and transfer their intra-class intensity, thereby effectively improving the robustness of the segmentation network to hard samples.

A Componentwise Approach to Weakly Supervised Semantic Segmentation Using Dual-Feedback Network.

Recent weakly supervised semantic segmentation methods generate pseudolabels to recover the lost position information in weak labels for training the segmentation network. Unfortunately, those pseudolabels often contain mislabeled regions and inaccurate boundaries due to the incomplete recovery of position information. It turns out that the result of semantic segmentation becomes determinate to a certain degree. In this article, we decompose the position information into two components: high-level semantic information and low-level physical information, and develop a componentwise approach to recover each component independently. Specifically, we propose a simple yet effective pseudolabels updating mechanism to iteratively correct mislabeled regions inside objects to precisely refine high-level semantic information. To reconstruct low-level physical information, we utilize a customized superpixel-based random walk mechanism to trim the boundaries. Finally, we design a novel network architecture, namely, a dual-feedback network (DFN), to integrate the two mechanisms into a unified model. Experiments on benchmark datasets show that DFN outperforms the existing state-of-the-art methods in terms of intersection-over-union (mIoU).

Discriminative Suprasphere Embedding for Fine-Grained Visual Categorization.

Despite the great success of the existing work in fine-grained visual categorization (FGVC), there are still several unsolved challenges, e.g., poor interpretation and vagueness contribution. To circumvent this drawback, motivated by the hypersphere embedding method, we propose a discriminative suprasphere embedding (DSE) framework, which can provide intuitive geometric interpretation and effectively extract discriminative features. Specifically, DSE consists of three modules. The first module is a suprasphere embedding (SE) block, which learns discriminative information by emphasizing weight and phase. The second module is a phase activation map (PAM) used to analyze the contribution of local descriptors to the suprasphere feature representation, which uniformly highlights the object region and exhibits remarkable object localization capability. The last module is a class contribution map (CCM), which quantitatively analyzes the network classification decision and provides insight into the domain knowledge about classified objects. Comprehensive experiments on three benchmark datasets demonstrate the effectiveness of our proposed method in comparison with state-of-the-art methods.

GNDAN: Graph Navigated Dual Attention Network for Zero-Shot Learning.

Zero-shot learning (ZSL) tackles the unseen class recognition problem by transferring semantic knowledge from seen classes to unseen ones. Typically, to guarantee desirable knowledge transfer, a direct embedding is adopted for associating the visual and semantic domains in ZSL. However, most existing ZSL methods focus on learning the embedding from implicit global features or image regions to the semantic space. Thus, they fail to: 1) exploit the appearance relationship priors between various local regions in a single image, which corresponds to the semantic information and 2) learn cooperative global and local features jointly for discriminative feature representations. In this article, we propose the novel graph navigated dual attention network (GNDAN) for ZSL to address these drawbacks. GNDAN employs a region-guided attention network (RAN) and a region-guided graph attention network (RGAT) to jointly learn a discriminative local embedding and incorporate global context for exploiting explicit global embeddings under the guidance of a graph. Specifically, RAN uses soft spatial attention to discover discriminative regions for generating local embeddings. Meanwhile, RGAT employs an attribute-based attention to obtain attribute-based region features, where each attribute focuses on the most relevant image regions. Motivated by the graph neural network (GNN), which is beneficial for structural relationship representations, RGAT further leverages a graph attention network to exploit the relationships between the attribute-based region features for explicit global embedding representations. Based on the self-calibration mechanism, the joint visual embedding learned is matched with the semantic embedding to form the final prediction. Extensive experiments on three benchmark datasets demonstrate that the proposed GNDAN achieves superior performances to the state-of-the-art methods. Our code and trained models are available at https://github.com/shiming-chen/GNDAN.

Altered brain network in first-episode, drug-naive patients with major depressive disorder.

