Search

Multiple modality clustering seeks to partition objects via leveraging cross-modality relations to provide comprehensive descriptions of the same objects. Current clustering methods rely heavily on accurate affinity measurements among samples. The samplewise affinity is costive to be constructed yet easy to corrupt by the heterogeneous gap. In the era of big data, fast and accurate clustering of multiple modality data remains challenging. To fill the gap, we propose a novel approach to achieve the clustering by focusing on feature matching across different modalities instead of samplewise affinity. First, a feature matching matrix is calculated by measuring the potential featurewise correlations. The obtained matching matrix is decomposed into two bases corresponding to the column and row spaces of feature matching, acting as coded bases within feature spaces of the different modalities. Then, the sample assignment is obtained by jointly reconstructing the samples by the two bases. The feature matching potential and sample assignment are collaboratively learned by an alternating optimization scheme. The proposed method dramatically reduces the computational cost by avoiding the costive samplewise affinity estimation, without sacrificing accuracy. Extensive experiments on the synthetic and real-world datasets demonstrate its superior speed and high accuracy.

The performance of most the clustering methods hinges on the used pairwise affinity, which is usually denoted by a similarity matrix. However, the pairwise similarity is notoriously known for its vulnerability of noise contamination or the imbalance in samples or features, and thus hinders accurate clustering. To tackle this issue, we propose to use information among samples to boost the clustering performance. We proved that a simplified similarity for pairs, denoted by a fourth order tensor, equals to the Kronecker product of pairwise similarity matrices under decomposable assumption, or provide complementary information for which the pairwise similarity missed under indecomposable assumption. Then a high order similarity matrix is obtained from the tensor similarity via eigenvalue decomposition. The high order similarity capturing spatial information serves as a robust complement for the pairwise similarity. It is further integrated with the popular pairwise similarity, named by IPS2, to boost the clustering performance. Extensive experiments demonstrated that the proposed IPS2 significantly outperformed previous similarity-based methods on real-world datasets and it was capable of handling the clustering task over under-sampled and noisy datasets., 10 pages, 7 figures, 2 tables, 4 pages supplementary information appendix

Accurate retinal vessel segmentation is very challenging. Recently, the deep learning based method has greatly improved performance. However, the non-vascular structures usually harm the performance and some low contrast small vessels are hard to be detected after several down-sampling operations. To solve these problems, we design a deep fusion network (DF-Net) including multiscale fusion, feature fusion and classifier fusion for multi-source vessel image segmentation. The multiscale fusion module allows the network to detect blood vessels with different scales. The feature fusion module fuses deep features with vessel responses extracted from a Frangi filter to obtain a compact yet domain invariant feature representation. The classifier fusion module provides the network more supervision. DF-Net also predicts the parameter of the Frangi filter to avoid manually picking the best parameters. The learned Frangi filter enhances the feature map of the multiscale network and restores the edge information loss caused by down-sampling operations. The proposed end-to-end network is easy to train and the inference time for one image is 41ms on a GPU. The model outperforms state-of-the-art methods and achieves the accuracy of 96.14%, 97.04%, 98.02% from three publicly available fundus image datasets DRIVE, STARE, CHASEDB1, respectively. The code is available at https://github.com/y406539259/DF-Net .

Image super-resolution from low-resolution fundus image has valuable applications in clinical practices. The popular methods yield unsatisfactory results when the fundus images are contaminated due to the bleeding or plaques caused by eye diseases. To this end, we propose a style-guided UNet (SGUNet) which incorporates a series of style-guided U-shape block (SUB) for fundus image super-resolution. Each SUB consists of trunk and mask branches. The proposed trunk branch is a U-shape structure that intends to enlarge the receptive field by down-sampling via large-stride convolution, and fuses the complementary information under the different receptive fields. The mask branch then dynamically estimates the relative importance of individual potential styles to reweigh the feature maps according to the significance of the potential styles. To fully leverage the hierarchical features, a dense feature fusion scheme is introduced by concatenating the output of preceding SUBs. We extensively validate the proposed network on low-resolution retina dataset with adversely affected by diseases. The experimental results demonstrate that our SGUNet achieves superior performance with excellent robustness and high accuracy by comparing with six popular methods.

Diabetic retinopathy, age-related macular degeneration and glaucoma, are the leading causes of visual impairment or blindness of the population across different ages. Retinal fundus imaging is a clinically regular tool for the diagnosis of retinal diseases. In the interest of having a comprehensive understanding of the fundus condition, it is valuable to leverage multiple fundus images from different modalities. However, a direct fusion of the multi-source fundus images eases to mis-align the physiological structure or spatial position due to possible eyeball rotations or head movements. The problem turns out to be more severe if the images were corrupted by ill conditions on eyes, such as micro-bleeding and plaques. To tackle this problem, we propose a multi-source registration model for retinal fundus images. Our proposed method considers multiple correspondences and dual structural constraints during the registration process. The method firstly selects adequate feature points by an adjustable threshold selection strategy. Then a feature-guided correspondence estimation model is established to build complementary features. Finally, their spatial transformation is built by using mean shift evolution. The evolution is guided by Tikhonov regularization on dual geometric structures. It overcomes the mess of mean shift vector field and mitigating the ill-posed displacement in field recovery. We have conducted our method on the collected 220 multi-source retinal fundus image pairs, which involve minor and larger displacement or severe retinopathy lesions, as well as additive different intensities of Gaussian noises. Extensive experiments demonstrate that the proposed method consistently outperforms seven feature-based methods.

Multi-view clustering seeks an underlying partition of the data from multiple views. Organizing the data into a tensor and then learning a self-expressive latent one to exploit cross-view information has attracted much attention. Most of the recent works mainly focus on the tensor representation, but they fail to directly extract the task-driving affinity matrix for clustering. Such method is typically modeled by a separated two-stage optimization, making the decoupled representation perform unsatisfactorily. One of the core issues is how to explore the common subspace across all views while learning self-representation tensor. To tackle the problem, we propose to jointly learn the two parts within a united optimization framework for consistent clustering performance, thus obtaining Task-driving Affinity Matrix for accurate Multi-view Clustering (TAMMC). First, the proposed method preserves the local affinities of all views via a graph regularization on self-expressive tensor. Second, by penalizing a Laplacian rank on a learned common subspace, our algorithm can guarantee superior clustering. An effective optimization procedure is proposed for the proposed model. Extensive experiments on eight benchmark datasets well demonstrate that our approach, named by TAMMC, achieves superior performances over other popular methods.

