Segmentation of the left atrial appendage based on fusion attention.

In clinical practice, the morphology of the left atrial appendage (LAA) plays an important role in the selection of LAA closure devices for LAA closure procedures. The morphology determination is influenced by the segmentation results. The LAA occupies only a small part of the entire 3D medical image, and the segmentation results are more likely to be biased towards the background region, making the segmentation of the LAA challenging. In this paper, we propose a lightweight attention mechanism called fusion attention, which imitates human visual behavior. We process the 3D image of the LAA using a method that involves overview observation followed by detailed observation. In the overview observation stage, the image features are pooled along the three dimensions of length, width, and height. The obtained features from the three dimensions are then separately input into the spatial attention and channel attention modules to learn the regions of interest. In the detailed observation stage, the attention results from the previous stage are fused using element-wise multiplication and combined with the original feature map to enhance feature learning. The fusion attention mechanism was evaluated on a left atrial appendage dataset provided by Liaoning Provincial People's Hospital, resulting in an average Dice coefficient of 0.8855. The results indicate that the fusion attention mechanism achieves better segmentation results on 3D images compared to existing lightweight attention mechanisms.

Electrostatic Smart Textiles for Braille-To-Speech Translation.

A wearable Braille-to-speech translation system is of great importance for providing auditory feedback in assisting blind people and people with speech impairment. However, previous reported Braille-to-speech translation systems still need to be improved in terms of comfortability or integration. Here, a Braille-to-speech translation system that uses dual-functional electrostatic transducers which are made of fabric-based materials and can be integrated into textiles is reported. Based on electrostatic induction, the electrostatic transducer can either serve as a tactile sensor or a loudspeaker with the same design. The proposed electrostatic transducers have excellent output performances, mechanical robustness, and working stability. By combining the devices with machine learning algorithms, it is possible to translate the Braille alphabet and 40 commonly used words (extensible) into speech with an accuracy of 99.09% and 97.08%, respectively. This work demonstrates a new approach for further developments of advanced assistive technology toward improving the lives of disabled people.

MOTC: Abdominal Multi-objective Segmentation Model with Parallel Fusion of Global and Local Information.

Convolutional Neural Networks have been widely applied in medical image segmentation. However, the existence of local inductive bias in convolutional operations restricts the modeling of long-term dependencies. The introduction of Transformer enables the modeling of long-term dependencies and partially eliminates the local inductive bias in convolutional operations, thereby improving the accuracy of tasks such as segmentation and classification. Researchers have proposed various hybrid structures combining Transformer and Convolutional Neural Networks. One strategy is to stack Transformer blocks and convolutional blocks to concentrate on eliminating the accumulated local bias of convolutional operations. Another strategy is to nest convolutional blocks and Transformer blocks to eliminate bias within each nested block. However, due to the granularity of bias elimination operations, these two strategies cannot fully exploit the potential of Transformer. In this paper, a parallel hybrid model is proposed for segmentation, which includes a Transformer branch and a Convolutional Neural Network branch in encoder. After parallel feature extraction, inter-layer information fusion and exchange of complementary information are performed between the two branches, simultaneously extracting local and global features while eliminating the local bias generated by convolutional operations within the current layer. A pure convolutional operation is used in decoder to obtain final segmentation results. To validate the impact of the granularity of bias elimination operations on the effectiveness of local bias elimination, the experiments in this paper were conducted on Flare21 dataset and Amos22 dataset. The average Dice coefficient reached 92.65% on Flare21 dataset, and 91.61% on Amos22 dataset, surpassing comparative methods. The experimental results demonstrate that smaller granularity of bias elimination operations leads to better performance.

Single Channel Based Interference-Free and Self-Powered Human-Machine Interactive Interface Using Eigenfrequency-Dominant Mechanism.

