Extract Representative Features Research Articles

Transformer and cross-attention-based multi-sensor in-situ monitoring of molten pool stability and part quality in laser powder bed fusion

The fluctuations in process parameters, as a random phenomenon in the laser powder bed fusion (LPBF) process, is occasional. These subtle variations can lead to defects such as porosity if not monitored and corrected promptly. Therefore, monitoring the stability of the molten pool and small changes in process parameters is significant. For process monitoring, the commonly used Transformer architectures have demonstrated remarkable performance in single-sensor signal processing. However, as the core component of the Transformer, the self-attention mechanism has limitations in data fusion of multi-sensor signals. To this end, a novel Transformer-based deep learning method associated with the cross-attention mechanism (Trans-Cross) is proposed. Trans-Cross leverages the collected photodiode and acoustic signals to monitor the fluctuation of process parameters as well as the quality of parts. Specifically, the cross-attention mechanism is introduced to enhance representative feature extraction from photodiode and acoustic signals, exploiting their complementary information. To better characterize the raw signals, an improved adaptive variational mode decomposition (VMD) algorithm is proposed for data preprocessing. The proposed Trans-Cross achieves impressive prediction accuracies of 97.41 % for identifying the fluctuations in the process parameters and 99.73 % for part quality classification. Trans-Cross exhibits superior performance and strong robustness when dealing with constraints such as limited input signal length and training data ratio. These results validate the feasibility of the proposed method in accurately identifying the fluctuations in process parameters as well as the part quality. This work provided data guidance to enhance the stability of the LPBF process.

Enhanced domain transfer deep fuzzy echo state network for rotating machinery fault diagnosis based on current signal

Although vibration-based fault diagnosis methods have achieved remarkable results in rotating machinery, they are still limited by various factors, such as environment noise and additional sensor installation. Meanwhile, due to inevitable distribution differences in practical engineering, deep transfer learning models are commonly used in fault diagnostic fields. Although deep architectures can extract representative domain-invariant features, they typically result in substantial computational burdens and training time. Therefore, to address the above problems, an enhanced domain transfer deep fuzzy echo state network (EDFESN) is proposed for rotating machinery fault diagnosis based on current signals. Firstly, current signals derived from the driving motor are directly collected without additional sensors. Secondly, an enhanced domain mapping method is raised to solve the problem of data distribution discrepancies between cross domain. Furthermore, a deep fuzzy echo state network (DFESN) integrated with fuzzy clustering is constructed, which utilizes layer-wise fuzzy-tuning learning paradigm instead of backpropagation step to reduce training time. Two datasets from gearbox and bearing both verify the effectiveness of the proposed method. Compared with other popular transfer methods, EDFESN not only achieves highest accuracy, but also has fastest training speed.

MH2AFormer: An Efficient Multiscale Hierarchical Hybrid Attention With a Transformer for Bladder Wall and Tumor Segmentation.

Achieving accurate bladder wall and tumor segmentation from MRI is critical for diagnosing and treating bladder cancer. However, automated segmentation remains challenging due to factors such as comparable density distributions, intricate tumor morphologies, and unclear boundaries. Considering the attributes of bladder MRI images, we propose an efficient multiscale hierarchical hybrid attention with a transformer (MH2AFormer) for bladder cancer and wall segmentation. Specifically, a multiscale hybrid attention and transformer (MHAT) module in the encoder is designed to adaptively extract and aggregate multiscale hybrid feature representations from the input image. In the decoder stage, we devise a multiscale hybrid attention (MHA) module to generate high-quality segmentation results from multiscale hybrid features. Combining these modules enhances the feature representation and guides the model to focus on tumor and wall regions, which helps to solve bladder image segmentation challenges. Moreover, MHAT utilizes the Fast Fourier Transformer with a large kernel (e.g., 224 × 224) to model global feature relationships while reducing computational complexity in the encoding stage. The model performance was evaluated on two datasets. As a result, the model achieves relatively best results regarding the intersection over union (IoU) and dice similarity coefficient (DSC) on both datasets (Dataset A: IoU = 80.26%, DSC = 88.20%; Dataset B: IoU = 89.74%, DSC = 94.48%). These advantageous outcomes substantiate the practical utility of our approach, highlighting its potential to alleviate the workload of radiologists when applied in clinical settings.