An image-based deep learning model was developed to detect different morphological patterns of diabetic macular edema based on optical coherence tomography images with high accuracy and transparency. This model could help ophthalmologists to make personalized therapeutic strategies for patients with diabetic macular edema., Purpose: To develop a deep learning (DL) model to detect morphologic patterns of diabetic macular edema (DME) based on optical coherence tomography (OCT) images. Methods: In the training set, 12,365 OCT images were extracted from a public data set and an ophthalmic center. A total of 656 OCT images were extracted from another ophthalmic center for external validation. The presence or absence of three OCT patterns of DME, including diffused retinal thickening, cystoid macular edema, and serous retinal detachment, was labeled with 1 or 0, respectively. A DL model was trained to detect three OCT patterns of DME. The occlusion test was applied for the visualization of the DL model. Results: Applying 5-fold cross-validation method in internal validation, the area under the receiver operating characteristic curve for the detection of three OCT patterns (i.e., diffused retinal thickening, cystoid macular edema, and serous retinal detachment) was 0.971, 0.974, and 0.994, respectively, with an accuracy of 93.0%, 95.1%, and 98.8%, respectively, a sensitivity of 93.5%, 94.5%, and 96.7%, respectively, and a specificity of 92.3%, 95.6%, and 99.3%, respectively. In external validation, the area under the receiver operating characteristic curve was 0.970, 0.997, and 0.997, respectively, with an accuracy of 90.2%, 95.4%, and 95.9%, respectively, a sensitivity of 80.1%, 93.4%, and 94.9%, respectively, and a specificity of 97.6%, 97.2%, and 96.5%, respectively. The occlusion test showed that the DL model could successfully identify the pathologic regions most critical for detection. Conclusion: Our DL model demonstrated high accuracy and transparency in the detection of OCT patterns of DME. These results emphasized the potential of artificial intelligence in assisting clinical decision-making processes in patients with DME.

Integrative multi-view subspace clustering aims to partition observed samples into underlying clusters through fusing representative subspace information from different views into a latent space. The clustering performance relies on the accuracy of sample affinity measurement. However, existing approaches leverage the subspace representation of each view and overlook the learning of appropriate sample affinities. This paper proposes to learn a consensus affinity directly by merging subspace representations of different views on a Grassmann manifold while maintaining their geometric structures across these views. The proposed method not only preserves the structure of the most informative individual view, but also discovers a latent common structure across all views. The associated constrained optimization problem is solved using the alternating direction method of multipliers. Extensive experiments on synthetic and real-world datasets show that the proposed method outperforms several state-of-the-art multi-view subspace clustering methods. The affinity matrix obtained by our method can extract highly representative and latent common information to enhance the clustering performance.

Multi-view clustering aims to partition objects based on multiple views through unsupervised learning. To exploit cross-view information, recent advances have shifted the focus from matrix-based to tensor-based subspace learning. Though, both approaches leverage the subspace representation of the data, further higher-order statistics extraction is often overlooked. To tackle the issue, in this paper, a novel multi-view clustering method is proposed to extract both intrinsic sample-wise and view-wise statistics from multi-view data. The multi-view data is represented in tensor form and mapped into a latent tensor subspace to exploit its sample-wise relationship. Through using multi-dimensional sparse coding in its similarity tensor, the view-wise relationship is exploited and evacuated. The two relationships are jointly learned and a latent clustering structure is essentially captured in a data-driven way. We conducted extensive experiments on face, object, digital ima ge, and text datasets and compared with ten state-of-the-art methods. The experimental results demonstrated that the proposed method, named as Exsavi, outperform the baselines in terms of various evaluation metrics.

Deep subspace clustering based on autoencoder and self-expression layer has become popular for clustering method. However, these models only focus on feature extraction by minimizing reconstruction error rather than specific clustering task, leading to unsatisfactory performance. To overcome the above shortcoming, we propose a deep subspace clustering framework which jointly extracts features via the embedding neural network and performs subspace learning. The model contains three modules, an autoencoder, a self-expression layer and a supervised competitive feature learning module. The proposed model is highly capable of capturing characteristic features to guide the clustering task by using relative entropy to minimize probabilistic cluster assignments and the target variables. The three modules are consolidated to be jointly trained and optimized competitively. Experimental results on five benchmark datasets demonstrate the effectiveness of the proposed deep subspace clustering by comparing with eleven baseline methods.

Nasopharyngeal carcinoma (NPC) is the most common malignant tumor of the nasopharynx. The delicate nature of the nasopharyngeal structures means that noninvasive magnetic resonance imaging (MRI) is the preferred diagnostic technique for NPC. However, NPC is a typically infiltrative tumor, usually with a small volume, and thus, it remains challenging to discriminate it from tightly connected surrounding tissues. To address this issue, this study proposes a voxel-wise discriminate method for locating and segmenting NPC from normal tissues in MRI sequences. The located NPC is refined to obtain its accurate segmentation results by an original multiviewed collaborative dictionary classification (CODL) model. The proposed CODL reconstructs a latent intact space and equips it with discriminative power for the collective multiview analysis task. Experiments on synthetic data demonstrate that CODL is capable of finding a discriminative space for multiview orthogonal data. We then evaluated the method on real NPC. Experimental results show that CODL could accurately discriminate and localize NPCs of different volumes. This method achieved superior performances in segmenting NPC compared with benchmark methods. Robust segmentation results show that CODL can effectively assist clinicians in locating NPC.

Multi-view clustering seeks to partition objects based on various observations by utilizing cross-views to provide a complementary description of the same objects. It remains challenging to effectively fuse the multi-view data with various dimensions as well as different structures into a new yet highly informative form, thus facilitating adequate assignment of the objects. To tackle the issue, we propose a common subspace integration (CSI) model. The CSI actively learns a common subspace by jointly preserving the local geometry of each view, while incorporating a global partition information to enhance its separability during the learning process. It can be easily generalized to its kernel version, thereby popularizing its general usages. An effective alternative optimization scheme is designed to solve the proposed model. Extensive experiments on six real-world datasets were conducted to demonstrate its superiority by comparing with the twelve state-of-art methods.

Analysis dictionary learning (ADL) has been successfully applied to a variety of learning systems. However, the ordinal locality of analysis dictionary has rarely been explored in constructing discriminative terms. In this paper, a discriminative low-rank analysis–synthesis dictionary learning (LR-ASDL) algorithm with the adaptively ordinal locality is proposed for object classification. Specifically, we first explicitly introduce the relations between the analysis atoms and profiles (i.e., row vectors of the coefficients matrix). That is, the similarity between two profiles depends on that between the corresponding analysis atoms. Moreover, an adaptively ordinal locality preserving(AOLP) term is constructed by simultaneously exploiting the profiles and analysis atoms, which can be learned in a supervised way. In this way, the neighborhood correlations between analysis atoms and the high-order ranking information of each analysis atom’s neighbors can be simultaneously preserved in the learning process. Particularly, this helps to uncover the intrinsic underlying data factors and inherit the geometry structure information of training samples. Furthermore, the low-rank model is imposed on the synthesis atoms to further facilitate the learned dictionaries to be more discriminative. Extensive experimental results on eight databases demonstrate that the LR-ASDL algorithm clearly outperforms some analysis and synthesis dictionary learning algorithms using deep and hand-crafted features.