The recent development of wearable devices is revolutionizing the way of human-machine interaction (HMI). Nowadays, an interactive interface that carries more embedded information is desired to fulfill the increasing demand in era of Internet of Things. However, present approach normally relies on sensor arrays for memory expansion, which inevitably brings the concern of wiring complexity, signal differentiation, power consumption, and miniaturization. Herein, a one-channel based self-powered HMI interface, which uses the eigenfrequency of magnetized micropillar (MMP) as identification mechanism, is reported. When manually vibrated, the inherent recovery of the MMP causes a damped oscillation that generates current signals because of Faraday's Law of induction. The time-to-frequency conversion explores the MMP-related eigenfrequency, which provides a specific solution to allocate diverse commands in an interference-free behavior even with one electric channel. A cylindrical cantilever model is built to regulate the MMP eigenfrequencies via precisely designing the dimensional parameters and material properties. It is shown that using one device and two electrodes, high-capacity HMI interface can be realized when the magnetic micropillars (MMPs) with different eigenfrequencies have been integrated. This study provides the reference value to design the future HMI system especially for situations that require a more intuitive and intelligent communication experience with high-memory demand.

Temporal topic model for clinical pathway mining from electronic medical records.

BACKGROUND: In recent years, the discovery of clinical pathways (CPs) from electronic medical records (EMRs) data has received increasing attention because it can directly support clinical doctors with explicit treatment knowledge, which is one of the key challenges in the development of intelligent healthcare services. However, the existing work has focused on topic probabilistic models, which usually produce treatment patterns with similar treatment activities, and such discovered treatment patterns do not take into account the temporal process of patient treatment which does not meet the needs of practical medical applications. METHODS: Based on the assumption that CPs can be derived from the data of EMRs which usually record the treatment process of patients, this paper proposes a new CPs mining method from EMRs, an extended form of the traditional topic model - the temporal topic model (TTM). The method can capture the treatment topics and the corresponding treatment timestamps for each treatment day. RESULTS: Experimental research conducted on a real-world dataset of patients' hospitalization processes, and the achieved results demonstrate the applicability and usefulness of the proposed methodology for CPs mining. Compared to existing benchmarks, our model shows significant improvement and robustness. CONCLUSION: Our TTM provides a more competitive way to mine potential CPs considering the temporal features of the EMR data, providing a very prospective tool to support clinical diagnostic decisions.

Crosstalk-Free Position Mapping for One-Step Reconstruction of Surface Topological Information via Eigenfrequency-Registered Wearable Interface.

Exploring flexible tactile sensors capable of recognizing surface information is significant for the development of virtual reality, artificial intelligence, soft robotics, and human-machine interactions (HMI). However, it is still a challenge for current tactile sensors to efficiently recognize the surface pattern information while maintaining the simplicity of the overall system. In this study, cantilever beam-like magnetized micropillars (MMPs) with height gradients are assembled as a position-registered array for rapid recognition of surface pattern information. After crossing the surface location with convex patterns, the deformed MMPs undergo an intrinsic oscillating process to induce damped electrical signals, which can then be converted to a frequency domain for eigenfrequency extraction. Via precisely defining the specific eigenfrequencies of different MMPs, position mapping is realized in crosstalk-free behavior even though all signals are processed by one communication channel and a pair of electrodes. With a customized LabVIEW program, the surface information (e.g., letters, numbers, and Braille) can be accurately reconstructed by the frequency sequence produced in a single scanning procedure. We expect that the proposed interface can be a convenient and powerful platform for intelligent surface information perception and an HMI system in the future.

A new segment method for pulmonary artery and vein.

Accurate differentiation between pulmonary arteries and veins (A/V) holds pivotal importance in the realm of diagnosing and treating pulmonary ailments. This study presents a new approach that leverages grayscale differences between A/V. Distinctions are measured using median and mean grayscale values within the vessel area. Initially, adherent regions are removed based on vessel structure. The trunk regions are segmented using gray level information near the heart region of the lung boundary. Incorrectly segmented vessels are corrected based on connectivity. For distal lung vessels, a similar distance field is established using a graph-cut method. Experimental results show the algorithm's superior segmentation accuracy, achieving 97.26% compared to the CNN-based average accuracy of 91.67%. Error branches are more concentrated, aiding subsequent manual and automatic correction. This demonstrates the algorithm's effective segmentation of pulmonary A/V.

Segmentation of Clinical Target Volume From CT Images for Cervical Cancer Using Deep Learning.