Distribution-based detection of radiographic changes in pneumonia patterns: A COVID-19 case study

Although the classification of chest radiographs has long been an extensively researched topic, interest increased significantly with the onset of the COVID-19 pandemic. Existing results are promising; however, the radiological similarities between COVID-19 and other types of respiratory diseases limit the success of conventional image classification approaches that focus on single instances. This study proposes a novel perspective that conceptualizes COVID-19 pneumonia as a deviation from a normative distribution of typical pneumonia patterns. Using a population-based approach, our approach utilizes distributional anomaly detection. This method diverges from traditional instance-wise approaches by focusing on sets of scans instead of individual images. Using an autoencoder to extract feature representations, we present instance-based and distribution-based assessments of the separability between COVID-positive and COVID-negative pneumonia radiographs. The results demonstrate that the proposed distribution-based methodology outperforms conventional instance-based techniques in identifying radiographic changes associated with COVID-positive cases. This underscores its potential as an early warning system capable of detecting significant distributional shifts in radiographic data. By continuously monitoring these changes, this approach offers a mechanism for early identification of emerging health trends, potentially signaling the onset of new pandemics and enabling prompt public health responses.

ESMformer: Error-aware self-supervised transformer for multi-view 3D human pose estimation

Multi-view 3D human pose estimation (HPE) currently faces several key challenges. Information from different viewpoints exhibits high variability due to complex environmental factors, posing difficulties in cross-view feature extraction and fusion. Additionally, multi-view 3D labeled pose data is rather scarce, and the impact of input 2D poses on 3D HPE accuracy has received little attention. To address these issues, we propose an Error-aware Self-supervised transformer framework for Multi-view 3D HPE (ESMformer). Firstly, we introduce a single-view multi-level feature extraction module to enhance pose features in individual viewpoints, which incorporates a novel relative attention mechanism for representative feature extraction at different levels. Subsequently, we develop multi-view intra-level and cross-level fusion modules to exploit spatio-temporal feature dependencies among human joints, and progressively fuse pose information from all views and levels. Furthermore, we explore an error-aware self-supervised learning strategy to reduce the model’s reliance on 3D pose annotations and mitigate the impact of incorrect 2D poses. This strategy adaptively selects reliable input 2D poses based on 3D pose prediction errors. Experiments on three popular benchmarks show that ESMformer achieves state-of-the-art results and maintains cost-effective computational complexity. Notably, ESMformer does not rely on any 3D pose annotations or prior human body knowledge, making it highly versatile and adaptable in practical applications.11The codes are available at https://github.com/z0911k/ESMformer.

Attentive Review Semantics-Aware Recommendation Model for Rating Prediction

Online reviews are crucial when building a recommendation model because they contain the specific and rich preferences of users related to different aspects of a particular item. Incorporating these online reviews into the recommendation model mitigates the data sparsity issue to some extent and contributes to better recommendation performance. Despite this success, review-based recommender systems have the limitation that they do not fully consider the relevance of the review text to the target item. Specifically, the review text should reflect the user’s detailed opinion about the target item to extract detailed preference information. Meanwhile, the review content must be directly related to the target item to extract the customer’s specific preferences for the item. However, previous studies have overlooked both of these aspects. Therefore, it is necessary to build a recommendation model that considers the relevance of the review content to the target item. To address this issue, this study proposes a novel recommendation model that accurately estimates users’ preferences by carefully considering the relevance of the review content to the items. The proposed model effectively extracts feature representations from the text using bidirectional encoder representations from a transformer and obtains fused features by considering the importance of features through the attention mechanism. To evaluate the performance of the model, we used a publicly accessible dataset of reviews from Amazon.com and compared it to various baseline models. The experimental results demonstrated better recommendation performance of the proposed model compared to other baseline models.

Porous-DeepONet: Learning the Solution Operators of Parametric Reactive Transport Equations in Porous Media

Reactive transport equations in porous media are critical in various scientific and engineering disciplines, but solving these equations can be computationally expensive when exploring different scenarios, such as varying porous structures and initial or boundary conditions. The deep operator network (DeepONet) has emerged as a popular deep learning framework for solving parametric partial differential equations. However, applying the DeepONet to porous media presents significant challenges due to its limited capability to extract representative features from intricate structures. To address this issue, we propose the Porous-DeepONet, a simple yet highly effective extension of the DeepONet framework that leverages convolutional neural networks (CNNs) to learn the solution operators of parametric reactive transport equations in porous media. By incorporating CNNs, we can effectively capture the intricate features of porous media, enabling accurate and efficient learning of the solution operators. We demonstrate the effectiveness of the Porous-DeepONet in accurately and rapidly learning the solution operators of parametric reactive transport equations with various boundary conditions, multiple phases, and multi-physical fields through five examples. This approach offers significant computational savings, potentially reducing the computation time by 50–1000 times compared with the finite-element method. Our work may provide a robust alternative for solving parametric reactive transport equations in porous media, paving the way for exploring complex phenomena in porous media.