Mammography is successfully used as an effective screening tool for cancer diagnosis. A calcification cluster on mammography is a primary sign of cancer. Early researches have proved the diagnostic value of the calcification, yet their performance is highly dependent on handcrafted image descriptors. Characterizing the calcification mammography in an automatic and robust way remains a challenge. In this paper, the calcification was characterized by descriptors obtained from deep learning and handcrafted descriptors. We compared the performances of different image feature sets on digital mammograms. The feature sets included the deep features alone, the handcrafted features, their combination, and the filtered deep features. Experimental results have demonstrated that the deep features outperform handcrafted features, but the handcrafted features can provide complementary information for deep features. We achieved a classification precision of 89.32% and sensitivity of 86.89% using the filtered deep features, which is the best performance among all the feature sets.

Background To develop a machine learning (ML) model for the prediction of the idiopathic macular hole (IMH) status at 1 month after vitrectomy and internal limiting membrane peeling (VILMP) surgery. Methods A total of 288 IMH eyes from four ophthalmic centers were enrolled. All eyes underwent optical coherence tomography (OCT) examinations upon admission and one month after VILMP. First, 1,792 preoperative macular OCT parameters and 768 clinical variables of 256 eyes from two ophthalmic centers were used to train and internally validate ML models. Second, 224 preoperative macular OCT parameters and 96 clinical variables of 32 eyes from the other two centers were utilized for external validation. To fulfill the purpose of predicting postoperative IMH status (i.e., closed or open), five ML algorithms were trained and internally validated by the ten-fold cross-validation method, while the best-performing algorithm was further tested by an external validation set. Results In the internal validation, the mean area under the receiver operating characteristic curves (AUCs) of the five ML algorithms were 0.882-0.951. The AUC, accuracy, sensitivity, and specificity of the best-performing algorithm (i.e., random forest, RF) were 0.951, 0.892, 0.973, and 0.904, respectively. In the external validation, the AUC of RF was 0.940, with an accuracy of 0.875, a specificity of 0.875, and a sensitivity of 0.958. Conclusions Based on the preoperative OCT parameters and clinical variables, our ML model achieved remarkable accuracy in predicting IMH status after VILMP. Therefore, ML models may help optimize surgical planning for IMH patients in the future.

Transferring makeup from the misaligned reference image is challenging. Previous methods overcome this barrier by computing pixel-wise correspondences between two images, which is inaccurate and computational-expensive. In this paper, we take a different perspective to break down the makeup transfer problem into a two-step extraction-assignment process. To this end, we propose a Style-based Controllable GAN model that consists of three components, each of which corresponds to target style-code encoding, face identity features extraction, and makeup fusion, respectively. In particular, a Part-specific Style Encoder encodes the component-wise makeup style of the reference image into a style-code in an intermediate latent space W. The style-code discards spatial information and therefore is invariant to spatial misalignment. On the other hand, the style-code embeds component-wise information, enabling flexible partial makeup editing from multiple references. This style-code, together with source identity features, is integrated into a Makeup Fusion Decoder equipped with multiple AdaIN layers to generate the final result. Our proposed method demonstrates great flexibility on makeup transfer by supporting makeup removal, shade-controllable makeup transfer, and part-specific makeup transfer, even with large spatial misalignment. Extensive experiments demonstrate the superiority of our approach over state-of-the-art methods. Code is available at https://github.com/makeuptransfer/SCGAN.

Corona Virus Disease (COVID-19) has spread globally quickly, and has resulted in a large number of causalities and medical resources insufficiency in many countries. Reverse-transcriptase polymerase chain reaction (RT-PCR) testing is adopted as biopsy tool for confirmation of virus infection. However, its accuracy is as low as 60-70%, which is inefficient to uncover the infected. In comparison, the chest CT has been considered as the prior choice in diagnosis and monitoring progress of COVID-19 infection. Although the COVID-19 diagnostic systems based on artificial intelligence have been developed for assisting doctors in diagnosis, the small sample size and the excessive time consumption limit their applications. To this end, this paper proposed a diagnosis prototype system for COVID-19 infection testing. The proposed deep learning model is trained and is tested on 2267 CT sequences from 1357 patients clinically confirmed with COVID-19 and 1235 CT sequences from non-infected people. The main highlights of the prototype system are: (1) no data augmentation is needed to accurately discriminate the COVID-19 from normal controls with the specificity of 0.92 and sensitivity of 0.93; (2) the raw DICOM image is not necessary in testing. Highly compressed image like Jpeg can be used to allow a quick diagnosis; and (3) it discriminates the virus infection within 6 seconds and thus allows an online test with light cost. We also applied our model on 48 asymptomatic patients diagnosed with COVID-19. We found that: (1) the positive rate of RT-PCR assay is 63.5% (687/1082). (2) 45.8% (22/48) of the RT-PCR assay is negative for asymptomatic patients, yet the accuracy of CT scans is 95.8%. The online detection system is available: http://212.64.70.65/covid. Graphic Abstract

AimsTo develop a deep learning (DL) model for automatic classification of macular hole (MH) aetiology (idiopathic or secondary), and a multimodal deep fusion network (MDFN) model for reliable prediction of MH status (closed or open) at 1 month after vitrectomy and internal limiting membrane peeling (VILMP).MethodsIn this multicentre retrospective cohort study, a total of 330 MH eyes with 1082 optical coherence tomography (OCT) images and 3300 clinical data enrolled from four ophthalmic centres were used to train, validate and externally test the DL and MDFN models. 266 eyes from three centres were randomly split by eye-level into a training set (80%) and a validation set (20%). In the external testing dataset, 64 eyes were included from the remaining centre. All eyes underwent macular OCT scanning at baseline and 1 month after VILMP. The area under the receiver operated characteristic curve (AUC), accuracy, specificity and sensitivity were used to evaluate the performance of the models.ResultsIn the external testing set, the AUC, accuracy, specificity and sensitivity of the MH aetiology classification model were 0.965, 0.950, 0.870 and 0.938, respectively; the AUC, accuracy, specificity and sensitivity of the postoperative MH status prediction model were 0.904, 0.825, 0.977 and 0.766, respectively; the AUC, accuracy, specificity and sensitivity of the postoperative idiopathic MH status prediction model were 0.947, 0.875, 0.815 and 0.979, respectively.ConclusionOur DL-based models can accurately classify the MH aetiology and predict the MH status after VILMP. These models would help ophthalmologists in diagnosis and surgical planning of MH.