Introduction: Segmentation of clinical target volume (CTV) from CT images is critical for cervical cancer brachytherapy, but this task is time-consuming, laborious, and not reproducible. In this work, we aim to propose an end-to-end model to segment CTV for cervical cancer brachytherapy accurately. Methods: In this paper, an improved M-Net model (Mnet_IM) is proposed to segment CTV of cervical cancer from CT images. An input and an output branch are both proposed to attach to the bottom layer to deal with CTV locating challenges due to its lower contrast than surrounding organs and tissues. A progressive fusion approach is then proposed to recover the prediction results layer by layer to enhance the smoothness of segmentation results. A loss function is defined on each of the multiscale outputs to form a deep supervision mechanism. Numbers of feature map channels that are directly connected to inputs are finally homogenized for each image resolution to reduce feature redundancy and computational burden. Result: Experimental results of the proposed model and some representative models on 5438 image slices from 53 cervical cancer patients demonstrate advantages of the proposed model in terms of segmentation accuracy, such as average surface distance, 95% Hausdorff distance, surface overlap, surface dice, and volumetric dice. Conclusion: A better agreement between the predicted CTV from the proposed model Mnet_IM and manually labeled ground truth is obtained compared to some representative state-of-the-art models.

Washable and Breathable Electret Sensors Based on a Hydro-Charging Technique for Smart Textiles.

Flexible electromechanical sensors based on electret materials have shown great application potential in wearable electronics. However, achieving great breathability yet maintaining good washability is still a challenge for traditional electret sensors. Herein, we report a washable and breathable electret sensor based on a hydro-charging technique, namely, hydro-charged electret sensor (HCES). The melt-blown polypropylene (MBPP) electret fabric can be charged while washing with water. The surface potential of MBPP electret fabric can be improved by optimizing the type of water, water pressure, water temperature, drying temperature, drying time, ambient air pressure, and ambient relative humidity. It is proposed that the single fiber has charges of different polarities on the upper and lower surfaces due to contact electrification with water, thereby forming electric dipoles between fibers, which can lead to better surface potential stability than the traditional corona-charging method. The HCES can achieve a high air permeability of ∼215 mm/s and sensitivity up to ∼0.21 V/Pa, with output voltage remaining stable after over 36,000 working cycles and multiple times of water washing. As a demonstration example, the HCES is integrated into a chest strap to monitor human respiration conditions.

Biodegradable Smart Face Masks for Machine Learning-Assisted Chronic Respiratory Disease Diagnosis.

Utilizing smart face masks to monitor and analyze respiratory signals is a convenient and effective method to give an early warning for chronic respiratory diseases. In this work, a smart face mask is proposed with an air-permeable and biodegradable self-powered breath sensor as the key component. This smart face mask is easily fabricated, comfortable to use, eco-friendly, and has sensitive and stable output performances in real wearable conditions. To verify the practicability, we use smart face masks to record respiratory signals of patients with chronic respiratory diseases when the patients do not have obvious symptoms. With the assistance of the machine learning algorithm of the bagged decision tree, the accuracy for distinguishing the healthy group and three groups of chronic respiratory diseases (asthma, bronchitis, and chronic obstructive pulmonary disease) is up to 95.5%. These results indicate that the strategy of this work is feasible and may promote the development of wearable health monitoring systems.

Self-supervised graph neural network with pre-training generative learning for recommendation systems.

The case assignment system is an essential system of case management and assignment within the procuratorate and is an important aspect of judicial fairness and efficiency. However, existing methods mostly use manual or random case assignment, which leads to unbalanced case distribution. Moreover, the relationship between prosecutors and case categories usually shows a power-law distribution in real-world data. Therefore, in this paper, we describe the case rationality assignment as a recommendation problem under the power-law distributed data. To solve the above problems, we propose an end-to-end Self-supervised Graph neural network model with Pre-training Generative learning for Recommendation (SGPGRec), the main idea of which is to capture self-supervised signals using intra-node features and inter-node correlations in the data, and generate the data representation by pre-training to improve the recommendation results. To be specific, we designed three auxiliary self-supervised tasks based on the prosecutor-case category interaction graph and the data distribution to obtain feature representations of prosecutors, case categories, and the interaction information between them. Then we constructed an end-to-end graph neural network recommendation model by the interaction information based on the data characteristics of the power-law distribution. Finally, extensive experimental consistency on a real-world dataset from three procuratorates shows that our method is effective compared to several yet competing baseline methods and further supports the development of an intelligent case assignment system with adequate performance.