Deep Learning for Skeleton-Based Human Activity Segmentation: An Autoencoder Approach

Automatic segmentation is essential for enhancing human activity recognition, especially given the limitations of publicly available datasets that often lack diversity in daily activities. This study introduces a novel segmentation method that utilizes skeleton data for a more accurate and efficient analysis of human actions. By employing an autoencoder, this method extracts representative features and reconstructs the dataset, using the discrepancies between the original and reconstructed data to establish a segmentation threshold. This innovative approach allows for the automatic segmentation of activity datasets into distinct segments. Rigorous evaluations against ground truth across three publicly available datasets demonstrate the method’s effectiveness, achieving impressive average annotation error, precision, recall, and F1-score values of 3.6, 90%, 87%, and 88%, respectively. This illustrates the robustness of the proposed method in accurately identifying change points and segmenting continuous skeleton-based activities as compared to two other state-of-the-art techniques: one based on deep learning and another using the classical time-series segmentation algorithm. Additionally, the dynamic thresholding mechanism enhances the adaptability of the segmentation process to different activity dynamics improving overall segmentation accuracy. This performance highlights the potential of the proposed method to significantly advance the field of human activity recognition by improving the accuracy and efficiency of identifying and categorizing human movements.

Clinical knowledge-based ECG abnormalities detection using dual-view CNN-Transformer and external attention mechanism

Background:Automatic abnormalities detection based on Electrocardiogram (ECG) contributes greatly to early prevention, computer aided diagnosis, and dynamic analysis of cardiovascular diseases. In order to achieve cardiologist-level performance, deep neural networks have been widely utilized to extract abstract feature representations. However, the mechanical stacking of numerous computationally intensive operations makes traditional deep neural networks suffer from inadequate learning, poor interpretability, and high complexity. Method:To address these limitations, a clinical knowledge-based ECG abnormalities detection model using dual-view CNN-Transformer and external attention mechanism is proposed by mimicking the diagnosis of the clinicians. Considering the clinical knowledge that both the detailed waveform changes within a single heartbeat and the global changes throughout the entire recording have complementary roles in abnormalities detection, we presented a dual-view CNN-Transformer to extract and fuse spatial–temporal features from different views. In addition, the locations of the ECG where abnormalities occur provide more information than other areas. Therefore, two external attention mechanisms are designed and added to the corresponding views to help the network learn efficiently. Results:Experiment results on the 9-class dataset show that the proposed model achieves an average F1-score of 0.854±0.01 with a higher interpretability and a lower complexity, outperforming the state-of-the-art model. Conclusions:Combining all these excellent features, this study provides a credible solution for automatic ECG abnormalities detection.

Remote Sensing Image Harmonization Method for Fine-Grained Ship Classification

Target recognition and fine-grained ship classification in remote sensing face challenges of high inter-class similarity and sample scarcity. A transfer fusion-based ship image harmonization algorithm is proposed to overcome these challenges. This algorithm designs a feature transfer fusion strategy based on the combination of a region-aware instantiation and attention mechanism. Adversarial learning is implemented through an image harmony generator and discriminator module to generate realistic remote sensing ship harmony images. Furthermore, the domain encoder and domain discriminator modules are responsible for extracting feature representations of the foreground and background, and further align the ship foreground with remote sensing ocean background features through feature discrimination. Compared with other advanced image conversion techniques, our algorithm delivers more realistic visuals, improving classification accuracy for six ship types by 3% and twelve types by 2.94%, outperforming Sim2RealNet. Finally, a mixed dataset containing data augmentation and harmonizing samples and real data was proposed for the fine-grained classification task of remote sensing ships. Evaluation experiments were conducted on eight typical fine-grained classification algorithms, and the accuracy of the fine-grained classification for all categories of ships was analyzed. The experimental results show that the mixed dataset proposed in this paper effectively alleviates the long-tail problem in real datasets, and the proposed remote sensing ship data augmentation framework performs better than state-of-the-art data augmentation methods in fine-grained ship classification tasks.