Accurate and reliable state estimation and mapping are the foundation of most autonomous driving systems. In recent years, researchers have focused on pose estimation through geometric feature matching. However, most of the works in the literature assume a static scenario. Moreover, a registration based on a geometric feature is vulnerable to the interference of a dynamic object, resulting in a decline of accuracy. With the development of a deep semantic segmentation network, we can conveniently obtain the semantic information from the point cloud in addition to geometric information. Semantic features can be used as an accessory to geometric features that can improve the performance of odometry and loop closure detection. In a more realistic environment, semantic information can filter out dynamic objects in the data, such as pedestrians and vehicles, which lead to information redundancy in generated map and map-based localization failure. In this paper, we propose a method called LiDAR inertial odometry (LIO) with loop closure combined with semantic information (LIO-CSI), which integrates semantic information to facilitate the front-end process as well as loop closure detection. First, we made a local optimization on the semantic labels provided by the Sparse Point-Voxel Neural Architecture Search (SPVNAS) network. The optimized semantic information is combined into the front-end process of tightly-coupled light detection and ranging (LiDAR) inertial odometry via smoothing and mapping (LIO-SAM), which allows us to filter dynamic objects and improve the accuracy of the point cloud registration. Then, we proposed a semantic assisted scan-context method to improve the accuracy and robustness of loop closure detection. The experiments were conducted on an extensively used dataset KITTI and a self-collected dataset on the Jilin University (JLU) campus. The experimental results demonstrate that our method is better than the purely geometric method, especially in dynamic scenarios, and it has a good generalization ability.

Multi-view graph-based clustering aims to partition samples via fusing similarity graphs from different views into a unified graph. The clustering performance relies on the accuracy of similarity measurement. However, most existing methods utilize a single metric whose similarity measurement can be easily corrupted by noises thus lacking high accuracy and generalization capability. We propose an effective multi-metric similarity graph refinement and fusion method for multi-view clustering. We construct multiple similarity graphs for each view by different metric, exploit a novel refined similarity through symmetric conditional probability to preserve the important similarity information and finally adaptively fuse multiple refined similarity graphs to an informative unified one. Extensive experiments on eight benchmark datasets have validated the effectiveness and superiority of our proposed method comparing to thirteen state-of-the-art methods.

Functional resonance magnetic imaging (fMRI) technology has been widely used in understanding cognition and behavior by characterizing the functional interaction between distant brain regions. Since the network topology of functional connectivity (FC) often dynamically shifts along with the change of brain states, it is challenging to identify the change point (transition between tasks) of functional connectivity without requiring prior knowledge of experiment settings. Although striking efforts have been made to detect changes on BOLD (blood-oxygen-level-dependent) signals, little attention has been paid to characterize the trajectory of whole-brain functional connectivity, which is more closely correlated to brain state change. Since FC is essentially a symmetric positive definite (SPD) correlation matrix, we present a change point detection network (CPD-Net) tailored to (1) learn the low-dimensional geometric feature representations of whole-brain functional connectivity on the Riemannian manifold of SPD matrices, and (2) automatically detect the brain state changes on the unseen functional neuroimages. It is worth noting that our CPD-Net is a manifold-based neural network to the extent that we leverage the alignment between the known functional tasks and the stratification underlying the learned low-dimensional FC feature representation on the Riemannian manifold of SPD matrices to steer the learning of geometric patterns from functional brain networks. We have evaluated the accuracy and replicability of our CPD-Net on task-based fMRI data from HCP (human connectome project) database, where our manifold-based CPD-Net achieves more accurate and consistent results than current learning-based CPD methods.

In recent years, deep learning has been increasingly used to decipher the relationships among protein sequence, structure, and function. Thus far these applications of deep learning have been mostly based on primary sequence information, while the vast amount of tertiary structure information remains untapped. In this study, we devised a self-supervised representation learning framework (PtsRep) to extract the fundamental features of unlabeled protein tertiary structures deposited in the PDB, a total of 35,568 structures. The learned embeddings were challenged with two commonly recognized protein engineering tasks: the prediction of protein stability and prediction of the fluorescence brightness of green fluorescent protein (GFP) variants, with training datasets of 16,431 and 26,198 proteins or variants, respectively. On both tasks, PtsRep outperformed the two benchmark methods UniRep and TAPE-BERT, which were pre-trained on two much larger sets of data of 24 and 32 million protein sequences, respectively. Protein clustering analyses demonstrated that PtsRep can capture the structural signatures of proteins. Further testing of the GFP dataset revealed two important implications for protein engineering: (1) a reduced and experimentally manageable training dataset (20%, or 5,239 variants) yielded a satisfactory prediction performance for PtsRep, achieving a recall rate of 70% for the top 26 brightest variants with 795 variants in the testing dataset retrieved; (2) counter-intuitively, when only the bright variants were used for training, the performances of PtsRep and the benchmarks not only did not worsen but they actually slightly improved. This study provides a new avenue for learning and exploring general protein structural representations for protein engineering.

COVID-19 causes burdens to the ICU. Evidence-based planning and optimal allocation of the scarce ICU resources is urgently needed but remains unaddressed. This study aims to identify variables and test the accuracy to predict the need for ICU admission, death despite ICU care, and among survivors, length of ICU stay, before patients were admitted to ICU. Retrospective data from 733 in-patients confirmed with COVD-19 in Wuhan, China, as of March 18, 2020. Demographic, clinical and laboratory were collected and analyzed using machine learning to build the predictive models. The built machine learning model can accurately assess ICU admission, length of ICU stay, and mortality in COVID-19 patients toward optimal allocation of ICU resources. The prediction can be done by using the clinical data collected within 1-15 days before the actual ICU admission. Lymphocyte absolute value involved in all prediction tasks with a higher AUC. The online predictive system is freely available to the public (http://212.64.70.65:8000/).

The human retinal surface resembles to a sphere while it is captured by two-dimensional (2D) planar imaging to have a stereo sequence in clinical practice. Reconstructing its three-dimensional (3D) structure from the 2D planar retinal images is crucial for analyzing the relationship between the topological morphology and clinical implication. In this regard, we propose to reconstruct the 3D retina structure from 2D stereo fundus images via dynamic registration. The fundus images from different viewpoints are first co-registrated by using multi-scale deep convolutional feature and geometric structure feature by building their transformation function. The aligned images are then mosaicked together and a 3D reconstruction is obtained by a learned weighted smoothing project the registered images onto 3D coordinates. We compare the proposed registration method with five state-of-the-art methods. Extensive experimental results demonstrate that the proposed framework achieves superior performances, even with challenging scenarios in which the tested images are severely degraded by illness, large eyeball rotation and low resolutions.