Dual-level diagnostic feature learning with recurrent neural networks for treatment sequence recommendation.

In recent years, the massive electronic medical records (EMRs) have supported the development of intelligent medical services such as treatment recommendations. However, existing treatment recommendations usually follow the traditional sequential recommendation strategies, ignoring the partial temporality of the practical process and the patient's diagnostic features. To this end, in this paper, we propose a new Dual-level Diagnostic Feature Learning with Recurrent Neural Network for treatment sequence recommendation (DDFL-RNN), where the dual-level diagnostic features including patients' historical medical records and current treatment results. Firstly, we divide the dataset into several sequential sets of treatment item based on the patient's treatment days. Secondly, we propose two kinds of attention mechanisms to learn diagnostic features, including the elemental attention mechanism and the sequential attention mechanism. Finally, the dual-level learned diagnostic features are brought into the recurrent neural network for encoding and recommendation. Extensive experiments on a breast cancer dataset from a first-rate hospital have shown that our model achieves significantly better performance than several classical and state-of-the-art baseline methods.

A novel biomarker selection method combining graph neural network and gene relationships applied to microarray data.

BACKGROUND: The discovery of critical biomarkers is significant for clinical diagnosis, drug research and development. Researchers usually obtain biomarkers from microarray data, which comes from the dimensional curse. Feature selection in machine learning is usually used to solve this problem. However, most methods do not fully consider feature dependence, especially the real pathway relationship of genes. RESULTS: Experimental results show that the proposed method is superior to classical algorithms and advanced methods in feature number and accuracy, and the selected features have more significance. METHOD: This paper proposes a feature selection method based on a graph neural network. The proposed method uses the actual dependencies between features and the Pearson correlation coefficient to construct graph-structured data. The information dissemination and aggregation operations based on graph neural network are applied to fuse node information on graph structured data. The redundant features are clustered by the spectral clustering method. Then, the feature ranking aggregation model using eight feature evaluation methods acts on each clustering sub-cluster for different feature selection. CONCLUSION: The proposed method can effectively remove redundant features. The algorithm's output has high stability and classification accuracy, which can potentially select potential biomarkers.

Deep learning-based breast region extraction of mammographic images combining pre-processing methods and semantic segmentation supported by Deeplab v3.

BACKGROUND: Breast cancer has long been one of the major global life-threatening illnesses among women. Surgery and adjuvant therapy, coupled with early detection, could save many lives. This underscores the importance of mammography, a cost-effective and accurate method for early detection. Due to the poor contrast, noise and artifacts which results in difficulty for radiologists to diagnose, Computer-Aided Diagnosis (CAD) systems are hence developed. The extraction of breast region is a fundamental and crucial preparation step for further development of CAD systems. OBJECTIVE: The proposed method aims to extract breast region accurately from mammographic images where noise is suppressed, contrast is enhanced and pectoral muscle region is removed. METHODS: This paper presents a new deep learning-based breast region extraction method that combines pre-processing methods containing noise suppression using median filter, contrast enhancement using CLAHE and semantic segmentation using Deeplab v3+ model. RESULTS: The method is trained and evaluated on mini-MIAS dataset. It has also been evaluated on INbreast dataset. The results outperform those generated by other recent researches and are indicative of the capacity of the model to retain its accuracy and runtime advantage across different databases with different image resolutions. CONCLUSIONS: The proposed method shows state-of-the-art performance at extracting breast region from mammographic images. Wide range of evaluation on two commonly used mammography datasets proves the ability and adaptability of the method.

A hybrid approach based on deep learning and level set formulation for liver segmentation in CT images.