BACKGROUND: Resting-state functional magnetic resonance imaging has been widely used for the assessment of brain functional network, yet with inconsistent results. The present study aimed to investigate intranetwork and internetwork connectivity differences between patients with major depressive disorder (MDD) and healthy controls at the integrity, network and edge levels of 8 well-defined resting state networks. METHODS: Thirty patients with MDD and sixty-three healthy control subjects were recruited in this study. RESULTS: Compared with healthy controls, patients with MDD showed increased node degree in the right amygdala and putamen, increased connectivity strength in the deep gray matter network (DGN) and increased functional connectivity in intranetwork and internetwork. Meanwhile, MDD showed decreased connectivity strength in visual network-DGN pair. LIMITATIONS: The sample size was small, and all patients in this study were of Asian ethnicity, especially Han individuals. CONCLUSIONS: These findings demonstrate that MDD cases and healthy controls may have divergent intranetwork and internetwork connectivity at an early stage without confounding influence of medication. These differences may underlie cognitive and behavioral alterations in patients with MDD. And these differences may help with the discrimination of patients and healthy people at an early stage of MDD. More studies in the future are warranted to assist in the diagnosis of this burdensome disease.

IAS-NET: Joint intraclassly adaptive GAN and segmentation network for unsupervised cross-domain in neonatal brain MRI segmentation.

PURPOSE: In neonatal brain magnetic resonance image (MRI) segmentation, the model we trained on the training set (source domain) often performs poorly in clinical practice (target domain). As the label of target-domain images is unavailable, this cross-domain segmentation needs unsupervised domain adaptation (UDA) to make the model adapt to the target domain. However, the shape and intensity distribution of neonatal brain MRI images across the domains are largely different from adults'. Current UDA methods aim to make synthesized images similar to the target domain as a whole. But it is impossible to synthesize images with intraclass similarity because of the regional misalignment caused by the cross-domain difference. This will result in generating intraclassly incorrect intensity information from target-domain images. To address this issue, we propose an IAS-NET (joint intraclassly adaptive generative adversarial network (GAN) (IA-NET) and segmentation) framework to bridge the gap between the two domains for intraclass alignment. METHODS: Our proposed IAS-NET is an elegant learning framework that transfers the appearance of images across the domains from both image and feature perspectives. It consists of the proposed IA-NET and a segmentation network (S-NET). The proposed IA-NET is a GAN-based adaptive network that contains one generator (including two encoders and one shared decoder) and four discriminators for cross-domain transfer. The two encoders are implemented to extract original image, mean, and variance features from source and target domains. The proposed local adaptive instance normalization algorithm is used to perform intraclass feature alignment to the target domain in the feature-map level. S-NET is a U-net structure network that is used to provide semantic constraint by a segmentation loss for the training of IA-NET. Meanwhile, it offers pseudo-label images for calculating intraclass features of the target domain. Source code (in Tensorflow) is available at https://github.com/lb-whu/RAS-NET/. RESULTS: Extensive experiments are carried out on two different data sets (NeoBrainS12 and dHCP), respectively. There exist great differences in the shape, size, and intensity distribution of magnetic resonance (MR) images in the two databases. Compared to baseline, we improve the average dice score of all tissues on NeoBrains12 by 6% through adaptive training with unlabeled dHCP images. Besides, we also conduct experiments on dHCP and improved the average dice score by 4%. The quantitative analysis of the mean and variance of the synthesized images shows that the synthesized image by the proposed is closer to the target domain both in the full brain or within each class than that of the compared methods. CONCLUSIONS: In this paper, the proposed IAS-NET can improve the performance of the S-NET effectively by its intraclass feature alignment in the target domain. Compared to the current UDA methods, the synthesized images by IAS-NET are more intraclassly similar to the target domain for neonatal brain MR images. Therefore, it achieves state-of-the-art results in the compared UDA models for the segmentation task.

A machine-learning-based method to predict adverse events in patients with dilated cardiomyopathy and severely reduced ejection fractions.

