Search

The decoder-only Transformer architecture with causal masking and relative position encoding (RPE) has become the de facto choice in language modeling. Despite its exceptional performance across various tasks, we have identified two limitations: First, it requires all attention scores to be non-zero and sum up to 1, even if the current embedding has sufficient self-contained information. This compels the model to assign disproportional excessive attention to specific tokens. Second, RPE-based Transformers are not universal approximators due to their limited capacity at encoding absolute positional information, which limits their application in position-critical tasks. In this work, we propose StableMask: a parameter-free method to address both limitations by refining the causal mask. It introduces pseudo-attention values to balance attention distributions and encodes absolute positional information via a progressively decreasing mask ratio. StableMask's effectiveness is validated both theoretically and empirically, showing significant enhancements in language models with parameter sizes ranging from 71M to 1.4B across diverse datasets and encoding methods. We further show that it naturally supports (1) efficient extrapolation without special tricks such as StreamingLLM and (2) easy integration with existing attention optimization techniques., Comment: Preprint

The core challenge of de novo protein design lies in creating proteins with specific functions or properties, guided by certain conditions. Current models explore to generate protein using structural and evolutionary guidance, which only provide indirect conditions concerning functions and properties. However, textual annotations of proteins, especially the annotations for protein domains, which directly describe the protein's high-level functionalities, properties, and their correlation with target amino acid sequences, remain unexplored in the context of protein design tasks. In this paper, we propose Protein-Annotation Alignment Generation (PAAG), a multi-modality protein design framework that integrates the textual annotations extracted from protein database for controllable generation in sequence space. Specifically, within a multi-level alignment module, PAAG can explicitly generate proteins containing specific domains conditioned on the corresponding domain annotations, and can even design novel proteins with flexible combinations of different kinds of annotations. Our experimental results underscore the superiority of the aligned protein representations from PAAG over 7 prediction tasks. Furthermore, PAAG demonstrates a nearly sixfold increase in generation success rate (24.7% vs 4.7% in zinc finger, and 54.3% vs 8.7% in the immunoglobulin domain) in comparison to the existing model.

Molecule-and-text cross-modal representation learning has emerged as a promising direction for enhancing the quality of molecular representation, thereby improving performance in various scientific fields, including drug discovery and materials science. Existing studies adopt a global alignment approach to learn the knowledge from different modalities. These global alignment approaches fail to capture fine-grained information, such as molecular fragments and their corresponding textual description, which is crucial for downstream tasks. Furthermore, it is incapable to model such information using a similar global alignment strategy due to data scarcity of paired local part annotated data from existing datasets. In this paper, we propose Atomas, a multi-modal molecular representation learning framework to jointly learn representations from SMILES string and text. We design a Hierarchical Adaptive Alignment model to concurrently learn the fine-grained fragment correspondence between two modalities and align these representations of fragments in three levels. Additionally, Atomas's end-to-end training framework incorporates the tasks of understanding and generating molecule, thereby supporting a wider range of downstream tasks. In the retrieval task, Atomas exhibits robust generalization ability and outperforms the baseline by 30.8% of recall@1 on average. In the generation task, Atomas achieves state-of-the-art results in both molecule captioning task and molecule generation task. Moreover, the visualization of the Hierarchical Adaptive Alignment model further confirms the chemical significance of our approach. Our codes can be found at https://anonymous.4open.science/r/Atomas-03C3.

Elucidating the cellular organization of the cerebral cortex is critical for understanding brain structure and function. Using large-scale single-nucleus RNA sequencing and spatial transcriptomic analysis of 143 macaque cortical regions, we obtained a comprehensive atlas of 264 transcriptome-defined cortical cell types and mapped their spatial distribution across the entire cortex. We characterized the cortical layer and region preferences of glutamatergic, GABAergic, and non-neuronal cell types, as well as regional differences in cell-type composition and neighborhood complexity. Notably, we discovered a relationship between the regional distribution of various cell types and the region’s hierarchical level in the visual and somatosensory systems. Cross-species comparison of transcriptomic data from human, macaque, and mouse cortices further revealed primate-specific cell types that are enriched in layer 4, with their marker genes expressed in a region-dependent manner. Our data provide a cellular and molecular basis for understanding the evolution, development, aging, and pathogenesis of the primate brain., Elucidating the cellular organization of the cerebral cortex is critical for understanding brain structure and function. Using large-scale single-nucleus RNA sequencing and spatial transcriptomic analysis of 143 macaque cortical regions, we obtained a comprehensive atlas of 264 transcriptome-defined cortical cell types and mapped their spatial distribution across the entire cortex. We characterized the cortical layer and region preferences of glutamatergic, GABAergic, and non-neuronal cell types, as well as regional differences in cell-type composition and neighborhood complexity. Notably, we discovered a relationship between the regional distribution of various cell types and the region's hierarchical level in the visual and somatosensory systems. Cross-species comparison of transcriptomic data from human, macaque, and mouse cortices further revealed primate-specific cell types that are enriched in layer 4, with their marker genes expressed in a region-dependent manner. Our data provide a cellular and molecular basis for understanding the evolution, development, aging, and pathogenesis of the primate brain.

Subpopulation shift exists widely in many real-world applications, which refers to the training and test distributions that contain the same subpopulation groups but with different subpopulation proportions. Ignoring subpopulation shifts may lead to significant performance degradation and fairness concerns. Importance reweighting is a classical and effective way to handle the subpopulation shift. However, recent studies have recognized that most of these approaches fail to improve the performance especially when applied to over-parameterized neural networks which are capable of fitting any training samples. In this work, we propose a simple yet practical framework, called reweighted mixup (RMIX), to mitigate the overfitting issue in over-parameterized models by conducting importance weighting on the ''mixed'' samples. Benefiting from leveraging reweighting in mixup, RMIX allows the model to explore the vicinal space of minority samples more, thereby obtaining more robust model against subpopulation shift. When the subpopulation memberships are unknown, the training-trajectories-based uncertainty estimation is equipped in the proposed RMIX to flexibly characterize the subpopulation distribution. We also provide insightful theoretical analysis to verify that RMIX achieves better generalization bounds over prior works. Further, we conduct extensive empirical studies across a wide range of tasks to validate the effectiveness of the proposed method., Comment: Journal version of arXiv:2209.08928

Elucidating the cellular organization of the cerebral cortex is critical for understanding brain structure and function. Using large-scale single-nucleus RNA sequencing and spatial transcriptomic analysis of 143 macaque cortical regions, we obtained a comprehensive atlas of 264 transcriptome-defined cortical cell types and mapped their spatial distribution across the entire cortex. We characterized the cortical layer and region preferences of glutamatergic, GABAergic, and non-neuronal cell types, as well as regional differences in cell-type composition and neighborhood complexity. Notably, we discovered a relationship between the regional distribution of various cell types and the region’s hierarchical level in the visual and somatosensory systems. Cross-species comparison of transcriptomic data from human, macaque, and mouse cortices further revealed primate-specific cell types that are enriched in layer 4, with their marker genes expressed in a region-dependent manner. Our data provide a cellular and molecular basis for understanding the evolution, development, aging, and pathogenesis of the primate brain., Elucidating the cellular organization of the cerebral cortex is critical for understanding brain structure and function. Using large-scale single-nucleus RNA sequencing and spatial transcriptomic analysis of 143 macaque cortical regions, we obtained a comprehensive atlas of 264 transcriptome-defined cortical cell types and mapped their spatial distribution across the entire cortex. We characterized the cortical layer and region preferences of glutamatergic, GABAergic, and non-neuronal cell types, as well as regional differences in cell-type composition and neighborhood complexity. Notably, we discovered a relationship between the regional distribution of various cell types and the region's hierarchical level in the visual and somatosensory systems. Cross-species comparison of transcriptomic data from human, macaque, and mouse cortices further revealed primate-specific cell types that are enriched in layer 4, with their marker genes expressed in a region-dependent manner. Our data provide a cellular and molecular basis for understanding the evolution, development, aging, and pathogenesis of the primate brain.