Dual-extraction modeling: A multi-modal deep-learning architecture for phenotypic prediction and functional gene mining of complex traits

Despite considerable advances in extracting crucial insights from bio-omics data to unravel the intricate mechanisms underlying complex traits, the absence of a universal multi-modal computational tool with robust interpretability for accurate phenotype prediction and identification of trait-associated genes remains a challenge. This study introduces the dual-extraction modeling (DEM) approach, a multi-modal deep-learning architecture designed to extract representative features from heterogeneous omics datasets, enabling the prediction of complex trait phenotypes. Through comprehensive benchmarking experiments, we demonstrate the efficacy of DEM in classification and regression prediction of complex traits. DEM consistently exhibits superior accuracy, robustness, generalizability, and flexibility. Notably, we establish its effectiveness in predicting pleiotropic genes that influence both flowering time and rosette leaf number, underscoring its commendable interpretability. In addition, we have developed user-friendly software to facilitate seamless utilization of DEM’s functions. In summary, this study presents a state-of-the-art approach with the ability to effectively predict qualitative and quantitative traits and identify functional genes, confirming its potential as a valuable tool for exploring the genetic basis of complex traits.

CVAE-DF: A hybrid deep learning framework for fertilization status detection of pre-incubation duck eggs based on VIS/NIR spectroscopy

Unfertilized duck eggs not removed prior to incubation will deteriorate quickly, posing a risk of contaminating the normally fertilized duck eggs. Thus, detecting the fertilization status of breeding duck eggs as early as possible is a meaningful and challenging task. Most existing work usually focus on the characteristics of chicken eggs during mid-term hatching. However, little attention has been paid to the detection for duck eggs prior to incubation. In this paper, we present a novel hybrid deep learning detection framework for the fertilization status of pre-incubation duck eggs, termed CVAE-DF, based on visible/near-infrared (VIS/NIR) transmittance spectroscopy. The framework comprises the encoder of a convolutional variational autoencoder (CVAE) and an improved deep forest (DF) model. More specifically, we first collected transmittance spectral data (400–1000 nm) of 255 duck eggs before hatching. The multiplicative scatter correction (MSC) method was then used to eliminate noise and extraneous information of the raw spectral data. Two efficient data augmentation methods were adopted to provide sufficient data. After that, CVAE was applied to extract representative features and reduce the feature dimension for the detection task. Finally, an improved DF model was employed to build the classification model on the enhanced feature set. The CVAE-DF model achieved an overall accuracy of 95.94 % on the test dataset. These experimental results in terms of four metrics demonstrate that our CVAE-DF method outperforms the traditional methods by a significant margin. Furthermore, the results also indicate that CVAE holds great promise as a novel feature extraction method for the VIS/NIR spectral analysis of other agricultural products. It is extremely beneficial to practical engineering.

Multichannel Multimodal Emotion Analysis of Cross-Modal Feedback Interactions Based on Knowledge Graph

Multimodal sentiment analysis is a downstream branch task of sentiment analysis with high attention at present. Previous work in multimodal sentiment analysis have focused on the representation and fusion of modalities, capturing the underlying semantic relationships between modalities by considering contextual information. While this approach is feasible for simple contextual comments, more complex comments require the integration of external knowledge to obtain more accurate sentiment information. However, incorporating external knowledge into sentiment analysis to enhance information complementarity has not been thoroughly investigated. To address this, we propose a multichannel cross-modal feedback interaction model that incorporates the knowledge graph into multimodal sentiment analysis. Our proposed model consists of two main components: the cross-modal feedback recurrent interaction module and the external knowledge module for capturing latent information. The cross-modal interaction employs a self-feedback mechanism during network training, extracting feature representations of each modality and using these representations to mask sensory inputs, allowing the model to perform feedback-based feature masking. The external knowledge graph captures potential semantic information representations in the textual data through knowledge graph embedding. Finally, a global feature fusion module is employed for multichannel multimodal information integration. On two publicly available datasets, our method demonstrates good performance in terms of accuracy and F1 scores, compared to state-of-the-art models and several baselines.