OBJECTIVE: Patients with dilated cardiomyopathy (DCM) and severely reduced left ventricular ejection fractions (LVEFs) are at very high risks of experiencing adverse cardiac events. A machine learning (ML) method could enable more effective risk stratification for these high-risk patients by incorporating various types of data. The aim of this study was to build an ML model to predict adverse events including all-cause deaths and heart transplantation in DCM patients with severely impaired LV systolic function. METHODS: One hundred and eighteen patients with DCM and severely reduced LVEFs (<35%) were included. The baseline clinical characteristics, laboratory data, electrocardiographic, and cardiac magnetic resonance (CMR) features were collected. Various feature selection processes and classifiers were performed to select an ML model with the best performance. The predictive performance of tested ML models was evaluated using the area under the curve (AUC) of the receiver operating characteristic curve using 10-fold cross-validation. RESULTS: Twelve patients died, and 17 patients underwent heart transplantation during the median follow-up of 508 days. The ML model included systolic blood pressure, left ventricular end-systolic and end-diastolic volume indices, and late gadolinium enhancement (LGE) extents on CMR imaging, and a support vector machine was selected as a classifier. The model showed excellent performance in predicting adverse events in DCM patients with severely reduced LVEF (the AUC and accuracy values were 0.873 and 0.763, respectively). CONCLUSIONS: This ML technique could effectively predict adverse events in DCM patients with severely reduced LVEF. ADVANCES IN KNOWLEDGE: The ML method has superior ability in risk stratification in severe DCM patients.

Prognostic Value of Late Enhanced Cardiac Magnetic Resonance Imaging Derived Texture Features in Dilated Cardiomyopathy Patients With Severely Reduced Ejection Fractions.

Background: Late enhanced cardiac magnetic resonance (CMR) images of the left ventricular myocardium contain an enormous amount of information that could provide prognostic value beyond that of late gadolinium enhancements (LGEs). With computational postprocessing and analysis, the heterogeneities and variations of myocardial signal intensities can be interpreted and measured as texture features. This study aimed to evaluate the value of texture features extracted from late enhanced CMR images of the myocardium to predict adverse outcomes in patients with dilated cardiomyopathy (DCM) and severe systolic dysfunction. Methods: This single-center study retrospectively enrolled patients with DCM with severely reduced left ventricular ejection fractions (LVEFs < 35%). Texture features were extracted from enhanced late scanning images, and the presence and extent of LGEs were also measured. Patients were followed-up for clinical endpoints composed of all-cause deaths and cardiac transplantation. Cox proportional hazard regression and Kaplan-Meier analyses were used to evaluate the prognostic value of texture features and conventional CMR parameters with event-free survival. Results: A total of 114 patients (37 women, median age 47.5 years old) with severely impaired systolic function (median LVEF, 14.0%) were followed-up for a median of 504.5 days. Twenty-nine patients experienced endpoint events, 12 died, and 17 underwent cardiac transplantations. Three texture features from a gray-level co-occurrence matrix (GLCM) (GLCM_contrast, GLCM_difference average, and GLCM_difference entropy) showed good prognostic value for adverse events when analyzed using univariable Cox hazard ratio regression (p = 0.007, p = 0.011, and p = 0.007, retrospectively). When each of the three features was analyzed using a multivariable Cox regression model that included the clinical parameter (systolic blood pressure) and LGE extent, they were found to be independently associated with adverse outcomes. Conclusion: Texture features related LGE heterogeneities and variations (GLCM_contrast, GLCM_difference average, and GLCM_difference entropy) are novel markers for risk stratification toward adverse events in DCM patients with severe systolic dysfunction.

Modal-Regression-Based Structured Low-Rank Matrix Recovery for Multiview Learning.

Low-rank Multiview Subspace Learning (LMvSL) has shown great potential in cross-view classification in recent years. Despite their empirical success, existing LMvSL-based methods are incapable of handling well view discrepancy and discriminancy simultaneously, which, thus, leads to performance degradation when there is a large discrepancy among multiview data. To circumvent this drawback, motivated by the block-diagonal representation learning, we propose structured low-rank matrix recovery (SLMR), a unique method of effectively removing view discrepancy and improving discriminancy through the recovery of the structured low-rank matrix. Furthermore, recent low-rank modeling provides a satisfactory solution to address the data contaminated by the predefined assumptions of noise distribution, such as Gaussian or Laplacian distribution. However, these models are not practical, since complicated noise in practice may violate those assumptions and the distribution is generally unknown in advance. To alleviate such a limitation, modal regression is elegantly incorporated into the framework of SLMR (termed MR-SLMR). Different from previous LMvSL-based methods, our MR-SLMR can handle any zero-mode noise variable that contains a wide range of noise, such as Gaussian noise, random noise, and outliers. The alternating direction method of multipliers (ADMM) framework and half-quadratic theory are used to optimize efficiently MR-SLMR. Experimental results on four public databases demonstrate the superiority of MR-SLMR and its robustness to complicated noise.

