XAI-enabled neural network analysis of metabolite spatial distributions.

We used deep neural networks to process the mass spectrometry imaging (MSI) data of mouse muscle (young vs aged) and human cancer (tumor vs normal adjacent) tissues, with the aim of using explainable artificial intelligence (XAI) methods to rapidly identify biomarkers that can distinguish different classes of tissues, from several thousands of metabolite features. We also modified classic neural network architectures to construct a deep convolutional neural network that is more suitable for processing high-dimensional MSI data directly, instead of using dimension reduction techniques, and compared it to seven other machine learning analysis methods' performance in classification accuracy. After ascertaining the superiority of Channel-ResNet10, we used a novel channel selection-based XAI method to identify the key metabolite features that were responsible for its learning accuracy. These key metabolite biomarkers were then processed using MetaboAnalyst for pathway enrichment mapping. We found that Channel-ResNet10 was superior to seven other machine learning methods for MSI analysis, reaching > 98% accuracy in muscle aging and colorectal cancer datasets. We also used a novel channel selection-based XAI method to find that in young and aged muscle tissues, the differentially distributed metabolite biomarkers were especially enriched in the propanoate metabolism pathway, suggesting it as a novel target pathway for anti-aging therapy.

A review of deep learning in medical imaging: Imaging traits, technology trends, case studies with progress highlights, and future promises.

Since its renaissance, deep learning has been widely used in various medical imaging tasks and has achieved remarkable success in many medical imaging applications, thereby propelling us into the so-called artificial intelligence (AI) era. It is known that the success of AI is mostly attributed to the availability of big data with annotations for a single task and the advances in high performance computing. However, medical imaging presents unique challenges that confront deep learning approaches. In this survey paper, we first present traits of medical imaging, highlight both clinical needs and technical challenges in medical imaging, and describe how emerging trends in deep learning are addressing these issues. We cover the topics of network architecture, sparse and noisy labels, federating learning, interpretability, uncertainty quantification, etc. Then, we present several case studies that are commonly found in clinical practice, including digital pathology and chest, brain, cardiovascular, and abdominal imaging. Rather than presenting an exhaustive literature survey, we instead describe some prominent research highlights related to these case study applications. We conclude with a discussion and presentation of promising future directions.

Which images to label for few-shot medical image analysis?

The success of deep learning methodologies hinges upon the availability of meticulously labeled extensive datasets. However, when dealing with medical images, the annotation process for such abundant training data often necessitates the involvement of experienced radiologists, thereby consuming their limited time resources. In order to alleviate this burden, few-shot learning approaches have been developed, which manage to achieve competitive performance levels with only several labeled images. Nevertheless, a crucial yet previously overlooked problem in few-shot learning is about the selection of template images for annotation before learning, which affects the final performance. In this study, we propose a novel TEmplate Choosing Policy (TECP) that aims to identify and select "the most worthy" images for annotation, particularly within the context of multiple few-shot medical tasks, including landmark detection, anatomy detection, and anatomy segmentation. TECP is composed of four integral components: (1) Self-supervised training, which entails training a pre-existing deep model to extract salient features from radiological images; (2) Alternative proposals for localizing informative regions within the images; and (3) Representative Score Estimation, which involves the evaluation and identification of the most representative samples or templates. (4) Ranking, which rank all candidates and select one with highest representative score. The efficacy of the TECP approach is demonstrated through a series of comprehensive experiments conducted on multiple public datasets. Across all three medical tasks, the utilization of TECP yields noticeable improvements in model performance.

Learning to Sketch: A Neural Approach to Item Frequency Estimation in Streaming Data.

Recently, there has been a trend of designing neural data structures to go beyond handcrafted data structures by leveraging patterns of data distributions for better accuracy and adaptivity. Sketches are widely used data structures in real-time web analysis, network monitoring, and self-driving to estimate item frequencies of data streams within limited space. However, existing sketches have not fully exploited the patterns of the data stream distributions, making it challenging to tightly couple them with neural networks that excel at memorizing pattern information. Starting from the premise, we envision a pure neural data structure as a base sketch, which we term the meta-sketch, to reinvent the base structure of conventional sketches. The meta-sketch learns basic sketching abilities from meta-tasks constituted with synthetic datasets following Zipf distributions in the pre-training phase and can be quickly adapted to real (skewed) distributions in the adaption phase. The meta-sketch not only surpasses its competitors in sketching conventional data streams but also holds good potential in supporting more complex streaming data, such as multimedia and graph stream scenarios. Extensive experiments demonstrate the superiority of the meta-sketch and offer insights into its working mechanism.