Accurate nasopharyngeal carcinoma (NPC) segmentation in magnetic resonance image (MRI) is crucial for diagnosis and treatment. However, most existing deep learning methods performed unsatisfactorily, since NPC is infiltrative and typically has a small or even tiny volume with indistinguishable boundary, making it indiscernible from tightly connected surrounding tissue in immense and complex background. To address the background dominant problem, this paper proposes a coarse-to-fine deep model. The proposed model starts with predicting a coarse mask with a well-designed segmentation module, followed by a boundary rendering module, which exploits semantic information from different layers of feature maps to refine the boundary of the coarse mask. The designed rendering module is shown to achieve superior performance with dramatically fewer parameters by operating only on the segmented mask, rather than on the whole feature maps as the popular methods do. Extensive experiments are conducted on a collected dataset consisting of 2000 MRI slices from 596 patients. Experimental results demonstrate that the proposed model not only outperforms six popular segmentation models but also has a considerable generalization capability on existing models.

Deep learning has been increasingly used in protein tertiary structure prediction, a major goal in life science. However, all the algorithms developed so far mostly use protein sequences as input, whereas the vast amount of protein tertiary structure information available in the Protein Data Bank (PDB) database remains largely unused, because of the inherent complexity of 3D data computation. In this study, we propose Protein Structure Camera (PSC) as an approach to convert protein structures into images. As a case study, we developed a deep learning method incorporating PSC (DeepPSC) to reconstruct protein backbone structures from alpha carbon traces. DeepPSC outperformed all the methods currently available for this task. This PSC approach provides a useful tool for protein structure representation, and for the application of deep learning in protein structure prediction and protein engineering.

Fundus image is commonly used in aiding the diagnosis of ophthalmic diseases. A high-resolution (HR) image is valuable to provide the anatomic information on the eye conditions. Recently, image super-resolution (SR) though learning model has been shown to be an economic yet effective way to satisfy the high demands in the clinical practice. However, the reported methods ignore the mutual dependencies of low-and high-resolution images and did not fully exploit the dependencies between channels. To tackle with the drawbacks, we propose a novel network for fundus image SR, named by Fundus Cascaded Channel-wise Attention Network (FC-CAN). The proposed FCCAN cascades channel attention module and dense module jointly to exploit the semantic interdependencies both frequency and domain information across channels. The channel attention module rescales channel maps in spatial domain, while the dense module preserves the HR components by up- and down-sampling operation. Experimental results demonstrate the superiority of our net-work in comparison with the six methods.

Measurement of the width of fetal lateral ventricles (LVs) in prenatal ultrasound (US) images is essential for antenatal neuronographic assessment. However, the manual measurement of LV width is highly subjective and relies on the clinical experience of scanners. To deal with this challenge, we propose a computer-aided detection framework for automatic measurement of fetal LVs in two-dimensional US images. First, we train a deep convolutional network on 2,400 images of LVs to perform pixel-wise segmentation. Then, the number of pixels per centimeter (PPC), a vital parameter for quantifying the caliper in US images, is obtained via morphological operations guided by prior knowledge. The estimated PPC, upon conversion to a physical length, is used to determine the diameter of the LV by employing the minimum enclosing rectangle method. Extensive experiments on a self-collected dataset demonstrate that the proposed method achieves superior performance over manual measurement, with a mean absolute measurement error of 1.8 mm. The proposed method is fully automatic and is shown to be capable of reducing measurement bias caused by improper US scanning.

Clustering objects with heterogeneous attributes captured from different dimensions remains challenging in integrating the multiple dimensional information. Most of the current multi-dimensional clustering models pin on direct sample-wised similarity and fail to exploit hidden mutual affinity among different sampling spaces. Thus, it is hard to capture a legible cluster structure. To tackle this issue, we propose a High-order multi-dimensional Spectral Clustering method (HSC). The proposed HSC aims to learn a high-order similarity to characterize the intrinsic relationship among different dimensional spaces instead of the ordinary similarity. It then performs a clustering task within a latent space by jointly learning the high-order similarity and ordinary similarity. Extensive experiments over synthetic and real-world data sets show that the proposed HSC outperforms benchmark multi-dimensional methods in most scenarios and is capable of revealing a reliable structure concealed across multi-dimensional spaces.

Retinal image contains rich information on the blood vessel and is highly related to vascular diseases. Fully automatic and accurate identification of arteries and veins from the complex background of retinal images is essential for analyzing eye-relevant diseases, and monitoring progressive eye diseases. However, popular methods, including deep learning-based models, performed unsatisfactorily in preserving the connectivity of both the arteries and veins. The results were shown to be disconnected or overlapped by the twos and thus manual calibration was needed to refine the results. To tackle the problem, this paper proposes a topological structure-constrained generative adversarial network (topGAN) to automatically identify and differentiate the arteries and veins from retinal images. The introduced topological structure term can automatically delineate the topological structure properties of retinal blood vessels and greatly improves the vascular connectivity of the entire arteriovenous classification results. We train and evaluate our model on both the AV-DRIVE public available dataset and the CVDG home-owned dataset, which consists of 40 images and 3119 images, respectively. Experiments demonstrate that integrating topological structure constraints can significantly improve the performance of arteriovenous classification. Our method achieves excellent performance with an accuracy of 94.3% on the AV-DRIVE dataset and 93.6% on the CVDG dataset.

Breast cancer is a highly heterogeneous disease that consists of subtypes with distinct genetic features and clinical symptoms. The patients with different subtypes react to different therapies, thus identifying molecular subtypes greatly contributes to precision diagnosis and personalized cancer treatment. PAM50 subtype is a widely accepted standard in breast cancer classification. The large amount of multi-omics data in public database like TCGA greatly contribute to the study of breast cancer subtype identification. However, the imbalance of sample subtypes in the existing database results in a large difficulty in correctly identifying subtypes with small sample size. In this paper, we proposed a novel method to accurately identify the PAM50 subtypes by utilizing the patients' omics profiles in TCGA database. Based on the integrated expression profiles of RNA-seq and Copy Number Alteration (CNA), the proposed method identifies subtype-related patterns by a multi-layer Convolutional Neural Network (CNN). A weighted loss function was applied to alleviate the effects of imbalanced samples, thus contributing to the accurate identification. We demonstrated that our method could identify PAM50 subtypes of patients with high precision (90.02%) and outperformed two benchmark methods.