Wave-CapNet: A Wavelet Neuron-based Wi-Fi Sensing Model for Human Identification

Gait is regarded as a unique feature for identifying people, and gait recognition is the basis of various customized services of the IoT. Unlike traditional techniques for identifying people, the Wi-Fi-based technique is unconstrained by illumination conditions and such that it eliminates the need for dense, specialized sensors and wearable devices. Although deep learning-based sensing models are conducive to the development of Wi-Fi-based identification, the latter technique relies on a large amount of data and requires a long training time, where this limits the scope of its use for identifying people. In this study, we propose a Wi-Fi sensing model called Wave-CapNet for human identification. We use data processing to eliminate errors in the raw data so that the model can extract the characteristics in channel state information (CSI). We also design a dedicated adaptive wavelet neural network to extract representative features from Wi-Fi signals with only a few epochs of training and a small number of parameters. Experiments show that it can identify human gait with an average accuracy of 99%. Moreover, it can achieve an average accuracy of 95% by using only 10% of the data and fewer than five epochs and outperforms state-of-the-art (SOTA) methods.

In situ monitoring with melt pool data based on multi-signal fusion method in laser powder bed fusion

In situ monitoring technologies for additive manufacturing are severely affected by signal-to-defect spatial registration, and a single monitoring solution that captures a specific process signature cannot identify defects with sufficient accuracy in industrial applications. In this paper, a data partitioning method was proposed to reduce registration errors and facilitate the extraction of defect-related representative features from data segments. Then, three multi-signal fusion methods based on convolutional neural networks were established from feature-level fusion, decision-level fusion, and data-level fusion, which enhanced the signal-to-defect correlation by means of complementary sensor information. The results showed that the feature-level fusion obtained the optimal classification performance compared with the other two fusion strategies. Among them, the classification accuracies of 100-, 150-, 200-, 300-, 400-, and 500-μm defects reached 74.41%, 89.84%, 90.82%, 93.95%, 97.85%, and 98.83%, respectively. The proposed data fusion approach outperformed the individual signal-based models for quality classification.

Consistency-constrained RGB-T crowd counting via mutual information maximization

The incorporation of thermal imaging data in RGB-T images has demonstrated its usefulness in cross-modal crowd counting by offering complementary information to RGB representations. Despite achieving satisfactory results in RGB-T crowd counting, many existing methods still face two significant limitations: (1) The oversight of the heterogeneous gap between modalities complicates the effective integration of multimodal features. (2) The absence of mining consistency hinders the full exploitation of the unique complementary strengths inherent in each modality. To this end, we present C4-MIM, a novel Consistency-constrained RGB-T Crowd Counting approach via Mutual Information Maximization. It effectively leverages multimodal information by learning the consistency between the RGB and thermal modalities, thereby enhancing the performance of cross-modal counting. Specifically, we first advocate extracting feature representations of different modalities in a shared encoder to moderate the heterogeneous gap since they obey the identical coding rules with shared parameters. Then, we intend to mine the consistent information of different modalities to better learn conducive information and improve the performance of feature representations. To this end, we formulate the complementarity of multimodality representations as a mutual information maximization regularizer to maximize the consistent information of different modalities, in which the consistency would be maximally attained before combining the multimodal information. Finally, we simply aggregate the feature representations of the different modalities and send them into a regressor to output the density maps. The proposed approach can be implemented by arbitrary backbone networks and is quite robust in the face of single modality unavailable or serious compromised. Extensively experiments have been conducted on the RGBT-CC and DroneRGBT benchmarks to evaluate the effectiveness and robustness of the proposed approach, demonstrating its superior performance compared to the SOTA approaches.

Few-Shot Learning from Augmented Label-Uncertain Queries in Bongard-HOI

Detecting human-object interactions (HOI) in a few-shot setting remains a challenge. Existing meta-learning methods struggle to extract representative features for classification due to the limited data, while existing few-shot HOI models rely on HOI text labels for classification. Moreover, some query images may display visual similarity to those outside their class, such as similar backgrounds between different HOI classes. This makes learning more challenging, especially with limited samples. Bongard-HOI epitomizes this HOI few-shot problem, making it the benchmark we focus on in this paper. In our proposed method, we introduce novel label-uncertain query augmentation techniques to enhance the diversity of the query inputs, aiming to distinguish the positive HOI class from the negative ones. As these augmented inputs may or may not have the same class label as the original inputs, their class label is unknown. Those belonging to a different class become hard samples due to their visual similarity to the original ones. Additionally, we introduce a novel pseudo-label generation technique that enables a mean teacher model to learn from the augmented label-uncertain inputs. We propose to augment the negative support set for the student model to enrich the semantic information, fostering diversity that challenges and enhances the student’s learning. Experimental results demonstrate that our method sets a new state-of-the-art (SOTA) performance by achieving 68.74% accuracy on the Bongard-HOI benchmark, a significant improvement over the existing SOTA of 66.59%. In our evaluation on HICO-FS, a more general few-shot recognition dataset, our method achieves 73.27% accuracy, outperforming the previous SOTA of 71.20% in the 5- way 5-shot task.