Lung Nodule Segmentation and Uncertain Region Prediction With an Uncertainty-Aware Attention Mechanism.

Radiologists possess diverse training and clinical experiences, leading to variations in the segmentation annotations of lung nodules and resulting in segmentation uncertainty. Conventional methods typically select a single annotation as the learning target or attempt to learn a latent space comprising multiple annotations. However, these approaches fail to leverage the valuable information inherent in the consensus and disagreements among the multiple annotations. In this paper, we propose an Uncertainty-Aware Attention Mechanism (UAAM) that utilizes consensus and disagreements among multiple annotations to facilitate better segmentation. To this end, we introduce the Multi-Confidence Mask (MCM), which combines a Low-Confidence (LC) Mask and a High-Confidence (HC) Mask. The LC mask indicates regions with low segmentation confidence, where radiologists may have different segmentation choices. Following UAAM, we further design an Uncertainty-Guide Multi-Confidence Segmentation Network (UGMCS-Net), which contains three modules: a Feature Extracting Module that captures a general feature of a lung nodule, an Uncertainty-Aware Module that produces three features for the annotations' union, intersection, and annotation set, and an Intersection-Union Constraining Module that uses distances between the three features to balance the predictions of final segmentation and MCM. To comprehensively demonstrate the performance of our method, we propose a Complex-Nodule Validation on LIDC-IDRI, which tests UGMCS-Net's segmentation performance on lung nodules that are difficult to segment using common methods. Experimental results demonstrate that our method can significantly improve the segmentation performance on nodules that are difficult to segment using conventional methods.

TRCA-Net: using TRCA filters to boost the SSVEP classification with convolutional neural network.

Objective.The steady-state visual evoked potential (SSVEP)-based brain-computer interface has received extensive attention in research due to its simple system, less training data, and high information transfer rate. There are currently two prominent methods dominating the classification of SSVEP signals. One is the knowledge-based task-related component analysis (TRCA) method, whose core idea is to find the spatial filters by maximizing the inter-trial covariance. The other is the deep learning-based approach, which directly learns a classification model from data. However, how to integrate the two methods to achieve better performance has not been studied before.Approach.In this study, we develop a novel algorithm named TRCA-Net (TRCA-Net) to enhance SSVEP signal classification, which enjoys the advantages of both the knowledge-based method and the deep model. Specifically, the proposed TRCA-Net first performs TRCA to obtain spatial filters, which extract task-related components of data. Then the TRCA-filtered features from different filters are rearranged as new multi-channel signals for a deep convolutional neural network (CNN) for classification. Introducing the TRCA filters to a deep learning-based approach improves the signal-to-noise ratio of input data, hence benefiting the deep learning model.Main results.We evaluate the performance of TRCA-Net using two publicly available large-scale benchmark datasets, and the results demonstrate the effectiveness of TRCA-Net. Additionally, offline and online experiments separately testing ten and five subjects further validate the robustness of TRCA-Net. Further, we conduct ablation studies on different CNN backbones and demonstrate that our approach can be transplanted into other CNN models to boost their performance.Significance.The proposed approach is believed to have a promising potential for SSVEP classification and promote its practical applications in communication and control. The code is available athttps://github.com/Sungden/TRCA-Net.

Transforming medical imaging with Transformers? A comparative review of key properties, current progresses, and future perspectives.

Transformer, one of the latest technological advances of deep learning, has gained prevalence in natural language processing or computer vision. Since medical imaging bear some resemblance to computer vision, it is natural to inquire about the status quo of Transformers in medical imaging and ask the question: can the Transformer models transform medical imaging? In this paper, we attempt to make a response to the inquiry. After a brief introduction of the fundamentals of Transformers, especially in comparison with convolutional neural networks (CNNs), and highlighting key defining properties that characterize the Transformers, we offer a comprehensive review of the state-of-the-art Transformer-based approaches for medical imaging and exhibit current research progresses made in the areas of medical image segmentation, recognition, detection, registration, reconstruction, enhancement, etc. In particular, what distinguishes our review lies in its organization based on the Transformer's key defining properties, which are mostly derived from comparing the Transformer and CNN, and its type of architecture, which specifies the manner in which the Transformer and CNN are combined, all helping the readers to best understand the rationale behind the reviewed approaches. We conclude with discussions of future perspectives.