Semi-supervised semantic segmentation aims to utilize limited labeled images and abundant unlabeled images to achieve label-efficient learning, wherein the weak-to-strong consistency regularization framework, popularized by FixMatch, is widely used as a benchmark scheme. Despite its effectiveness, we observe that such scheme struggles with satisfactory segmentation for the local regions. This can be because it originally stems from the image classification task and lacks specialized mechanisms to capture fine-grained local semantics that prioritizes in dense prediction. To address this issue, we propose a novel framework called \texttt{MaskMatch}, which enables fine-grained locality learning to achieve better dense segmentation. On top of the original teacher-student framework, we design a masked modeling proxy task that encourages the student model to predict the segmentation given the unmasked image patches (even with 30\% only) and enforces the predictions to be consistent with pseudo-labels generated by the teacher model using the complete image. Such design is motivated by the intuition that if the predictions are more consistent given insufficient neighboring information, stronger fine-grained locality perception is achieved. Besides, recognizing the importance of reliable pseudo-labels in the above locality learning and the original consistency learning scheme, we design a multi-scale ensembling strategy that considers context at different levels of abstraction for pseudo-label generation. Extensive experiments on benchmark datasets demonstrate the superiority of our method against previous approaches and its plug-and-play flexibility.

Immune repertoire classification, a typical multiple instance learning (MIL) problem, is a frontier research topic in computational biology that makes transformative contributions to new vaccines and immune therapies. However, the traditional instance-space MIL, directly assigning bag-level labels to instances, suffers from the massive amount of noisy labels and extremely low witness rate. In this work, we propose a noisy-label-learning formulation to solve the immune repertoire classification task. To remedy the inaccurate supervision of repertoire-level labels for a sequence-level classifier, we design a robust training strategy: The initial labels are smoothed to be asymmetric and are progressively corrected using the model's predictions throughout the training process. Furthermore, two models with the same architecture but different parameter initialization are co-trained simultaneously to remedy the known "confirmation bias" problem in the self-training-like schema. As a result, we obtain accurate sequence-level classification and, subsequently, repertoire-level classification. Experiments on the Cytomegalovirus (CMV) and Cancer datasets demonstrate our method's effectiveness and superior performance on sequence-level and repertoire-level tasks., Comment: IJCAI 2023

Pre-trained large language models demonstrate potential in extracting information from DNA sequences, yet adapting to a variety of tasks and data modalities remains a challenge. To address this, we propose DNAGPT, a generalized DNA pre-training model trained on over 200 billion base pairs from all mammals. By enhancing the classic GPT model with a binary classification task (DNA sequence order), a numerical regression task (guanine-cytosine content prediction), and a comprehensive token language, DNAGPT can handle versatile DNA analysis tasks while processing both sequence and numerical data. Our evaluation of genomic signal and region recognition, mRNA abundance regression, and artificial genomes generation tasks demonstrates DNAGPT's superior performance compared to existing models designed for specific downstream tasks, benefiting from pre-training using the newly designed model structure.

Subpopulation shift widely exists in many real-world machine learning applications, referring to the training and test distributions containing the same subpopulation groups but varying in subpopulation frequencies. Importance reweighting is a normal way to handle the subpopulation shift issue by imposing constant or adaptive sampling weights on each sample in the training dataset. However, some recent studies have recognized that most of these approaches fail to improve the performance over empirical risk minimization especially when applied to over-parameterized neural networks. In this work, we propose a simple yet practical framework, called uncertainty-aware mixup (UMIX), to mitigate the overfitting issue in over-parameterized models by reweighting the “mixed” samples according to the sample uncertainty. The training-trajectories-based uncertainty estimation is equipped in the proposed UMIX for each sample to flexibly characterize the subpopulation distribution. We also provide insightful theoretical analysis to verify that UMIX achieves better generalization bounds over prior works. Further, we conduct extensive empirical studies across a wide range of tasks to validate the effectiveness of our method both qualitatively and quantitatively. Code is available at this URL.

Recently, deep neural networks have greatly advanced histopathology image segmentation but usually require abundant annotated data. However, due to the gigapixel scale of whole slide images and pathologists' heavy daily workload, obtaining pixel-level labels for supervised learning in clinical practice is often infeasible. Alternatively, weakly-supervised segmentation methods have been explored with less laborious image-level labels, but their performance is unsatisfactory due to the lack of dense supervision. Inspired by the recent success of self-supervised learning methods, we present a label-efficient tissue prototype dictionary building pipeline and propose to use the obtained prototypes to guide histopathology image segmentation. Particularly, taking advantage of self-supervised contrastive learning, an encoder is trained to project the unlabeled histopathology image patches into a discriminative embedding space where these patches are clustered to identify the tissue prototypes by efficient pathologists' visual examination. Then, the encoder is used to map the images into the embedding space and generate pixel-level pseudo tissue masks by querying the tissue prototype dictionary. Finally, the pseudo masks are used to train a segmentation network with dense supervision for better performance. Experiments on two public datasets demonstrate that our human-machine interactive tissue prototype learning method can achieve comparable segmentation performance as the fully-supervised baselines with less annotation burden and outperform other weakly-supervised methods. Codes will be available upon publication., Comment: IPMI2023 camera ready

Learning with Noisy Labels (LNL) has attracted significant attention from the research community. Many recent LNL methods rely on the assumption that clean samples tend to have "small loss". However, this assumption always fails to generalize to some real-world cases with imbalanced subpopulations, i.e., training subpopulations varying in sample size or recognition difficulty. Therefore, recent LNL methods face the risk of misclassifying those "informative" samples (e.g., hard samples or samples in the tail subpopulations) into noisy samples, leading to poor generalization performance. To address the above issue, we propose a novel LNL method to simultaneously deal with noisy labels and imbalanced subpopulations. It first leverages sample correlation to estimate samples' clean probabilities for label correction and then utilizes corrected labels for Distributionally Robust Optimization (DRO) to further improve the robustness. Specifically, in contrast to previous works using classification loss as the selection criterion, we introduce a feature-based metric that takes the sample correlation into account for estimating samples' clean probabilities. Then, we refurbish the noisy labels using the estimated clean probabilities and the pseudo-labels from the model's predictions. With refurbished labels, we use DRO to train the model to be robust to subpopulation imbalance. Extensive experiments on a wide range of benchmarks demonstrate that our technique can consistently improve current state-of-the-art robust learning paradigms against noisy labels, especially when encountering imbalanced subpopulations.