Interpretable medical deep framework by logits-constraint attention guiding graph-based multi-scale fusion for Alzheimer’s disease analysis

Deep learning using structural MRI has been widely applied to early diagnosis study of Alzheimer’s disease. Among existing methods, attention-based 3D subject-level methods can not only provide diagnosis results but also interpret the significant brain regions, thereby attracting considerable attention. However, the performance of previous attention-based methods might be still restricted by: (i) the gap between attention scores and semantic significant regions; (ii) using only single-scale features or simply fusing multi-scale information by addition or concatenation for classification decision-making. To overcome these two issues, we propose an innovative dual-branch model called LA-GMF, which consists of two major modules: logits-constraint attention (LA) and graph-based multi-scale fusion (GMF). The LA module is designed to guide the model to focus on key areas to enhance the diagnostic performance of local lesions, by reducing the inconsistency between attention scores and class prediction probabilities. Meanwhile, by combining the graph neural network and the self-attention mechanism, the GMF module not only introduces the interaction between patches, but also explores the correlation and complementarity between features at different scales, thereby extracting feature representations more comprehensively. Experiments on the popular ADNI and AIBL datasets validate the potential of our model in boosting early AD diagnosis accuracy. Additionally, our interpretation experiments demonstrate the superior interpretability performance of the proposed method over recent state-of-the-art attention-based methods. Our source codes are released at: https://github.com/nollexu/LA-GMF.

Filtering and regret network for spacecraft component segmentation based on gray images and depth maps

Identifying and segmenting spacecraft components is vital in many on-orbit space missions, such as on-orbit maintenance and component recovery. Integrating depth maps with visual images has been proven effective in improving segmentation accuracy. However, existing methods ignore the noise and fallacy in collected depth maps, which interfere with the network to extract representative features, decreasing the final segmentation accuracy. To this end, this paper proposes a Filtering and Regret Network (FRNet) for spacecraft component segmentation. The FRNet incorporates filtering and regret mechanisms to suppress the abnormal depth response in shallow layers and selectively reuses the filtered cues in deep layers, avoiding the detrimental effects of low-quality depth information while preserving the semantic context inherent in depth maps. Furthermore, a two-stage feature fusion module is proposed, which involves information interaction and aggregation. This module effectively explores the feature correlation and unifies the multi-modal features into a comprehensive representation. Finally, a large-scale spacecraft component recognition dataset is constructed for training and evaluating spacecraft component segmentation algorithms. Experimental results demonstrate that the FRNet achieves a state-of-the-art performance with a mean Intersection Over Union (mIOU) of 84.13% and an average inference time of 133.2 ms when tested on an NVIDIA RTX 2080 SUPER GPU.

Encoder-Decoder Contrast for Unsupervised Anomaly Detection in Medical Images.

Unsupervised anomaly detection (UAD) aims to recognize anomalous images based on the training set that contains only normal images. In medical image analysis, UAD benefits from leveraging the easily obtained normal (healthy) images, avoiding the costly collecting and labeling of anomalous (unhealthy) images. Most advanced UAD methods rely on frozen encoder networks pre-trained using ImageNet for extracting feature representations. However, the features extracted from the frozen encoders that are borrowed from natural image domains coincide little with the features required in the target medical image domain. Moreover, optimizing encoders usually causes pattern collapse in UAD. In this paper, we propose a novel UAD method, namely Encoder-Decoder Contrast (EDC), which optimizes the entire network to reduce biases towards pre-trained image domain and orient the network in the target medical domain. We start from feature reconstruction approach that detects anomalies from reconstruction errors. Essentially, a contrastive learning paradigm is introduced to tackle the problem of pattern collapsing while optimizing the encoder and the reconstruction decoder simultaneously. In addition, to prevent instability and further improve performances, we propose to bring globality into the contrastive objective function. Extensive experiments are conducted across four medical image modalities including optical coherence tomography, color fundus image, brain MRI, and skin lesion image, where our method outperforms all current state-of-the-art UAD methods. Code is available at: https://github.com/guojiajeremy/EDC.