Nanopore sequencing is promising because of its long read length and high speed. During sequencing, a strand of DNA/RNA passes through a biological nanopore, which causes the current in the pore to fluctuate. During basecalling, context-dependent current measurements are translated into the base sequence of the DNA/RNA strand. Accurate and fast basecalling is vital for downstream analyses such as genome assembly and detecting single-nucleotide polymorphisms and genomic structural variants. However, owing to the various changes in DNA/RNA molecules, noise during sequencing, and limitations of basecalling methods, accurate basecalling remains a challenge. In this paper, we propose Causalcall, which uses an end-to-end temporal convolution-based deep learning model for accurate and fast nanopore basecalling. Developed on a temporal convolutional network (TCN) and a connectionist temporal classification decoder, Causalcall directly identifies base sequences of varying lengths from current measurements in long time series. In contrast to the basecalling models using recurrent neural networks (RNNs), the convolution-based model of Causalcall can speed up basecalling by matrix computation. Experiments on multiple species have demonstrated the great potential of the TCN-based model to improve basecalling accuracy and speed when compared to an RNN-based model. Besides, experiments on genome assembly indicate the utility of Causalcall in reference-based genome assembly.

Background: Coronavirus Disease (COVID-19) causes burdens to the intensive care unit (ICU). Evidence-based planning and optimal allocation of the scarce ICU resources is urgently needed but remains unaddressed. Methods: Retrospective data from 733 in-patients with laboratory-confirmed COVD-19 in Wuhan, China, as of March 18, 2020. Demographic, clinical course and laboratory were collected and analyzed using machine learning and Least Absolute Shrinkage and Selection Operator (LASSO) to build the predictive models. Findings: First, a predictive model using an ensemble learning strategy and classifier voting mechanism was built to identify the ICU admission based on ten factors identified in 909 potential predictors. It yielded a sensitivity of 0.86 and specificity of 0.82. Second, a model built on ten significant variables predicted whether patients would die despite entering the ICU with accuracy and AUC of 92% and 98% using 3-fold cross-validation. Third, a sparse regression model was built to estimate the length of stay in ICU. It yielded that the average difference between the predicted and empirical time is less than one day. Lymphocyte absolute value appeared in all prediction tasks, thus it was a very noteworthy factor. Interpretation: We identified variables and tested the accuracy to predict the need for ICU admission, death despite ICU admission, and among survivors, length of ICU stay, before patients were admitted to ICU. Our predictions provided quantitative and objective evidence for the optimal planning and allocation of ICU resources. We revealed the most predictive variables to assist clinical workflow. Funding Statement: This work was supported by the grant from the National Natural Science Foundation of China (no.61771007). Declaration of Interests: All authors declare no competing interests. Ethics Approval Statement: This study was approved by the First Affiliated Hospital of Guangxi Medical University Hospital Ethics Committee, with the informed consent being waived.

Objectives To evaluate the feasibility of using deep-learning algorithms to classify as normal or abnormal sonographic images of the fetal brain obtained in standard axial planes. Methods We included in the study images retrieved from a large hospital database from 10 251 normal and 2529 abnormal pregnancies. Abnormal cases were confirmed by neonatal ultrasound, follow-up examination or autopsy. After a series of pretraining data processing steps, 15 372 normal and 14 047 abnormal fetal brain images in standard axial planes were obtained. These were divided into training and test datasets (at case level rather than image level), at a ratio of approximately 8:2. The training data were used to train the algorithms for three purposes: performance of image segmentation along the fetal skull, classification of the image as normal or abnormal and localization of the lesion. The accuracy was then tested on the test datasets, with performance of segmentation being assessed using precision, recall and Dice's coefficient (DICE), calculated to measure the extent of overlap between human-labeled and machine-segmented regions. We assessed classification accuracy by calculating the sensitivity and specificity for abnormal images. Additionally, for 2491 abnormal images, we determined how well each lesion had been localized by overlaying heat maps created by an algorithm on the segmented ultrasound images; an expert judged these in terms of how satisfactory was the lesion localization by the algorithm, classifying this as having been done precisely, closely or irrelevantly. Results Segmentation precision, recall and DICE were 97.9%, 90.9% and 94.1%, respectively. For classification, the overall accuracy was 96.3%. The sensitivity and specificity for identification of abnormal images were 96.9% and 95.9%, respectively, and the area under the receiver-operating-characteristics curve was 0.989 (95% CI, 0.986-0.991). The algorithms located lesions precisely in 61.6% (1535/2491) of the abnormal images, closely in 24.6% (614/2491) and irrelevantly in 13.7% (342/2491). Conclusions Deep-learning algorithms can be trained for segmentation and classification of normal and abnormal fetal brain ultrasound images in standard axial planes and can provide heat maps for lesion localization. This study lays the foundation for further research on the differential diagnosis of fetal intracranial abnormalities. Copyright © 2020 ISUOG. Published by John Wiley & Sons Ltd.

Automatic nasopharyngeal carcinoma (NPC) segmentation in magnetic resonance (MR) images remains challenging since NPC is infiltrative and typically has a small or even tiny volume, making it indiscernible from tightly connected surrounding tissues. Recent methods using deep learning models performed unsatisfactorily since the boundary between NPC and its neighbor tissues is difficult to distinguish. In this paper, a novel Convolutional Neural Network (CNN) with recurrent attention modules (RAMs) is proposed to tackle the problem. To enhance the performance of NPC segmentation, the proposed fully automatic NPC segmentation method with recurrent attention exploits the semantic features in higher layers to guide the learning of features in lower layers. Features are fed into RAMs iteratively from the higher layers to the lower ones. The lower layers are updated iteratively by the guidance of higher layers to render with discriminative capability. Our proposed method was validated in a dataset including 596 patients, experimental results demonstrate that our method outperforms state-of-the-art methods.

Magnetic resonance imaging exhibits high sensitivity but low specificity for breast cancer. The present study aimed to investigate whether combining morphology, texture features and kinetic features with diffusion-weighted imaging using quantitative analysis improves the accuracy of discriminating malignant from benign breast masses. In total, 104 and 171 malignant lesions in 205 women were included. Additionally, 13 texture and 11 morphology features were computed from each lesion using a semi-automated segmentation method. To increase prediction accuracy, a newly designed classification model, difference-weighted local hyperplane, was used for statistical analysis of the combined effects of the features for predicting lesion type. The mean apparent diffusion coefficient (ADC) value for each lesion was calculated. Diagnostic performances of morphology and texture features, kinetic features and ADC alone and the combination of them were evaluated using receiver operating characteristics analysis. Malignant lesions had lower mean ADCs than benign lesions. By using 10-fold cross validation scheme, combined morphological and kinetic features achieved a diagnostic average accuracy of 0.87. Adding an ADC threshold of 1.37×10−3 mm2/sec increased the overall averaged accuracy to 0.90. A multivariate model combining ADC values with 6 morphological and kinetic parameters best discriminated malignant from benign lesions. Incorporating morphology and texture features, kinetic features and ADC into a multivariable diagnostic model improves the discriminatory power of breast lesions.