Subpopulation shift widely exists in many real-world machine learning applications, referring to the training and test distributions containing the same subpopulation groups but varying in subpopulation frequencies. Importance reweighting is a normal way to handle the subpopulation shift issue by imposing constant or adaptive sampling weights on each sample in the training dataset. However, some recent studies have recognized that most of these approaches fail to improve the performance over empirical risk minimization especially when applied to over-parameterized neural networks. In this work, we propose a simple yet practical framework, called uncertainty-aware mixup (UMIX), to mitigate the overfitting issue in over-parameterized models by reweighting the ''mixed'' samples according to the sample uncertainty. The training-trajectories-based uncertainty estimation is equipped in the proposed UMIX for each sample to flexibly characterize the subpopulation distribution. We also provide insightful theoretical analysis to verify that UMIX achieves better generalization bounds over prior works. Further, we conduct extensive empirical studies across a wide range of tasks to validate the effectiveness of our method both qualitatively and quantitatively. Code is available at https://github.com/TencentAILabHealthcare/UMIX., Comment: NeurIPS 2022

Multiple instance learning (MIL) is a powerful approach to classify whole slide images (WSIs) for diagnostic pathology. A fundamental challenge of MIL on WSI classification is to discover the \textit{critical instances} that trigger the bag label. However, previous methods are primarily designed under the independent and identical distribution hypothesis (\textit{i.i.d}), ignoring either the correlations between instances or heterogeneity of tumours. In this paper, we propose a novel multiplex-detection-based multiple instance learning (MDMIL) to tackle the issues above. Specifically, MDMIL is constructed by the internal query generation module (IQGM) and the multiplex detection module (MDM) and assisted by the memory-based contrastive loss during training. Firstly, IQGM gives the probability of instances and generates the internal query (IQ) for the subsequent MDM by aggregating highly reliable features after the distribution analysis. Secondly, the multiplex-detection cross-attention (MDCA) and multi-head self-attention (MHSA) in MDM cooperate to generate the final representations for the WSI. In this process, the IQ and trainable variational query (VQ) successfully build up the connections between instances and significantly improve the model's robustness toward heterogeneous tumours. At last, to further enforce constraints in the feature space and stabilize the training process, we adopt a memory-based contrastive loss, which is practicable for WSI classification even with a single sample as input in each iteration. We conduct experiments on three computational pathology datasets, e.g., CAMELYON16, TCGA-NSCLC, and TCGA-RCC datasets. The superior accuracy and AUC demonstrate the superiority of our proposed MDMIL over other state-of-the-art methods.

Detectingandsegmentingobjectswithinwholeslideimagesis essential in computational pathology workflow. Self-supervised learning (SSL) is appealing to such annotation-heavy tasks. Despite the extensive benchmarks in natural images for dense tasks, such studies are, unfortunately, absent in current works for pathology. Our paper intends to narrow this gap. We first benchmark representative SSL methods for dense prediction tasks in pathology images. Then, we propose concept contrastive learning (ConCL), an SSL framework for dense pre-training. We explore how ConCL performs with concepts provided by different sources and end up with proposing a simple dependency-free concept generating method that does not rely on external segmentation algorithms or saliency detection models. Extensive experiments demonstrate the superiority of ConCL over previous state-of-the-art SSL methods across different settings. Along our exploration, we distll several important and intriguing components contributing to the success of dense pre-training for pathology images. We hope this work could provide useful data points and encourage the community to conduct ConCL pre-training for problems of interest. Code is available., Comment: Accepted as an ECCV 2022 paper. Code is available at https://github.com/Jiawei-Yang/ConCL or https://github.com/TencentAILabHealthcare/ConCL

Whole slide image (WSI) classification often relies on deep weakly supervised multiple instance learning (MIL) methods to handle gigapixel resolution images and slide-level labels. Yet the decent performance of deep learning comes from harnessing massive datasets and diverse samples, urging the need for efficient training pipelines for scaling to large datasets and data augmentation techniques for diversifying samples. However, current MIL-based WSI classification pipelines are memory-expensive and computation-inefficient since they usually assemble tens of thousands of patches as bags for computation. On the other hand, despite their popularity in other tasks, data augmentations are unexplored for WSI MIL frameworks. To address them, we propose ReMix, a general and efficient framework for MIL based WSI classification. It comprises two steps: reduce and mix. First, it reduces the number of instances in WSI bags by substituting instances with instance prototypes, i.e., patch cluster centroids. Then, we propose a ``Mix-the-bag'' augmentation that contains four online, stochastic and flexible latent space augmentations. It brings diverse and reliable class-identity-preserving semantic changes in the latent space while enforcing semantic-perturbation invariance. We evaluate ReMix on two public datasets with two state-of-the-art MIL methods. In our experiments, consistent improvements in precision, accuracy, and recall have been achieved but with orders of magnitude reduced training time and memory consumption, demonstrating ReMix's effectiveness and efficiency. Code is available., Comment: Published in MICCAI 2022. Code: https://github.com/Jiawei-Yang/ReMix or https://github.com/TencentAILabHealthcare/ReMix

Tungsten carbide based cemented carbides, often simply termed hardmetals, are established forefront materials for tools, structural components, and wear parts with stringent requirements. Several of the technological applications in which they are used include exposure to chemically aggressive media. Under these conditions, failure induced under applied load may be accelerated; and consequently, the service life may be decreased. Within this context, this work addresses the influence of corrosion-induced damage on the mechanical integrity and load-bearing capability of hardmetals at different length scales, i.e., from 100s nanometers to 1000s microns. Experimental data acquired by means of nanoindentation, pyramidal, and spherical indentation, as well as sliding contact (micro- and nanoscratch) techniques, are presented. The attained results allow for identifying guidelines for the microstructural design of these materials under combined consideration of corrosion and mechanical contact as service-like conditions. Discussion of the reported findings includes a critical analysis of corrosion effects on the evolution of microstructure-property-performance interrelations for the materials under consideration., Peer Reviewed, Postprint (published version)

Dense prediction in medical volume provides enriched guidance for clinical analysis. CNN backbones have met bottleneck due to lack of long-range dependencies and global context modeling power. Recent works proposed to combine vision transformer with CNN, due to its strong global capture ability and learning capability. However, most works are limited to simply applying pure transformer with several fatal flaws (i.e., lack of inductive bias, heavy computation and little consideration for 3D data). Therefore, designing an elegant and efficient vision transformer learner for dense prediction in medical volume is promising and challenging. In this paper, we propose a novel 3D Shuffle-Mixer network of a new Local Vision Transformer-MLP paradigm for medical dense prediction. In our network, a local vision transformer block is utilized to shuffle and learn spatial context from full-view slices of rearranged volume, a residual axial-MLP is designed to mix and capture remaining volume context in a slice-aware manner, and a MLP view aggregator is employed to project the learned full-view rich context to the volume feature in a view-aware manner. Moreover, an Adaptive Scaled Enhanced Shortcut is proposed for local vision transformer to enhance feature along spatial and channel dimensions adaptively, and a CrossMerge is proposed to skip-connects the multi-scale feature appropriately in the pyramid architecture. Extensive experiments demonstrate the proposed model outperforms other state-of-the-art medical dense prediction methods.