Neuroanatomical underpinning of diffusion kurtosis measurements in the cerebral cortex of healthy macaque brains.

PURPOSE: To investigate the neuroanatomical underpinning of healthy macaque brain cortical microstructure measured by diffusion kurtosis imaging (DKI), which characterizes non-Gaussian water diffusion. METHODS: High-resolution DKI was acquired from 6 postmortem macaque brains. Neurofilament density (ND) was quantified based on structure tensor from neurofilament histological images of a different macaque brain sample. After alignment of DKI-derived mean kurtosis (MK) maps to the histological images, MK and histology-based ND were measured at corresponding regions of interests characterized by distinguished cortical MK values in the prefrontal/precentral-postcentral and temporal cortices. Pearson correlation was performed to test significant correlation between these cortical MK and ND measurements. RESULTS: Heterogeneity of cortical MK across different cortical regions was revealed, with significantly and consistently higher MK measurements in the prefrontal/precentral-postcentral cortex compared to those in the temporal cortex across all six scanned macaque brains. Corresponding higher ND measurements in the prefrontal/precentral-postcentral cortex than in the temporal cortex were also found. The heterogeneity of cortical MK is associated with heterogeneity of histology-based ND measurements, with significant correlation between cortical MK and corresponding ND measurements (P < .005). CONCLUSION: These findings suggested that DKI-derived MK can potentially be an effective noninvasive biomarker quantifying underlying neuroanatomical complexity inside the cerebral cortical mantle for clinical and neuroscientific research.

Diffusion-MRI-based regional cortical microstructure at birth for predicting neurodevelopmental outcomes of 2-year-olds.

Cerebral cortical architecture at birth encodes regionally differential dendritic arborization and synaptic formation. It underlies behavioral emergence of 2-year-olds. Brain changes in 0-2 years are most dynamic across the lifespan. Effective prediction of future behavior with brain microstructure at birth will reveal structural basis of behavioral emergence in typical development and identify biomarkers for early detection and tailored intervention in atypical development. Here we aimed to evaluate the neonate whole-brain cortical microstructure quantified by diffusion MRI for predicting future behavior. We found that individual cognitive and language functions assessed at the age of 2 years were robustly predicted by neonate cortical microstructure using support vector regression. Remarkably, cortical regions contributing heavily to the prediction models exhibited distinctive functional selectivity for cognition and language. These findings highlight regional cortical microstructure at birth as a potential sensitive biomarker in predicting future neurodevelopmental outcomes and identifying individual risks of brain disorders.

Regularized-Ncut: Robust and homogeneous functional parcellation of neonate and adult brain networks.

Brain network parcellation based on resting-state functional MRI (rs-fMRI) is affected by noise, resulting in spurious small patches and decreased functional homogeneity within each network. Obtaining robust and homogeneous parcellation of neonate brain is more difficult, because neonate rs-fMRI is associated with relatively higher level of noise and no prior knowledge from a functional neonate atlas is available as spatial constraints. To meet these challenges, we developed a novel data-driven Regularized Normalized-cut (RNcut) method. RNcut is formulated by adding two regularization terms, a smoothing term using Markov random fields and a small-patch removal term, to conventional normalized-cut (Ncut) method. The RNcut and competing methods were tested with simulated datasets with known ground truth and then applied to both adult and neonate rs-fMRI datasets. Based on the parcellated networks generated by RNcut, intra-network connectivity was quantified. The test results from simulated datasets demonstrated that the RNcut method is more robust (p < 0.01) to noise and can delineate parcellated functional networks with significantly better (p < 0.01) spatial contiguity and significantly higher (p < 0.01) functional homogeneity than competing methods. Application of RNcut to neonate and adult rs-fMRI dataset revealed distinctive functional brain organization of neonate brains from that of adult brains. Collectively, we developed a novel data-driven RNcut method by integrating conventional Ncut with two regularization terms, generating robust and homogeneous functional parcellation without imposing spatial constraints. A broad range of brain network applications and analyses, especially neonate and infant brain parcellation with noisy and large sample of datasets, can potentially benefit from this RNcut method.