Copy-number variations (CNVs) are associated with complex diseases and particular tumor types. Array-based comparative genomic hybridization (aCGH) is a common approach for the detection of CNVs. Traditional CNV detection methods for multiple aCGH samples mainly use batch samples to find common variations, not accounting for the individual characteristics of each sample. Accurately differentiating both the commonly shared and the individual CNV patterns is pivotal to identify cell populations, or to distinguish cell growth (as in cancer) from invasion of new cells. Our preliminary results have now demonstrated that both the shared and individual CNV patterns have distinctive characteristics after wavelet transform. Methods: To exploit these characteristics, we propose to formulate a quadratic data-separation problem within the wavelet space to discriminate the shared and individual CNVs from raw data. We have elaborated a numerical solution and shown that the solution can be obtained by solving decoupled subproblems. By this approach, computational costs can be limited, enabling efficient application in the analysis of large sequencing datasets. Results: The advantages of our proposed method, called WaveDec, have been demonstrated by comparison with popular CNV-detection methods using synthetic and empirical aCGH data. The performance of WaveDec was further validated by experiments with single-cell-sequencing data. Conclusion: WaveDec can successfully differentiate shared and individual patterns, and performs well even in data contaminated with high levels of noise. Significance: Both the shared and individual patterns can be uniquely characterized as well as effectively decomposed within the wavelet space.

The removal of impulse noise is a prerequisite step in image analysis. The classic partial differential equation (PDE) has achieved a great success in suppressing Gaussian noise, but its performance in reducing impulse noise is less satisfactory. The main difficulty arises from finding a nice diffusion function. To tackle this problem, the paper develops a novel diffusion system to suppress impulse noise. The proposed diffusion system consists of two phases. In the first phase, an effective image filter called Clean Pixel Excluder (CPE) is designed to identify clean pixels from the noisy ones. In the second phase, a robust diffusion model is reformulated by developing a novel diffusion tensor to control the smoothing on both direction and strength adaptively. A numerical scheme based on the multi-scale technique is provided. Extensive experiments on both synthetic and real images show that the proposed system achieves a superior performance over several standard methods in terms of noise suppression and detail preservation.

Users submit queries to an online database via its query interface. Query interface parsing, which is important for many applications, understands the query capabilities of a query interface. Since most query interfaces are organized hierarchically, we present a novel query interface parsing method, StatParser (Statistical Parser), to automatically extract the hierarchical query capabilities of query interfaces. StatParser automatically learns from a set of parsed query interfaces and parses new query interfaces. StatParser starts from a small grammar and enhances the grammar with a set of probabilities learned from parsed query interfaces under the maximum-entropy principle. Given a new query interface, the probability-enhanced grammar identifies the parse tree with the largest global probability to be the query capabilities of the query interface. Experimental results show that StatParser very accurately extracts the query capabilities and can effectively overcome the problems of existing query interface parsers.

Microcalcification is an effective indicator of early breast cancer. To improve the diagnostic accuracy of microcalcifications, this study evaluates the performance of deep learning-based models on large datasets for its discrimination. A semi-automated segmentation method was used to characterize all microcalcifications. A discrimination classifier model was constructed to assess the accuracies of microcalcifications and breast masses, either in isolation or combination, for classifying breast lesions. Performances were compared to benchmark models. Our deep learning model achieved a discriminative accuracy of 87.3% if microcalcifications were characterized alone, compared to 85.8% with a support vector machine. The accuracies were 61.3% for both methods with masses alone and improved to 89.7% and 85.8% after the combined analysis with microcalcifications. Image segmentation with our deep learning model yielded 15, 26 and 41 features for the three scenarios, respectively. Overall, deep learning based on large datasets was superior to standard methods for the discrimination of microcalcifications. Accuracy was increased by adopting a combinatorial approach to detect microcalcifications and masses simultaneously. This may have clinical value for early detection and treatment of breast cancer.

Background Classifying cancers by gene selection is among the most important and challenging procedures in biomedicine. A major challenge is to design an effective method that eliminates irrelevant, redundant, or noisy genes from the classification, while retaining all of the highly discriminative genes. Results We propose a gene selection method, called local hyperplane-based discriminant analysis (LHDA). LHDA adopts two central ideas. First, it uses a local approximation rather than global measurement; second, it embeds a recently reported classification model, K-Local Hyperplane Distance Nearest Neighbor(HKNN) classifier, into its discriminator. Through classification accuracy-based iterations, LHDA obtains the feature weight vector and finally extracts the optimal feature subset. The performance of the proposed method is evaluated in extensive experiments on synthetic and real microarray benchmark datasets. Eight classical feature selection methods, four classification models and two popular embedded learning schemes, including k-nearest neighbor (KNN), hyperplane k-nearest neighbor (HKNN), Support Vector Machine (SVM) and Random Forest are employed for comparisons. Conclusion The proposed method yielded comparable to or superior performances to seven state-of-the-art models. The nice performance demonstrate the superiority of combining feature weighting with model learning into an unified framework to achieve the two tasks simultaneously. Electronic supplementary material The online version of this article (doi:10.1186/s12859-015-0629-6) contains supplementary material, which is available to authorized users.

Background The apparent diffusion coefficient (ADC) is a highly diagnostic factor in discriminating malignant and benign breast masses in diffusion-weighted magnetic resonance imaging (DW-MRI). The combination of ADC and other pictorial characteristics has improved lesion type identification accuracy. The objective of this study was to reassess the findings on an independent patient group by changing the magnetic field from 1.5-Tesla to 3.0-Tesla. Methods This retrospective study consisted of a training group of 234 female patients, including 85 benign and 149 malignant lesions, imaged using 1.5-Tesla MRI, and a test group of 95 female patients, including 19 benign and 85 malignant lesions, imaged using 3.0-Tesla MRI. The lesion of interest was segmented from the raw image and four sets of measurements describing the morphology, kinetics, DW-MRI, and texture of the pictorial properties of each lesion were obtained. Each lesion was characterized by 28 features in total. Three classical machine-learning algorithms were used to build prediction models on the training group, which evaluated the prognostic performance of the multi-sided features in three scenarios. To reduce information redundancy, five highly diagnostic factors were selected to obtain a compact yet informative characterization of the lesion status. Results Three classification models were built on the training of 1.5-Tesla patients and were tested on the independent 3.0-Tesla test group. The following results were found. i) Characterization of breast masses in a multi-sided way dramatically increased prediction performance. The usage of all features gave a higher performance in both sensitivity and specificity than any individual feature groups or their combinations. ii) ADC was a highly effective factor in improving the sensitivity in discriminating malignant from benign masses. iii) Five features, namely ADC, Sum Average, Entropy, Elongation, and Sum Variance, were selected to achieve the highest performance in diagnosis of the 3.0-Tesla patient group. Conclusions The combination of ADC and other multi-sided characteristics can increase the capability of discriminating malignant and benign breast lesions, even under different imaging protocols. The selected compact feature subsets achieved a high diagnostic performance and thus are promising in clinical applications for discriminating lesion type and for personalized treatment planning.