Accurate liver segmentation is essential for radiation therapy planning of hepatocellular carcinoma and absorbed dose calculation. However, liver segmentation is a challenging task due to the anatomical variability in both shape and size and the low contrast between liver and its surrounding organs. Thus we propose a convolutional neural network (CNN) for automated liver segmentation. In our method, fractional differential enhancement is firstly applied for preprocessing. Subsequently, an initial liver segmentation is obtained by using a CNN. Finally, accurate liver segmentation is achieved by the evolution of an active contour model. Experimental results show that the proposed method outperforms existing methods. One hundred fifty CT scans are evaluated for the experiment. For liver segmentation, Dice of 95.8%, true positive rate of 95.1%, positive predictive value of 93.2%, and volume difference of 7% are calculated. In addition, the values of these evaluation measures show that the proposed method is able to provide a precise and robust segmentation estimate, which can also assist the manual liver segmentation task.

Abdomen tissues segmentation from computed tomography images using deep learning and level set methods.

Accurate abdomen tissues segmentation is one of the crucial tasks in radiation therapy planning of related diseases. However, abdomen tissues segmentation (liver, kidney) is difficult because the low contrast between abdomen tissues and their surrounding organs. In this paper, an attention-based deep learning method for automated abdomen tissues segmentation is proposed. In our method, image cropping is first applied to the original images. U-net model with attention mechanism is then constructed to obtain the initial abdomen tissues. Finally, level set evolution which consists of three energy terms is used for optimize the initial abdomen segmentation. The proposed model is evaluated across 470 subsets. For liver segmentation, the mean dice are 96.2 and 95.1% for the FLARE21 datasets and the LiTS datasets, respectively. For kidney segmentation, the mean dice are 96.6 and 95.7% for the FLARE21 datasets and the LiTS datasets, respectively. Experimental evaluation exhibits that the proposed method can obtain better segmentation results than other methods.

A Global Inhomogeneous Intensity Clustering- (GINC-) Based Active Contour Model for Image Segmentation and Bias Correction.

Image segmentation is still an open problem especially when intensities of the objects of interest are overlapped due to the presence of intensity inhomogeneities. A bias correction embedded level set model is proposed in this paper where inhomogeneities are estimated by orthogonal primary functions. First, an inhomogeneous intensity clustering energy is defined based on global distribution characteristics of the image intensities, and membership functions of the clusters described by the level set function are then introduced to define the data term energy of the proposed model. Second, a regularization term and an arc length term are also included to regularize the level set function and smooth its zero-level set contour, respectively. Third, the proposed model is extended to multichannel and multiphase patterns to segment colorful images and images with multiple objects, respectively. Experimental results and comparison with relevant models demonstrate the advantages of the proposed model in terms of bias correction and segmentation accuracy on widely used synthetic and real images and the BrainWeb and the IBSR image repositories.

MISS-D: A fast and scalable framework of medical image storage service based on distributed file system.

Background and Objective Processing of medical imaging big data is deeply challenging due to the size of data, computational complexity, security storage and inherent privacy issues. Traditional picture archiving and communication system, which is an imaging technology used in the healthcare industry, generally uses centralized high performance disk storage arrays in the practical solutions. The existing storage solutions are not suitable for the diverse range of medical imaging big data that needs to be stored reliably and accessed in a timely manner. The economical solution is emerging as the cloud computing which provides scalability, elasticity, performance and better managing cost. Cloud based storage architecture for medical imaging big data has attracted more and more attention in industry and academia. Methods This study presents a novel, fast and scalable framework of medical image storage service based on distributed file system. Two innovations of the framework are introduced in this paper. An integrated medical imaging content indexing file model for large-scale image sequence is designed to adapt to the high performance storage efficiency on distributed file system. A virtual file pooling technology is proposed, which uses the memory-mapped file method to achieve an efficient data reading process and provides the data swapping strategy in the pool. Result The experiments show that the framework not only has comparable performance of reading and writing files which meets requirements in real-time application domain, but also bings greater convenience for clinical system developers by multiple client accessing types. The framework supports different user client types through the unified micro-service interfaces which basically meet the needs of clinical system development especially for online applications. The experimental results demonstrate the framework can meet the needs of real-time data access as well as traditional picture archiving and communication system. Conclusions This framework aims to allow rapid data accessing for massive medical images, which can be demonstrated by the online web client for MISS-D framework implemented in this paper for real-time data interaction. The framework also provides a substantial subset of features to existing open-source and commercial alternatives, which has a wide range of potential applications.