Few-shot learning is an established topic in natural images for years, but few work is attended to histology images, which is of high clinical value since well-labeled datasets and rare abnormal samples are expensive to collect. Here, we facilitate the study of few-shot learning in histology images by setting up three cross-domain tasks that simulate real clinics problems. To enable label-efficient learning and better generalizability, we propose to incorporate contrastive learning (CL) with latent augmentation (LA) to build a few-shot system. CL learns useful representations without manual labels, while LA transfers semantic variations of the base dataset in an unsupervised way. These two components fully exploit unlabeled training data and can scale gracefully to other label-hungry problems. In experiments, we find i) models learned by CL generalize better than supervised learning for histology images in unseen classes, and ii) LA brings consistent gains over baselines. Prior studies of self-supervised learning mainly focus on ImageNet-like images, which only present a dominant object in their centers. Recent attention has been paid to images with multi-objects and multi-textures. Histology images are a natural choice for such a study. We show the superiority of CL over supervised learning in terms of generalization for such data and provide our empirical understanding for this observation. The findings in this work could contribute to understanding how the model generalizes in the context of both representation learning and histological image analysis. Code is available.

Recently, deep neural networks have greatly advanced undersampled Magnetic Resonance Image (MRI) reconstruction, wherein most studies follow the one-anatomy-one-network fashion, i.e., each expert network is trained and evaluated for a specific anatomy. Apart from inefficiency in training multiple independent models, such convention ignores the shared de-aliasing knowledge across various anatomies which can benefit each other. To explore the shared knowledge, one naive way is to combine all the data from various anatomies to train an all-round network. Unfortunately, despite the existence of the shared de-aliasing knowledge, we reveal that the exclusive knowledge across different anatomies can deteriorate specific reconstruction targets, yielding overall performance degradation. Observing this, in this study, we present a novel deep MRI reconstruction framework with both anatomy-shared and anatomy-specific parameterized learners, aiming to "seek common ground while reserving differences" across different anatomies.Particularly, the primary anatomy-shared learners are exposed to different anatomies to model flourishing shared knowledge, while the efficient anatomy-specific learners are trained with their target anatomy for exclusive knowledge. Four different implementations of anatomy-specific learners are presented and explored on the top of our framework in two MRI reconstruction networks. Comprehensive experiments on brain, knee and cardiac MRI datasets demonstrate that three of these learners are able to enhance reconstruction performance via multiple anatomy collaborative learning., Comment: submitted to an IEEE journal

Objective. To develop an artificial intelligence method predicting lymph node metastasis (LNM) for patients with colorectal cancer (CRC). Impact Statement. A novel interpretable multimodal AI-based method to predict LNM for CRC patients by integrating information of pathological images and serum tumor-specific biomarkers. Introduction. Preoperative diagnosis of LNM is essential in treatment planning for CRC patients. Existing radiology imaging and genomic tests approaches are either unreliable or too costly. Methods. A total of 1338 patients were recruited, where 1128 patients from one centre were included as the discovery cohort and 210 patients from other two centres were involved as the external validation cohort. We developed a Multimodal Multiple Instance Learning (MMIL) model to learn latent features from pathological images and then jointly integrated the clinical biomarker features for predicting LNM status. The heatmaps of the obtained MMIL model were generated for model interpretation. Results. The MMIL model outperformed preoperative radiology-imaging diagnosis and yielded high area under the curve (AUCs) of 0.926, 0.878, 0.809, and 0.857 for patients with stage T1, T2, T3, and T4 CRC, on the discovery cohort. On the external cohort, it obtained AUCs of 0.855, 0.832, 0.691, and 0.792, respectively (T1-T4), which indicates its prediction accuracy and potential adaptability among multiple centres. Conclusion. The MMIL model showed the potential in the early diagnosis of LNM by referring to pathological images and tumor-specific biomarkers, which is easily accessed in different institutes. We revealed the histomorphologic features determining the LNM prediction indicating the model ability to learn informative latent features.

Doctors often make diagonostic decisions based on patient's image scans, such as magnetic resonance imaging (MRI), and patient's electronic health records (EHR) such as age, gender, blood pressure and so on. Despite a lot of automatic methods have been proposed for either image or text analysis in computer vision or natural language research areas, much fewer studies have been developed for the fusion of medical image and EHR data for medical problems. Among existing early or intermediate fusion methods, concatenation of features from both modalities is still a mainstream. For a better exploiting of image and EHR data, we propose a multi-modal attention module which use EHR data to help the selection of important regions during image feature extraction process conducted by traditional CNN. Moreover, we propose to incorporate multi-head machnism to gated multimodal unit (GMU) to make it able to parallelly fuse image and EHR features in different subspaces. With the help of the two modules, existing CNN architecture can be enhanced using both modalities. Experiments on predicting Glasgow outcome scale (GOS) of intracerebral hemorrhage patients and classifying Alzheimer's Disease showed the proposed method can automatically focus on task-related areas and achieve better results by making better use of image and EHR features.

The outbreak of novel coronavirus disease 2019 (COVID-19) has already infected millions of people and is still rapidly spreading all over the globe. Most COVID-19 patients suffer from lung infection, so one important diagnostic method is to screen chest radiography images, e.g., X-Ray or CT images. However, such examinations are time-consuming and labor-intensive, leading to limited diagnostic efficiency. To solve this issue, AI-based technologies, such as deep learning, have been used recently as effective computer-aided means to improve diagnostic efficiency. However, one practical and critical difficulty is the limited availability of annotated COVID-19 data, due to the prohibitive annotation costs and urgent work of doctors to fight against the pandemic. This makes the learning of deep diagnosis models very challenging. To address this, motivated by that typical pneumonia has similar characteristics with COVID-19 and many pneumonia datasets are publicly available, we propose to conduct domain knowledge adaptation from typical pneumonia to COVID-19. There are two main challenges: 1) the discrepancy of data distributions between domains; 2) the task difference between the diagnosis of typical pneumonia and COVID-19. To address them, we propose a new deep domain adaptation method for COVID-19 diagnosis, namely COVID-DA. Specifically, we alleviate the domain discrepancy via feature adversarial adaptation and handle the task difference issue via a novel classifier separation scheme. In this way, COVID-DA is able to diagnose COVID-19 effectively with only a small number of COVID-19 annotations. Extensive experiments verify the effectiveness of COVID-DA and its great potential for real-world applications.

Artificial Intelligence (AI)-powered pathology is a revolutionary step in the world of digital pathology and shows great promise to increase both diagnosis accuracy and efficiency. However, defocus and motion blur can obscure tissue or cell characteristics hence compromising AI algorithms'accuracy and robustness in analyzing the images. In this paper, we demonstrate a deep-learning-based approach that can alleviate the defocus and motion blur of a microscopic image and output a sharper and cleaner image with retrieved fine details without prior knowledge of the blur type, blur extent and pathological stain. In this approach, a deep learning classifier is first trained to identify the image blur type. Then, two encoder-decoder networks are trained and used alone or in combination to deblur the input image. It is an end-to-end approach and introduces no corrugated artifacts as traditional blind deconvolution methods do. We test our approach on different types of pathology specimens and demonstrate great performance on image blur correction and the subsequent improvement on the diagnosis outcome of AI algorithms.