LE-UDA: Label-Efficient Unsupervised Domain Adaptation for Medical Image Segmentation.

While deep learning methods hitherto have achieved considerable success in medical image segmentation, they are still hampered by two limitations: (i) reliance on large-scale well-labeled datasets, which are difficult to curate due to the expert-driven and time-consuming nature of pixel-level annotations in clinical practices, and (ii) failure to generalize from one domain to another, especially when the target domain is a different modality with severe domain shifts. Recent unsupervised domain adaptation (UDA) techniques leverage abundant labeled source data together with unlabeled target data to reduce the domain gap, but these methods degrade significantly with limited source annotations. In this study, we address this underexplored UDA problem, investigating a challenging but valuable realistic scenario, where the source domain not only exhibits domain shift w.r.t. the target domain but also suffers from label scarcity. In this regard, we propose a novel and generic framework called "Label-Efficient Unsupervised Domain Adaptation" (LE-UDA). In LE-UDA, we construct self-ensembling consistency for knowledge transfer between both domains, as well as a self-ensembling adversarial learning module to achieve better feature alignment for UDA. To assess the effectiveness of our method, we conduct extensive experiments on two different tasks for cross-modality segmentation between MRI and CT images. Experimental results demonstrate that the proposed LE-UDA can efficiently leverage limited source labels to improve cross-domain segmentation performance, outperforming state-of-the-art UDA approaches in the literature.

Radiology report generation with a learned knowledge base and multi-modal alignment.

In clinics, a radiology report is crucial for guiding a patient's treatment. However, writing radiology reports is a heavy burden for radiologists. To this end, we present an automatic, multi-modal approach for report generation from a chest x-ray. Our approach, motivated by the observation that the descriptions in radiology reports are highly correlated with specific information of the x-ray images, features two distinct modules: (i) Learned knowledge base: To absorb the knowledge embedded in the radiology reports, we build a knowledge base that can automatically distill and restore medical knowledge from textual embedding without manual labor; (ii) Multi-modal alignment: to promote the semantic alignment among reports, disease labels, and images, we explicitly utilize textual embedding to guide the learning of the visual feature space. We evaluate the performance of the proposed model using metrics from both natural language generation and clinic efficacy on the public IU-Xray and MIMIC-CXR datasets. Our ablation study shows that each module contributes to improving the quality of generated reports. Furthermore, the assistance of both modules, our approach outperforms state-of-the-art methods over almost all the metrics. Code is available at https://github.com/LX-doctorAI1/M2KT.

Histopathological bladder cancer gene mutation prediction with hierarchical deep multiple-instance learning.

Gene mutation detection is usually carried out by molecular biological methods, which is expensive and has a long-time cycle. In contrast, pathological images are ubiquitous. If clinically significant gene mutations can be predicted only through pathological images, it will greatly promote the widespread use of gene mutation detection in clinical practice. However, current gene mutation prediction methods based on pathological images are ineffective because of the inability to identify mutated regions in gigapixel Whole Slide Image (WSI). To address this challenge, hereby we propose a carefully designed framework for WSI-based gene mutation prediction, which consists of three parts. (i) The first part of cancerous area segmentation, based on supervised learning, quickly filters out a large number of non-mutated regions; (ii) the second part of cancerous patch clustering, based on the representations derived from contrastive learning, ensures the comprehensiveness of patch selection; and (iii) the third part of mutation classification, based on the proposed hierarchical deep multi-instance learning method (HDMIL), ensures that sufficient patches are considered and inaccurate selections are ignored. In addition, benefiting from a two-stage attention mechanism in HDMIL, the patches that are highly correlated with gene mutations can be identified. This interpretability can help a pathologist to analyze the correlation between gene mutation and histopathological morphology. Experimental results demonstrate that the proposed gene mutation prediction framework significantly outperforms the state-of-the-art methods. In the TCGA bladder cancer dataset, five clinically relevant gene mutations are well predicted.

Multilevel effective surgical workflow recognition in robotic left lateral sectionectomy with deep learning: experimental research.