We present a new method in image segmentation that is based on Otsu's method but iteratively searches for subregions of the image for segmentation, instead of treating the full image as a whole region for processing. The iterative method starts with Otsu's threshold and computes the mean values of the two classes as separated by the threshold. Based on the Otsu's threshold and the two mean values, the method separates the image into three classes instead of two as the standard Otsu's method does. The first two classes are determined as the foreground and background and they will not be processed further. The third class is denoted as a to-be-determined (TBD) region that is processed at next iteration. At the succeeding iteration, Otsu's method is applied on the TBD region to calculate a new threshold and two class means and the TBD region is again separated into three classes, namely, foreground, background, and a new TBD region, which by definition is smaller than the previous TBD regions. Then, the new TBD region is processed in the similar manner. The process stops when the Otsu's thresholds calculated between two iterations is less than a preset threshold. Then, all the intermediate foreground and background regions are, respectively, combined to create the final segmentation result. Tests on synthetic and real images showed that the new iterative method can achieve better performance than the standard Otsu's method in many challenging cases, such as identifying weak objects and revealing fine structures of complex objects while the added computational cost is minimal.

PurposeDynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is increasingly used for breast cancer diagnosis as supplementary to conventional imaging techniques. Combining of diffusion-weighted imaging (DWI) of morphology and kinetic features from DCE-MRI to improve the discrimination power of malignant from benign breast masses is rarely reported.Materials and methodsThe study comprised of 234 female patients with 85 benign and 149 malignant lesions. Four distinct groups of features, coupling with pathological tests, were estimated to comprehensively characterize the pictorial properties of each lesion, which was obtained by a semi-automated segmentation method. Classical machine learning scheme including feature subset selection and various classification schemes were employed to build prognostic model, which served as a foundation for evaluating the combined effects of the multi-sided features for predicting of the types of lesions. Various measurements including cross validation and receiver operating characteristics were used to quantify the diagnostic performances of each feature as well as their combination.ResultsSeven features were all found to be statistically different between the malignant and the benign groups and their combination has achieved the highest classification accuracy. The seven features include one pathological variable of age, one morphological variable of slope, three texture features of entropy, inverse difference and information correlation, one kinetic feature of SER and one DWI feature of apparent diffusion coefficient (ADC). Together with the selected diagnostic features, various classical classification schemes were used to test their discrimination power through cross validation scheme. The averaged measurements of sensitivity, specificity, AUC and accuracy are 0.85, 0.89, 90.9% and 0.93, respectively.ConclusionMulti-sided variables which characterize the morphological, kinetic, pathological properties and DWI measurement of ADC can dramatically improve the discriminatory power of breast lesions.

Background Modeling high-dimensional data involving thousands of variables is particularly important for gene expression profiling experiments, nevertheless,it remains a challenging task. One of the challenges is to implement an effective method for selecting a small set of relevant genes, buried in high-dimensional irrelevant noises. RELIEF is a popular and widely used approach for feature selection owing to its low computational cost and high accuracy. However, RELIEF based methods suffer from instability, especially in the presence of noisy and/or high-dimensional outliers. Results We propose an innovative feature weighting algorithm, called LHR, to select informative genes from highly noisy data. LHR is based on RELIEF for feature weighting using classical margin maximization. The key idea of LHR is to estimate the feature weights through local approximation rather than global measurement, which is typically used in existing methods. The weights obtained by our method are very robust in terms of degradation of noisy features, even those with vast dimensions. To demonstrate the performance of our method, extensive experiments involving classification tests have been carried out on both synthetic and real microarray benchmark datasets by combining the proposed technique with standard classifiers, including the support vector machine (SVM), k-nearest neighbor (KNN), hyperplane k-nearest neighbor (HKNN), linear discriminant analysis (LDA) and naive Bayes (NB). Conclusion Experiments on both synthetic and real-world datasets demonstrate the superior performance of the proposed feature selection method combined with supervised learning in three aspects: 1) high classification accuracy, 2) excellent robustness to noise and 3) good stability using to various classification algorithms.

Currently, most feature weights estimation methods are independent on the classification algorithms. The combination of discriminant analysis and classifiers for effective pattern classification remains heuristic. The present study address the topics of learning of feature weights by using a recently reported classification algorithm, K-Local Hyperplane Distance Nearest Neighbor (HKNN) [18], in which the data is modeled as embedded in a linear hyperplane. Motivated by the encouraging performance of the Learning Discriminative Projections and Prototypes, the feature weights are estimated by minimizing the classifier leave-one-out cross validation error of HKNN. Approximated explicit solution is obtained to give feature estimation. Therefore, the feature weighting and classification are perfectly matched. The performance of the combinational model is extensively assessed via experiments on both synthetic and benchmark datasets. The superior results demonstrate that the method is competitive compared with some state-of-art models.

Morphology of lymph nodal metastasis is critical for diagnosis and prognosis of cancer patients. However, accurate prediction of lymph node type based on morphological information is rarely available due to lack of pathological validation. To obtain correct morphological information, lymph nodes must be segmented from computed tomography (CT) image accurately. In this paper we described a novel approach to segment and predict the status of lymph nodes from CT images and confirmed the diagnostic performance by clinical pathological results. We firstly removed noise and preserved edge details using a revised nonlinear diffusion equation, and secondly we used a repulsive-force-based snake method to segment the lymph nodes. Morphological measurements for the characterization of the node status were obtained from the segmented node image. These measurements were further selected to derive a highly representative set of node status, called feature vector. Finally, classical classification scheme based on support vector machine model was employed to simulate the prediction of nodal status. Experiments on real clinical rectal cancer data showed that the prediction performance with the proposed framework is highly consistent with pathological results. Therefore, this novel algorithm is promising for status prediction of lymph nodes.