The overexpression of human epidermal growth factor receptor 2 (HER2) has been established as a therapeutic target in multiple types of cancers, such as breast and gastric cancers. Immunohistochemistry (IHC) is employed as a basic HER2 test to identify the HER2-positive, borderline, and HER2-negative patients. However, the reliability and accuracy of HER2 scoring are affected by many factors, such as pathologists' experience. Recently, artificial intelligence (AI) has been used in various disease diagnosis to improve diagnostic accuracy and reliability, but the interpretation of diagnosis results is still an open problem. In this paper, we propose a real-time HER2 scoring system, which follows the HER2 scoring guidelines to complete the diagnosis, and thus each step is explainable. Unlike the previous scoring systems based on whole-slide imaging, our HER2 scoring system is integrated into an augmented reality (AR) microscope that can feedback AI results to the pathologists while reading the slide. The pathologists can help select informative fields of view (FOVs), avoiding the confounding regions, such as DCIS. Importantly, we illustrate the intermediate results with membrane staining condition and cell classification results, making it possible to evaluate the reliability of the diagnostic results. Also, we support the interactive modification of selecting regions-of-interest, making our system more flexible in clinical practice. The collaboration of AI and pathologists can significantly improve the robustness of our system. We evaluate our system with 285 breast IHC HER2 slides, and the classification accuracy of 95\% shows the effectiveness of our HER2 scoring system., Comment: 11 pages, 5 figures, and 4 tables

Bone metastases are a frequent occurrence with cancer, and early detection can guide the patient's treatment regimen. Metastatic bone disease can present in density extremes as sclerotic (high density) and lytic (low density) or in a continuum with an admixture of both sclerotic and lytic components. We design a framework to detect and characterize the varying spectrum of presentation of spine metastasis on positron emission tomography/computed tomography (PET/CT) data. A technique is proposed to synthesize CT and PET images to enhance the lesion appearance for computer detection. A combination of watershed, graph cut, and level set algorithms is first run to obtain the initial detections. Detections are then sent to multiple classifiers for sclerotic, lytic, and mixed lesions. The system was tested on 44 cases with 225 sclerotic, 139 lytic, and 92 mixed lesions. The results showed that sensitivity (false positive per patient) was 0.81 (2.1), 0.81 (1.3), and 0.76 (2.1) for sclerotic, lytic, and mixed lesions, respectively. It also demonstrates that using PET/CT data significantly improves the computer aided detection performance over using CT alone.

Bone metastases are a frequent occurrence with cancer, and early detection can guide the patient's treatment regimen. Metastatic bone disease can present in density extremes as sclerotic (high density) and lytic (low density) or in a continuum with an admixture of both sclerotic and lytic components. We design a framework to detect and characterize the varying spectrum of presentation of spine metastasis on positron emission tomography/computed tomography (PET/CT) data. A technique is proposed to synthesize CT and PET images to enhance the lesion appearance for computer detection. A combination of watershed, graph cut, and level set algorithms is first run to obtain the initial detections. Detections are then sent to multiple classifiers for sclerotic, lytic, and mixed lesions. The system was tested on 44 cases with 225 sclerotic, 139 lytic, and 92 mixed lesions. The results showed that sensitivity (false positive per patient) was 0.81 (2.1), 0.81 (1.3), and 0.76 (2.1) for sclerotic, lytic, and mixed lesions, respectively. It also demonstrates that using PET/CT data significantly improves the computer aided detection performance over using CT alone.

ObjectiveTo investigate a fully automated abdominal CT-based muscle tool in a large adult screening population.MethodsA fully automated validated muscle segmentation algorithm was applied to 9310 non-contrast CT scans, including a primary screening cohort of 8037 consecutive asymptomatic adults (mean age, 57.1±7.8 years; 3555M/4482F). Sequential follow-up scans were available in a subset of 1171 individuals (mean interval, 5.1 years). Muscle tissue cross-sectional area and attenuation (Hounsfield unit, HU) at the L3 level were assessed, including change over time.ResultsMean values were significantly higher in males for both muscle area (190.6±33.6 vs 133.3±24.1 cm2, p<0.001) and density (34.3±11.1 HU vs 27.3±11.7 HU, p<0.001). Age-related losses were observed, with mean muscle area reduction of -1.5 cm2/year and attenuation reduction of -1.5 HU/year. Overall age-related muscle density (attenuation) loss was steeper than for muscle area for both sexes up to the age of 70 years. Between ages 50 and 70, relative muscle attenuation decreased significantly more in females (-30.6% vs -18.0%, p<0.001), whereas relative rates of muscle area loss were similar (-8%). Between ages 70 and 90, males lost more density (-22.4% vs -7.5%) and area (-13.4% vs -6.9%, p<0.001). Of the 1171 patients with longitudinal follow-up, 1013 (86.5%) showed a decrease in muscle attenuation, 739 (63.1%) showed a decrease in area, and 1119 (95.6%) showed a decrease in at least one of these measures.ConclusionThis fully automated CT muscle tool allows for both individualized and population-based assessment. Such data could be automatically derived at abdominal CT regardless of study indication, allowing for opportunistic sarcopenia detection.Advances in knowledgeThis fully automated tool can be applied to routine abdominal CT scans for prospective or retrospective opportunistic sarcopenia assessment, regardless of the original clinical indication. Mean values were significantly higher i

Prognostic tumor growth modeling via volumetric medical imaging observations can potentially lead to better outcomes of tumor treatment and surgical planning. Recent advances of convolutional networks have demonstrated higher accuracy than traditional mathematical models in predicting future tumor volumes. This indicates that deep learning-based techniques may have great potentials on addressing such problem. However, current 2D patch-based modeling approaches cannot make full use of the spatio-temporal imaging context of the tumor's longitudinal 4D (3D + time) data. Moreover, they are incapable to predict clinically-relevant tumor properties, other than volumes. In this paper, we exploit to formulate the tumor growth process through convolutional Long Short-Term Memory (ConvLSTM) that extract tumor's static imaging appearances and capture its temporal dynamic changes within a single network. We extend ConvLSTM into the spatio-temporal domain (ST-ConvLSTM) by jointly learning the inter-slice 3D contexts and the longitudinal or temporal dynamics from multiple patient studies. Our approach can incorporate other non-imaging patient information in an end-to-end trainable manner. Experiments are conducted on the largest 4D longitudinal tumor dataset of 33 patients to date. Results validate that the ST-ConvLSTM produces a Dice score of 83.2%+-5.1% and a RVD of 11.2%+-10.8%, both significantly outperforming (p<0.05) other compared methods of linear model, ConvLSTM, and generative adversarial network (GAN) under the metric of predicting future tumor volumes. Additionally, our new method enables the prediction of both cell density and CT intensity numbers. Last, we demonstrate the generalizability of ST-ConvLSTM by employing it in 4D medical image segmentation task, which achieves an averaged Dice score of 86.3+-1.2% for left-ventricle segmentation in 4D ultrasound with 3 seconds per patient.