BACKGROUND: Automated surgical workflow recognition is the foundation for computational models of medical knowledge to interpret surgical procedures. The fine-grained segmentation of the surgical process and the improvement of the accuracy of surgical workflow recognition facilitate the realization of autonomous robotic surgery. This study aimed to construct a multigranularity temporal annotation dataset of the standardized robotic left lateral sectionectomy (RLLS) and develop a deep learning-based automated model for multilevel overall and effective surgical workflow recognition. METHODS: From December 2016 to May 2019, 45 cases of RLLS videos were enrolled in our dataset. All frames of RLLS videos in this study are labeled with temporal annotations. The authors defined those activities that truly contribute to the surgery as effective frames, while other activities are labeled as under-effective frames. Effective frames of all RLLS videos are annotated with three hierarchical levels of 4 steps, 12 tasks, and 26 activities. A hybrid deep learning model were used for surgical workflow recognition of steps, tasks, activities, and under-effective frames. Moreover, the authors also carried out multilevel effective surgical workflow recognition after removing under-effective frames. RESULTS: The dataset comprises 4 383 516 annotated RLLS video frames with multilevel annotation, of which 2 418 468 frames are effective. The overall accuracies of automated recognition for Steps, Tasks, Activities, and under-effective frames are 0.82, 0.80, 0.79, and 0.85, respectively, with corresponding precision values of 0.81, 0.76, 0.60, and 0.85. In multilevel effective surgical workflow recognition, the overall accuracies were increased to 0.96, 0.88, and 0.82 for Steps, Tasks, and Activities, respectively, while the precision values were increased to 0.95, 0.80, and 0.68. CONCLUSION: In this study, the authors created a dataset of 45 RLLS cases with multilevel annotations and developed a hybrid deep learning model for surgical workflow recognition. The authors demonstrated a fairly higher accuracy in multilevel effective surgical workflow recognition when under-effective frames were removed. Our research could be helpful in the development of autonomous robotic surgery.

18F-FDG-PET/CT-based deep learning model for fully automated prediction of pathological grading for pancreatic ductal adenocarcinoma before surgery.

BACKGROUND: The determination of pathological grading has a guiding significance for the treatment of pancreatic ductal adenocarcinoma (PDAC) patients. However, there is a lack of an accurate and safe method to obtain pathological grading before surgery. The aim of this study is to develop a deep learning (DL) model based on 18F-fluorodeoxyglucose-positron emission tomography/computed tomography (18F-FDG-PET/CT) for a fully automatic prediction of preoperative pathological grading of pancreatic cancer. METHODS: A total of 370 PDAC patients from January 2016 to September 2021 were collected retrospectively. All patients underwent 18F-FDG-PET/CT examination before surgery and obtained pathological results after surgery. A DL model for pancreatic cancer lesion segmentation was first developed using 100 of these cases and applied to the remaining cases to obtain lesion regions. After that, all patients were divided into training set, validation set, and test set according to the ratio of 5:1:1. A predictive model of pancreatic cancer pathological grade was developed using the features computed from the lesion regions obtained by the lesion segmentation model and key clinical characteristics of the patients. Finally, the stability of the model was verified by sevenfold cross-validation. RESULTS: The Dice score of the developed PET/CT-based tumor segmentation model for PDAC was 0.89. The area under curve (AUC) of the PET/CT-based DL model developed on the basis of the segmentation model was 0.74, with an accuracy, sensitivity, and specificity of 0.72, 0.73, and 0.72, respectively. After integrating key clinical data, the AUC of the model improved to 0.77, with its accuracy, sensitivity, and specificity boosted to 0.75, 0.77, and 0.73, respectively. CONCLUSION: To the best of our knowledge, this is the first deep learning model to end-to-end predict the pathological grading of PDAC in a fully automatic manner, which is expected to improve clinical decision-making.

Semi-Supervised CT Lesion Segmentation Using Uncertainty-Based Data Pairing and SwapMix.

Semi-supervised learning (SSL) methods show their powerful performance to deal with the issue of data shortage in the field of medical image segmentation. However, existing SSL methods still suffer from the problem of unreliable predictions on unannotated data due to the lack of manual annotations for them. In this paper, we propose an unreliability-diluted consistency training (UDiCT) mechanism to dilute the unreliability in SSL by assembling reliable annotated data into unreliable unannotated data. Specifically, we first propose an uncertainty-based data pairing module to pair annotated data with unannotated data based on a complementary uncertainty pairing rule, which avoids two hard samples being paired off. Secondly, we develop SwapMix, a mixed sample data augmentation method, to integrate annotated data into unannotated data for training our model in a low-unreliability manner. Finally, UDiCT is trained by minimizing a supervised loss and an unreliability-diluted consistency loss, which makes our model robust to diverse backgrounds. Extensive experiments on three chest CT datasets show the effectiveness of our method for semi-supervised CT lesion segmentation.