Molecular Subtypes Recognition of Breast Cancer in Dynamic Contrast-Enhanced Breast Magnetic Resonance Imaging Phenotypes from Radiomics Data.

BACKGROUND AND OBJECTIVE: Breast cancer is a major cause of mortality among women if not treated in early stages. Recognizing molecular markers from DCE-MRI directly to distinguish the four molecular subtypes without invasive biopsy is helpful for guiding treatment plans for breast cancer, which provides a fast way to consequential treatment plan decision in early time and best opportunity for patients. METHODS: This study presents an approach of molecular subtypes recognition from breast cancer image phenotypes by radiomics. An improved region growth algorithm with dynamic threshold without user interaction is proposed for cancer lesion segmentation, which gives the precise border of lesion other than area with background. The lesions are extracted automatically based on radiologists' annotation which guarantees the lesion is segmented correctly. Various features are extracted on lesions data including texture, morphology, dynamic kinetics, and statistics features carried out on a large patient cohort, which are used to validate the relationship between image phenotypes and the molecular subtypes. A new algorithm of multimodel-based recursive feature elimination is applied on the radiomics data generated by the feature extraction process. This method obtains the feature subset with stable performance for different classification models, and the gradient boosting decision tree model gets the best results of both classification performance and imbalance performance on molecular subtypes. RESULT: From the experimental results, 69 optimal features from 143 original features are found by the multimodel-based recursive feature elimination algorithms and the gradient boosting decision tree classifier obtains a good performance with accuracy 0.87, precise 0.88, recall 0.87, and F1-score 0.87. The dataset with 637 patients in this paper has serious imbalance problem on different molecular subtypes, and the the robust features that are generated by multimodel-based recursive feature eliminiation algorithm make the gradient boosting decision tree classifier have good behaviors. The recognition precision for the four molecular subtypes of luminal A, luminal B, HER-2, and basal-like are 0.91, 0.89, 0.83, and 0.87, respectively. CONCLUSIONS: The improved lesion segmentation method gives more precise lesion edge, which not only saves the time of automatic extraction of lesion region of interest without threshold setting for each case, but also prevents the segmentation error by manual and prejudice from different radiologists. The feature selection algorithm of multimodel-based recursive feature elimination has the ability to find robust and optimal features that distinguish the four molecular subtypes from image phenotypes. The gradient boosting decision tree classifier rather plays a main role in recognition than other models used in this paper.

Fused Group Lasso Regularized Multi-Task Feature Learning and Its Application to the Cognitive Performance Prediction of Alzheimer's Disease.

Alzheimer's disease (AD) is characterized by gradual neurodegeneration and loss of brain function, especially for memory during early stages. Regression analysis has been widely applied to AD research to relate clinical and biomarker data such as predicting cognitive outcomes from MRI measures. Recently, multi-task based feature learning (MTFL) methods with sparsity-inducing [Formula: see text]-norm have been widely studied to select a discriminative feature subset from MRI features by incorporating inherent correlations among multiple clinical cognitive measures. However, existing MTFL assumes the correlation among all tasks is uniform, and the task relatedness is modeled by encouraging a common subset of features via sparsity-inducing regularizations that neglect the inherent structure of tasks and MRI features. To address this issue, we proposed a fused group lasso regularization to model the underlying structures, involving 1) a graph structure within tasks and 2) a group structure among the image features. To this end, we present a multi-task feature learning framework with a mixed norm of fused group lasso and [Formula: see text]-norm to model these more flexible structures. For optimization, we employed the alternating direction method of multipliers (ADMM) to efficiently solve the proposed non-smooth formulation. We evaluated the performance of the proposed method using the Alzheimer's Disease Neuroimaging Initiative (ADNI) datasets. The experimental results demonstrate that incorporating the two prior structures with fused group lasso norm into the multi-task feature learning can improve prediction performance over several competing methods, with estimated correlations of cognitive functions and identification of cognition-relevant imaging markers that are clinically and biologically meaningful.