A multiple center milestone study of clinical vertebra segmentation is presented in this paper. Vertebra segmentation is a fundamental step for spinal image analysis and intervention. The first half of the study was conducted in the spine segmentation challenge in 2014 International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI) Workshop on Computational Spine Imaging (CSI 2014). The objective was to evaluate the performance of several state-of-the-art vertebra segmentation algorithms on computed tomography (CT) scans using ten training and five testing dataset, all healthy cases; the second half of the study was conducted after the challenge, where additional 5 abnormal cases are used for testing to evaluate the performance under abnormal cases. Dice coefficients and absolute surface distances were used as evaluation metrics. Segmentation of each vertebra as a single geometric unit, as well as separate segmentation of vertebra substructures, was evaluated. Five teams participated in the comparative study. The top performers in the study achieved Dice coefficient of 0.93 in the upper thoracic, 0.95 in the lower thoracic and 0.96 in the lumbar spine for healthy cases, and 0.88 in the upper thoracic, 0.89 in the lower thoracic and 0.92 in the lumbar spine for osteoporotic and fractured cases. The strengths and weaknesses of each method as well as future suggestion for improvement are discussed. This is the first multi-center comparative study for vertebra segmentation methods, which will provide an up-to-date performance milestone for the fast growing spinal image analysis and intervention.

Fine-grained classification of cervical cells into different abnormality levels is of great clinical importance but remains very challenging. Contrary to traditional classification methods that rely on hand-crafted or engineered features, convolution neural network (CNN) can classify cervical cells based on automatically learned deep features. However, CNN in previous studies do not involve cell morphological information, and it is unknown whether morphological features can be directly modeled by CNN to classify cervical cells. This paper presents a CNN-based method that combines cell image appearance with cell morphology for classification of cervical cells in Pap smear. The training cervical cell dataset consists of adaptively re-sampled image patches coarsely centered on the nuclei. Several CNN models (AlexNet, GoogleNet, ResNet and DenseNet) pre-trained on ImageNet dataset were fine-tuned on the cervical dataset for comparison. The proposed method is evaluated on the Herlev cervical dataset by five-fold cross-validation at patient level splitting. Results show that by adding cytoplasm and nucleus masks as raw morphological information into appearance-based CNN learning, higher classification accuracies can be achieved in general. Among the four CNN models, GoogleNet fed with both morphological and appearance information obtains the highest classification accuracies of 94.5% for 2-class classification task and 64.5% for 7-class classification task. Our method demonstrates that combining cervical cell morphology with appearance information can provide improved classification performance, which is clinically important for early diagnosis of cervical dysplastic changes., Comment: 7 pages, 4 figures

Tumor growth is associated with cell invasion and mass-effect, which are traditionally formulated by mathematical models, namely reaction-diffusion equations and biomechanics. Such models can be personalized based on clinical measurements to build the predictive models for tumor growth. In this paper, we investigate the possibility of using deep convolutional neural networks (ConvNets) to directly represent and learn the cell invasion and mass-effect, and to predict the subsequent involvement regions of a tumor. The invasion network learns the cell invasion from information related to metabolic rate, cell density and tumor boundary derived from multimodal imaging data. The expansion network models the mass-effect from the growing motion of tumor mass. We also study different architectures that fuse the invasion and expansion networks, in order to exploit the inherent correlations among them. Our network can easily be trained on population data and personalized to a target patient, unlike most previous mathematical modeling methods that fail to incorporate population data. Quantitative experiments on a pancreatic tumor data set show that the proposed method substantially outperforms a state-of-the-art mathematical model-based approach in both accuracy and efficiency, and that the information captured by each of the two subnetworks are complementary.

Image segmentation is a fundamental problem in medical image analysis. In recent years, deep neural networks achieve impressive performances on many medical image segmentation tasks by supervised learning on large manually annotated data. However, expert annotations on big medical datasets are tedious, expensive or sometimes unavailable. Weakly supervised learning could reduce the effort for annotation but still required certain amounts of expertise. Recently, deep learning shows a potential to produce more accurate predictions than the original erroneous labels. Inspired by this, we introduce a very weakly supervised learning method, for cystic lesion detection and segmentation in lung CT images, without any manual annotation. Our method works in a self-learning manner, where segmentation generated in previous steps (first by unsupervised segmentation then by neural networks) is used as ground truth for the next level of network learning. Experiments on a cystic lung lesion dataset show that the deep learning could perform better than the initial unsupervised annotation, and progressively improve itself after self-learning., Comment: 4 pages, 3 figures

This paper reports Deep LOGISMOS approach to 3D tumor segmentation by incorporating boundary information derived from deep contextual learning to LOGISMOS - layered optimal graph image segmentation of multiple objects and surfaces. Accurate and reliable tumor segmentation is essential to tumor growth analysis and treatment selection. A fully convolutional network (FCN), UNet, is first trained using three adjacent 2D patches centered at the tumor, providing contextual UNet segmentation and probability map for each 2D patch. The UNet segmentation is then refined by Gaussian Mixture Model (GMM) and morphological operations. The refined UNet segmentation is used to provide the initial shape boundary to build a segmentation graph. The cost for each node of the graph is determined by the UNet probability maps. Finally, a max-flow algorithm is employed to find the globally optimal solution thus obtaining the final segmentation. For evaluation, we applied the method to pancreatic tumor segmentation on a dataset of 51 CT scans, among which 30 scans were used for training and 21 for testing. With Deep LOGISMOS, DICE Similarity Coefficient (DSC) and Relative Volume Difference (RVD) reached 83.2+-7.8% and 18.6+-17.4% respectively, both are significantly improved (p<0.05) compared with contextual UNet and/or LOGISMOS alone., Comment: 4 pages,3 figures

Automation-assisted cervical screening via Pap smear or liquid-based cytology (LBC) is a highly effective cell imaging based cancer detection tool, where cells are partitioned into "abnormal" and "normal" categories. However, the success of most traditional classification methods relies on the presence of accurate cell segmentations. Despite sixty years of research in this field, accurate segmentation remains a challenge in the presence of cell clusters and pathologies. Moreover, previous classification methods are only built upon the extraction of hand-crafted features, such as morphology and texture. This paper addresses these limitations by proposing a method to directly classify cervical cells - without prior segmentation - based on deep features, using convolutional neural networks (ConvNets). First, the ConvNet is pre-trained on a natural image dataset. It is subsequently fine-tuned on a cervical cell dataset consisting of adaptively re-sampled image patches coarsely centered on the nuclei. In the testing phase, aggregation is used to average the prediction scores of a similar set of image patches. The proposed method is evaluated on both Pap smear and LBC datasets. Results show that our method outperforms previous algorithms in classification accuracy (98.3%), area under the curve (AUC) (0.99) values, and especially specificity (98.3%), when applied to the Herlev benchmark Pap smear dataset and evaluated using five-fold cross-validation. Similar superior performances are also achieved on the HEMLBC (H&E stained manual LBC) dataset. Our method is promising for the development of automation-assisted reading systems in primary cervical screening.

The vertebral body is the main axial load-bearing structure of the spinal vertebra. Assessment of acute injury and chronic deformity of the vertebral body is difficult to assess accurately and quantitatively by simple visual inspection. We propose a cortical shell unwrapping method to examine the vertebral body for injury such as fractures and degenerative osteophytes. The spine is first segmented and partitioned into vertebrae. Then the cortical shell of the vertebral body is extracted using deformable dual-surface models. The cortical shell is then unwrapped onto a 2D map and the complex 3D detection problem is effectively converted to a pattern recognition problem on a 2D plane. Characteristic features adapted for different applications are computed and sent to a committee of support vector machines for classification. The system was evaluated on two applications, one for fracture detection on trauma CT datasets and the other on degenerative osteophyte assessment on sodium fluoride PET/CT. The fracture CAD achieved 93.6% sensitivity at 3.2 false positive per patient and the degenerative osteophyte CAD achieved 82% sensitivity at 4.7 false positive per patient.