Multi-Modal Tumor Segmentation With Deformable Aggregation and Uncertain Region Inpainting.

Multi-modal tumor segmentation exploits complementary information from different modalities to help recognize tumor regions. Known multi-modal segmentation methods mainly have deficiencies in two aspects: First, the adopted multi-modal fusion strategies are built upon well-aligned input images, which are vulnerable to spatial misalignment between modalities (caused by respiratory motions, different scanning parameters, registration errors, etc). Second, the performance of known methods remains subject to the uncertainty of segmentation, which is particularly acute in tumor boundary regions. To tackle these issues, in this paper, we propose a novel multi-modal tumor segmentation method with deformable feature fusion and uncertain region refinement. Concretely, we introduce a deformable aggregation module, which integrates feature alignment and feature aggregation in an ensemble, to reduce inter-modality misalignment and make full use of cross-modal information. Moreover, we devise an uncertain region inpainting module to refine uncertain pixels using neighboring discriminative features. Experiments on two clinical multi-modal tumor datasets demonstrate that our method achieves promising tumor segmentation results and outperforms state-of-the-art methods.

Federated Transfer Learning for Low-dose PET Denoising: A Pilot Study with Simulated Heterogeneous Data.

Positron emission tomography (PET) with a reduced injection dose, i.e., low-dose PET, is an efficient way to reduce radiation dose. However, low-dose PET reconstruction suffers from a low signal-to-noise ratio (SNR), affecting diagnosis and other PET-related applications. Recently, deep learning-based PET denoising methods have demonstrated superior performance in generating high-quality reconstruction. However, these methods require a large amount of representative data for training, which can be difficult to collect and share due to medical data privacy regulations. Moreover, low-dose PET data at different institutions may use different low-dose protocols, leading to non-identical data distribution. While previous federated learning (FL) algorithms enable multi-institution collaborative training without the need of aggregating local data, it is challenging for previous methods to address the large domain shift caused by different low-dose PET settings, and the application of FL to PET is still under-explored. In this work, we propose a federated transfer learning (FTL) framework for low-dose PET denoising using heterogeneous low-dose data. Our experimental results on simulated multi-institutional data demonstrate that our method can efficiently utilize heterogeneous low-dose data without compromising data privacy for achieving superior low-dose PET denoising performance for different institutions with different low-dose settings, as compared to previous FL methods.

Rethinking Semi-Supervised Medical Image Segmentation: A Variance-Reduction Perspective.

For medical image segmentation, contrastive learning is the dominant practice to improve the quality of visual representations by contrasting semantically similar and dissimilar pairs of samples. This is enabled by the observation that without accessing ground truth labels, negative examples with truly dissimilar anatomical features, if sampled, can significantly improve the performance. In reality, however, these samples may come from similar anatomical regions and the models may struggle to distinguish the minority tail-class samples, making the tail classes more prone to misclassification, both of which typically lead to model collapse. In this paper, we propose ARCO, a semi-supervised contrastive learning (CL) framework with stratified group theory for medical image segmentation. In particular, we first propose building ARCO through the concept of variance-reduced estimation and show that certain variance-reduction techniques are particularly beneficial in pixel/voxel-level segmentation tasks with extremely limited labels. Furthermore, we theoretically prove these sampling techniques are universal in variance reduction. Finally, we experimentally validate our approaches on eight benchmarks, i.e., five 2D/3D medical and three semantic segmentation datasets, with different label settings, and our methods consistently outperform state-of-the-art semi-supervised methods. Additionally, we augment the CL frameworks with these sampling techniques and demonstrate significant gains over previous methods. We believe our work is an important step towards semi-supervised medical image segmentation by quantifying the limitation of current self-supervision objectives for accomplishing such challenging safety-critical tasks.

FedFTN: Personalized federated learning with deep feature transformation network for multi-institutional low-count PET denoising.