Multi-instance Deep Learning of Ultrasound Imaging Data for Pattern Classification of Congenital Abnormalities of the Kidney and Urinary Tract in Children.

OBJECTIVE: To reliably and quickly diagnose children with posterior urethral valves (PUV), we developed a multi-instance deep learning method to automate image analysis. METHODS: We built a robust pattern classifier to distinguish 86 children with PUV from 71 children with mild unilateral hydronephrosis based on ultrasound images (3504 in sagittal view and 2558 in transverse view) obtained during routine clinical care. RESULTS: The multi-instance deep learning classifier performed better than classifiers built on either single sagittal images or single transverse images. Particularly, the deep learning classifiers built on single images in the sagittal view and single images in the transverse view obtained area under the receiver operating characteristic curve (AUC) values of 0.796 ± 0.064 and 0.815 ± 0.071, respectively. AUC values of the multi-instance deep learning classifiers built on images in the sagittal and transverse views with mean pooling operation were 0.949 ± 0.035 and 0.954 ± 0.033, respectively. The multi-instance deep learning classifiers built on images in both the sagittal and transverse views with a mean pooling operation obtained an AUC of 0.961 ± 0.026 with a classification rate of 0.925 ± 0.060, specificity of 0.986 ± 0.032, and sensitivity of 0.873 ± 0.120, respectively. Discriminative regions of the kidney located using classification activation mapping demonstrated that the deep learning techniques could identify meaningful anatomical features from ultrasound images. CONCLUSION: The multi-instance deep learning method provides an automatic and accurate means to extract informative features from ultrasound images and discriminate infants with PUV from male children with unilateral hydronephrosis.

Differential White Matter Maturation from Birth to 8 Years of Age.

Comprehensive delineation of white matter (WM) microstructural maturation from birth to childhood is critical for understanding spatiotemporally differential circuit formation. Without a relatively large sample of datasets and coverage of critical developmental periods of both infancy and early childhood, differential maturational charts across WM tracts cannot be delineated. With diffusion tensor imaging (DTI) of 118 typically developing (TD) children aged 0-8 years and 31 children with autistic spectrum disorder (ASD) aged 2-7 years, the microstructure of every major WM tract and tract group was measured with DTI metrics to delineate differential WM maturation. The exponential model of microstructural maturation of all WM was identified. The WM developmental curves were separated into fast, intermediate, and slow phases in 0-8 years with distinctive time period of each phase across the tracts. Shorter periods of the fast and intermediate phases in certain tracts, such as the commissural tracts, indicated faster earlier development. With TD WM maturational curves as the reference, higher residual variance of WM microstructure was found in children with ASD. The presented comprehensive and differential charts of TD WM microstructural maturation of all major tracts and tract groups in 0-8 years provide reference standards for biomarker detection of neuropsychiatric disorders.

Automatic kidney segmentation in ultrasound images using subsequent boundary distance regression and pixelwise classification networks.

It remains challenging to automatically segment kidneys in clinical ultrasound (US) images due to the kidneys' varied shapes and image intensity distributions, although semi-automatic methods have achieved promising performance. In this study, we propose subsequent boundary distance regression and pixel classification networks to segment the kidneys automatically. Particularly, we first use deep neural networks pre-trained for classification of natural images to extract high-level image features from US images. These features are used as input to learn kidney boundary distance maps using a boundary distance regression network and the predicted boundary distance maps are classified as kidney pixels or non-kidney pixels using a pixelwise classification network in an end-to-end learning fashion. We also adopted a data-augmentation method based on kidney shape registration to generate enriched training data from a small number of US images with manually segmented kidney labels. Experimental results have demonstrated that our method could automatically segment the kidney with promising performance, significantly better than deep learning-based pixel classification networks.