Tumor growth prediction, a highly challenging task, has long been viewed as a mathematical modeling problem, where the tumor growth pattern is personalized based on imaging and clinical data of a target patient. Though mathematical models yield promising results, their prediction accuracy may be limited by the absence of population trend data and personalized clinical characteristics. In this paper, we propose a statistical group learning approach to predict the tumor growth pattern that incorporates both the population trend and personalized data, in order to discover high-level features from multimodal imaging data. A deep convolutional neural network approach is developed to model the voxel-wise spatio-temporal tumor progression. The deep features are combined with the time intervals and the clinical factors to feed a process of feature selection. Our predictive model is pretrained on a group data set and personalized on the target patient data to estimate the future spatio-temporal progression of the patient's tumor. Multimodal imaging data at multiple time points are used in the learning, personalization and inference stages. Our method achieves a Dice coefficient of 86.8% +- 3.6% and RVD of 7.9% +- 5.4% on a pancreatic tumor data set, outperforming the DSC of 84.4% +- 4.0% and RVD 13.9% +- 9.8% obtained by a previous state-of-the-art model-based method.

The recent rapid and tremendous success of deep convolutional neural networks (CNN) on many challenging computer vision tasks largely derives from the accessibility of the well-annotated ImageNet and PASCAL VOC datasets. Nevertheless, unsupervised image categorization (i.e., without the ground-truth labeling) is much less investigated, yet critically important and difficult when annotations are extremely hard to obtain in the conventional way of "Google Search" and crowd sourcing. We address this problem by presenting a looped deep pseudo-task optimization (LDPO) framework for joint mining of deep CNN features and image labels. Our method is conceptually simple and rests upon the hypothesized "convergence" of better labels leading to better trained CNN models which in turn feed more discriminative image representations to facilitate more meaningful clusters/labels. Our proposed method is validated in tackling two important applications: 1) Large-scale medical image annotation has always been a prohibitively expensive and easily-biased task even for well-trained radiologists. Significantly better image categorization results are achieved via our proposed approach compared to the previous state-of-the-art method. 2) Unsupervised scene recognition on representative and publicly available datasets with our proposed technique is examined. The LDPO achieves excellent quantitative scene classification results. On the MIT indoor scene dataset, it attains a clustering accuracy of 75.3%, compared to the state-of-the-art supervised classification accuracy of 81.0% (when both are based on the VGG-VD model)., Comment: WACV 2017. arXiv admin note: text overlap with arXiv:1603.07965 V2: supplementary material appended

PurposeTo design and validate a fully automated computer system for the detection and anatomic localization of traumatic thoracic and lumbar vertebral body fractures at computed tomography (CT).Materials and methodsThis retrospective study was HIPAA compliant. Institutional review board approval was obtained, and informed consent was waived. CT examinations in 104 patients (mean age, 34.4 years; range, 14-88 years; 32 women, 72 men), consisting of 94 examinations with positive findings for fractures (59 with vertebral body fractures) and 10 control examinations (without vertebral fractures), were performed. There were 141 thoracic and lumbar vertebral body fractures in the case set. The locations of fractures were marked and classified by a radiologist according to Denis column involvement. The CT data set was divided into training and testing subsets (37 and 67 subsets, respectively) for analysis by means of prototype software for fully automated spinal segmentation and fracture detection. Free-response receiver operating characteristic analysis was performed.ResultsTraining set sensitivity for detection and localization of fractures within each vertebra was 0.82 (28 of 34 findings; 95% confidence interval [CI]: 0.68, 0.90), with a false-positive rate of 2.5 findings per patient. The sensitivity for fracture localization to the correct vertebra was 0.88 (23 of 26 findings; 95% CI: 0.72, 0.96), with a false-positive rate of 1.3. Testing set sensitivity for the detection and localization of fractures within each vertebra was 0.81 (87 of 107 findings; 95% CI: 0.75, 0.87), with a false-positive rate of 2.7. The sensitivity for fracture localization to the correct vertebra was 0.92 (55 of 60 findings; 95% CI: 0.79, 0.94), with a false-positive rate of 1.6. The most common cause of false-positive findings was nutrient foramina (106 of 272 findings [39%]).ConclusionThe fully automated computer system detects and anatomically localizes vertebral body fractures in the thora

A multiple center milestone study of clinical vertebra segmentation is presented in this paper. Vertebra segmentation is a fundamental step for spinal image analysis and intervention. The first half of the study was conducted in the spine segmentation challenge in 2014 International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI) Workshop on Computational Spine Imaging (CSI 2014). The objective was to evaluate the performance of several state-of-the-art vertebra segmentation algorithms on computed tomography (CT) scans using ten training and five testing dataset, all healthy cases; the second half of the study was conducted after the challenge, where additional 5 abnormal cases are used for testing to evaluate the performance under abnormal cases. Dice coefficients and absolute surface distances were used as evaluation metrics. Segmentation of each vertebra as a single geometric unit, as well as separate segmentation of vertebra substructures, was evaluated. Five teams participated in the comparative study. The top performers in the study achieved Dice coefficient of 0.93 in the upper thoracic, 0.95 in the lower thoracic and 0.96 in the lumbar spine for healthy cases, and 0.88 in the upper thoracic, 0.89 in the lower thoracic and 0.92 in the lumbar spine for osteoporotic and fractured cases. The strengths and weaknesses of each method as well as future suggestion for improvement are discussed. This is the first multi-center comparative study for vertebra segmentation methods, which will provide an up-to-date performance milestone for the fast growing spinal image analysis and intervention. (C) 2016 Elsevier Ltd. All rights reserved.

Dataflow networks have application in various forms of stream processing, for example for parallel processing of multimedia data. The description of dataflow graphs, including their firing behavior, is typically non-compositional and not amenable to separate compilation. This article considers a dataflow language with a type and effect system that captures the firing behavior of actors. This system allows definitions to abstract over actor firing rates, supporting the definition and safe composition of actor definitions where firing rates are not instantiated until a dataflow graph is launched., Comment: In Proceedings QAPL'16, arXiv:1610.07696

Obtaining semantic labels on a large scale radiology image database (215,786 key images from 61,845 unique patients) is a prerequisite yet bottleneck to train highly effective deep convolutional neural network (CNN) models for image recognition. Nevertheless, conventional methods for collecting image labels (e.g., Google search followed by crowd-sourcing) are not applicable due to the formidable difficulties of medical annotation tasks for those who are not clinically trained. This type of image labeling task remains non-trivial even for radiologists due to uncertainty and possible drastic inter-observer variation or inconsistency. In this paper, we present a looped deep pseudo-task optimization procedure for automatic category discovery of visually coherent and clinically semantic (concept) clusters. Our system can be initialized by domain-specific (CNN trained on radiology images and text report derived labels) or generic (ImageNet based) CNN models. Afterwards, a sequence of pseudo-tasks are exploited by the looped deep image feature clustering (to refine image labels) and deep CNN training/classification using new labels (to obtain more task representative deep features). Our method is conceptually simple and based on the hypothesized "convergence" of better labels leading to better trained CNN models which in turn feed more effective deep image features to facilitate more meaningful clustering/labels. We have empirically validated the convergence and demonstrated promising quantitative and qualitative results. Category labels of significantly higher quality than those in previous work are discovered. This allows for further investigation of the hierarchical semantic nature of the given large-scale radiology image database.