Low-count PET is an efficient way to reduce radiation exposure and acquisition time, but the reconstructed images often suffer from low signal-to-noise ratio (SNR), thus affecting diagnosis and other downstream tasks. Recent advances in deep learning have shown great potential in improving low-count PET image quality, but acquiring a large, centralized, and diverse dataset from multiple institutions for training a robust model is difficult due to privacy and security concerns of patient data. Moreover, low-count PET data at different institutions may have different data distribution, thus requiring personalized models. While previous federated learning (FL) algorithms enable multi-institution collaborative training without the need of aggregating local data, addressing the large domain shift in the application of multi-institutional low-count PET denoising remains a challenge and is still highly under-explored. In this work, we propose FedFTN, a personalized federated learning strategy that addresses these challenges. FedFTN uses a local deep feature transformation network (FTN) to modulate the feature outputs of a globally shared denoising network, enabling personalized low-count PET denoising for each institution. During the federated learning process, only the denoising network's weights are communicated and aggregated, while the FTN remains at the local institutions for feature transformation. We evaluated our method using a large-scale dataset of multi-institutional low-count PET imaging data from three medical centers located across three continents, and showed that FedFTN provides high-quality low-count PET images, outperforming previous baseline FL reconstruction methods across all low-count levels at all three institutions.

Learning to Localize Cross-Anatomy Landmarks in X-Ray Images with a Universal Model.

Objective and Impact Statement. In this work, we develop a universal anatomical landmark detection model which learns once from multiple datasets corresponding to different anatomical regions. Compared with the conventional model trained on a single dataset, this universal model not only is more light weighted and easier to train but also improves the accuracy of the anatomical landmark location. Introduction. The accurate and automatic localization of anatomical landmarks plays an essential role in medical image analysis. However, recent deep learning-based methods only utilize limited data from a single dataset. It is promising and desirable to build a model learned from different regions which harnesses the power of big data. Methods. Our model consists of a local network and a global network, which capture local features and global features, respectively. The local network is a fully convolutional network built up with depth-wise separable convolutions, and the global network uses dilated convolution to enlarge the receptive field to model global dependencies. Results. We evaluate our model on four 2D X-ray image datasets totaling 1710 images and 72 landmarks in four anatomical regions. Extensive experimental results show that our model improves the detection accuracy compared to the state-of-the-art methods. Conclusion. Our model makes the first attempt to train a single network on multiple datasets for landmark detection. Experimental results qualitatively and quantitatively show that our proposed model performs better than other models trained on multiple datasets and even better than models trained on a single dataset separately.

GAN-based disentanglement learning for chest X-ray rib suppression.

Clinical evidence has shown that rib-suppressed chest X-rays (CXRs) can improve the reliability of pulmonary disease diagnosis. However, previous approaches on generating rib-suppressed CXR face challenges in preserving details and eliminating rib residues. We hereby propose a GAN-based disentanglement learning framework called Rib Suppression GAN, or RSGAN, to perform rib suppression by utilizing the anatomical knowledge embedded in unpaired computed tomography (CT) images. In this approach, we employ a residual map to characterize the intensity difference between CXR and the corresponding rib-suppressed result. To predict the residual map in CXR domain, we disentangle the image into structure- and contrast-specific features and transfer the rib structural priors from digitally reconstructed radiographs (DRRs) computed by CT. Furthermore, we employ additional adaptive loss to suppress rib residue and preserve more details. We conduct extensive experiments based on 1673 CT volumes, and four benchmarking CXR datasets, totaling over 120K images, to demonstrate that (i) our proposed RSGAN achieves superior image quality compared to the state-of-the-art rib suppression methods; (ii) combining CXR with our rib-suppressed result leads to better performance in lung disease classification and tuberculosis area detection.

Knowledge matters: Chest radiology report generation with general and specific knowledge.

Automatic chest radiology report generation is critical in clinics which can relieve experienced radiologists from the heavy workload and remind inexperienced radiologists of misdiagnosis or missed diagnose. Existing approaches mainly formulate chest radiology report generation as an image captioning task and adopt the encoder-decoder framework. However, in the medical domain, such pure data-driven approaches suffer from the following problems: 1) visual and textual bias problem; 2) lack of expert knowledge. In this paper, we propose a knowledge-enhanced radiology report generation approach introduces two types of medical knowledge: 1) General knowledge, which is input independent and provides the broad knowledge for report generation; 2) Specific knowledge, which is input dependent and provides the fine-grained knowledge for chest X-ray report generation. To fully utilize both the general and specific knowledge, we also propose a knowledge-enhanced multi-head attention mechanism. By merging the visual features of the radiology image with general knowledge and specific knowledge, the proposed model can improve the quality of generated reports. The experimental results on the publicly available IU-Xray dataset show that the proposed knowledge-enhanced approach outperforms state-of-the-art methods in almost all metrics. And the results of MIMIC-CXR dataset show that the proposed knowledge-enhanced approach is on par with state-of-the-art methods. Ablation studies also demonstrate that both general and specific knowledge can help to improve the performance of chest radiology report generation.