Despite the recent advances in automatically describing image contents, their applications have been mostly limited to image caption datasets containing natural images (e.g., Flickr 30k, MSCOCO). In this paper, we present a deep learning model to efficiently detect a disease from an image and annotate its contexts (e.g., location, severity and the affected organs). We employ a publicly available radiology dataset of chest x-rays and their reports, and use its image annotations to mine disease names to train convolutional neural networks (CNNs). In doing so, we adopt various regularization techniques to circumvent the large normal-vs-diseased cases bias. Recurrent neural networks (RNNs) are then trained to describe the contexts of a detected disease, based on the deep CNN features. Moreover, we introduce a novel approach to use the weights of the already trained pair of CNN/RNN on the domain-specific image/text dataset, to infer the joint image/text contexts for composite image labeling. Significantly improved image annotation results are demonstrated using the recurrent neural cascade model by taking the joint image/text contexts into account.

Remarkable progress has been made in image recognition, primarily due to the availability of large-scale annotated datasets and the revival of deep CNN. CNNs enable learning data-driven, highly representative, layered hierarchical image features from sufficient training data. However, obtaining datasets as comprehensively annotated as ImageNet in the medical imaging domain remains a challenge. There are currently three major techniques that successfully employ CNNs to medical image classification: training the CNN from scratch, using off-the-shelf pre-trained CNN features, and conducting unsupervised CNN pre-training with supervised fine-tuning. Another effective method is transfer learning, i.e., fine-tuning CNN models pre-trained from natural image dataset to medical image tasks. In this paper, we exploit three important, but previously understudied factors of employing deep convolutional neural networks to computer-aided detection problems. We first explore and evaluate different CNN architectures. The studied models contain 5 thousand to 160 million parameters, and vary in numbers of layers. We then evaluate the influence of dataset scale and spatial image context on performance. Finally, we examine when and why transfer learning from pre-trained ImageNet (via fine-tuning) can be useful. We study two specific computer-aided detection (CADe) problems, namely thoraco-abdominal lymph node (LN) detection and interstitial lung disease (ILD) classification. We achieve the state-of-the-art performance on the mediastinal LN detection, with 85% sensitivity at 3 false positive per patient, and report the first five-fold cross-validation classification results on predicting axial CT slices with ILD categories. Our extensive empirical evaluation, CNN model analysis and valuable insights can be extended to the design of high performance CAD systems for other medical imaging tasks.

Accurate spine segmentation allows for improved identification and quantitative characterization of abnormalities of the vertebra, such as vertebral fractures. However, in existing automated vertebra segmentation methods on computed tomography (CT) images, leakage into nearby bones such as ribs occurs due to the close proximity of these visibly intense structures in a 3D CT volume. To reduce this error, we propose the use of joint vertebra-rib atlases to improve the segmentation of vertebrae via multi-atlas joint label fusion. Segmentation was performed and evaluated on CTs containing 106 thoracic and lumbar vertebrae from 10 pathological and traumatic spine patients on an individual vertebra level basis. Vertebra atlases produced errors where the segmentation leaked into the ribs. The use of joint vertebra-rib atlases produced a statistically significant increase in the Dice coefficient from 92.5 $\pm$ 3.1% to 93.8 $\pm$ 2.1% for the left and right transverse processes and a decrease in the mean and max surface distance from 0.75 $\pm$ 0.60mm and 8.63 $\pm$ 4.44mm to 0.30 $\pm$ 0.27mm and 3.65 $\pm$ 2.87mm, respectively., Comment: Manuscript to be presented at SPIE Medical Imaging 2016, 27 February - 3 March, 2016, San Diego, California, USA

Injuries of the spine, and its posterior elements in particular, are a common occurrence in trauma patients, with potentially devastating consequences. Computer-aided detection (CADe) could assist in the detection and classification of spine fractures. Furthermore, CAD could help assess the stability and chronicity of fractures, as well as facilitate research into optimization of treatment paradigms. In this work, we apply deep convolutional networks (ConvNets) for the automated detection of posterior element fractures of the spine. First, the vertebra bodies of the spine with its posterior elements are segmented in spine CT using multi-atlas label fusion. Then, edge maps of the posterior elements are computed. These edge maps serve as candidate regions for predicting a set of probabilities for fractures along the image edges using ConvNets in a 2.5D fashion (three orthogonal patches in axial, coronal and sagittal planes). We explore three different methods for training the ConvNet using 2.5D patches along the edge maps of 'positive', i.e. fractured posterior-elements and 'negative', i.e. non-fractured elements. An experienced radiologist retrospectively marked the location of 55 displaced posterior-element fractures in 18 trauma patients. We randomly split the data into training and testing cases. In testing, we achieve an area-under-the-curve of 0.857. This corresponds to 71% or 81% sensitivities at 5 or 10 false-positives per patient, respectively. Analysis of our set of trauma patients demonstrates the feasibility of detecting posterior-element fractures in spine CT images using computer vision techniques such as deep convolutional networks., Comment: To be presented at SPIE Medical Imaging, 2016, San Diego

Classification of vertebral compression fractures (VCF) having osteoporotic or neoplastic origin is fundamental to the planning of treatment. We developed a fracture classification system by acquiring quantitative morphologic and bone density determinants of fracture progression through the use of automated measurements from longitudinal studies. A total of 250 CT studies were acquired for the task, each having previously identified VCFs with osteoporosis or neoplasm. Thirty-six features or each identified VCF were computed and classified using a committee of support vector machines. Ten-fold cross validation on 695 identified fractured vertebrae showed classification accuracies of 0.812, 0.665, and 0.820 for the measured, longitudinal, and combined feature sets respectively., Comment: Contributed 4-Page Paper to be presented at the 2016 IEEE International Symposium on Biomedical Imaging (ISBI), April 13-16, 2016, Prague, Czech Republic

The precise and accurate segmentation of the vertebral column is essential in the diagnosis and treatment of various orthopedic, neurological, and oncological traumas and pathologies. Segmentation is especially challenging in the presence of pathology such as vertebral compression fractures. In this paper, we propose a method to produce segmentations for osteoporotic compression fractured vertebrae by applying a multi-atlas joint label fusion technique for clinical CT images. A total of 170 thoracic and lumbar vertebrae were evaluated using atlases from five patients with varying degrees of spinal degeneration. In an osteoporotic cohort of bundled atlases, registration provided an average Dice coefficient and mean absolute surface distance of 2.7$\pm$4.5% and 0.32$\pm$0.13mm for osteoporotic vertebrae, respectively, and 90.9$\pm$3.0% and 0.36$\pm$0.11mm for compression fractured vertebrae., Comment: MICCAI 2015 Computational Methods and Clinical Applications for Spine Imaging Workshop