CGFTrans: Cross-Modal Global Feature Fusion Transformer for Medical Report Generation.

Medical report generation, as a cross-modal automatic text generation task, can be highly significant both in research and clinical fields. The core is to generate diagnosis reports in clinical language from medical images. However, several limitations persist, including a lack of global information, inadequate cross-modal fusion capabilities, and high computational demands. To address these issues, we propose cross-modal global feature fusion Transformer (CGFTrans) to extract global information meanwhile reduce computational strain. Firstly, we introduce mesh recurrent network to capture inter-layer information at different levels to address the absence of global features. Then, we design feature fusion decoder and define 'mid-fusion' strategy to separately fuse visual and global features with medical report embeddings, which enhances the ability of the cross-modal joint learning. Finally, we integrate shifted window attention into Transformer encoder to alleviate computational pressure and capture pathological information at multiple scales. Extensive experiments conducted on three datasets demonstrate that the proposed method achieves average increments of 2.9%, 1.5%, and 0.7% in terms of the BLEU-1, METEOR and ROUGE-L metrics, respectively. Besides, it achieves average increments -22.4% and 17.3% training time and images throughput, respectively.

Cardiocerebrovascular benefits of early rhythm control in patients with atrial fibrillation detected after stroke: a systematic review and meta-analysis.

Objective: This study aimed to evaluate the impact of early rhythm control (ERC) on the occurrence of cardiocerebrovascular events in patients diagnosed with atrial fibrillation detected after stroke (AFDAS). Methods: A systematic search was conducted across nine databases from inception to October 15, 2023 to identify clinical trials comparing ERC with usual care interventions in AFDAS patients. The primary outcome assessed was recurrent stroke, with secondary outcomes including all-cause mortality, adverse events related to arrhythmias, and dementia. Results: Analysis of five studies, consisting of two randomized clinical trials (RCTs) involving 490 patients and three cohort studies involving 95,019 patients, revealed a reduced rate of recurrent stroke [odds ratio (OR) = 0.30, 95% confidence interval (CI) 0.11-0.80, P = 0.016 in RCTs; OR = 0.64, 95% CI 0.61-0.68, P < 0.00001 in cohort studies] and all-cause mortality (hazards ratio = 0.94, 95% CI 0.90-0.98, P = 0.005 in cohort studies) in the ERC group compared to the usual care group. In addition, ERC was associated with superior outcomes in terms of dementia. Conclusions: Patients with AFDAS who underwent ERC treatment exhibited a decreased risk of cardiocerebrovascular events compared to those receiving usual care. These results support the potential benefits of implementing an ERC strategy for this specific patient population. Systematic Review Registration: https://www.crd.york.ac.uk/PROSPERO/, Identifier [CRD42023465994].

LRFNet: A real-time medical image fusion method guided by detail information.

Multimodal medical image fusion (MMIF) technology plays a crucial role in medical diagnosis and treatment by integrating different images to obtain fusion images with comprehensive information. Deep learning-based fusion methods have demonstrated superior performance, but some of them still encounter challenges such as imbalanced retention of color and texture information and low fusion efficiency. To alleviate the above issues, this paper presents a real-time MMIF method, called a lightweight residual fusion network. First, a feature extraction framework with three branches is designed. Two independent branches are used to fully extract brightness and texture information. The fusion branch enables different modal information to be interactively fused at a shallow level, thereby better retaining brightness and texture information. Furthermore, a lightweight residual unit is designed to replace the conventional residual convolution in the model, thereby improving the fusion efficiency and reducing the overall model size by approximately 5 times. Finally, considering that the high-frequency image decomposed by the wavelet transform contains abundant edge and texture information, an adaptive strategy is proposed for assigning weights to the loss function based on the information content in the high-frequency image. This strategy effectively guides the model toward preserving intricate details. The experimental results on MRI and functional images demonstrate that the proposed method exhibits superior fusion performance and efficiency compared to alternative approaches. The code of LRFNet is available at https://github.com/HeDan-11/LRFNet.

Identifying atomically thin isolated-band channels for intrinsic steep-slope transistors by high-throughput study.

Developing low-power FETs holds significant importance in advancing logic circuits, especially as the feature size of MOSFETs approaches sub-10 nanometers. However, this has been restricted by the thermionic limitation of SS, which is limited to 60 mV per decade at room temperature. Herein, we proposed a strategy that utilizes 2D semiconductors with an isolated-band feature as channels to realize sub-thermionic SS in MOSFETs. Through high-throughput calculations, we established a guiding principle that combines the atomic structure and orbital interaction to identify their sub-thermionic transport potential. This guides us to screen 192 candidates from the 2D material database comprising 1608 systems. Additionally, the physical relationship between the sub-thermionic transport performances and electronic structures is further revealed, which enables us to predict 15 systems with promising device performances for low-power applications with supply voltage below 0.5 V. This work opens a new way for the low-power electronics based on 2D materials and would inspire extensive interests in the experimental exploration of intrinsic steep-slope MOSFETs.

Multi-scale adversarial learning with difficult region supervision learning models for primary tumor segmentation.

Objective.Recently, deep learning techniques have found extensive application in accurate and automated segmentation of tumor regions. However, owing to the variety of tumor shapes, complex types, and unpredictability of spatial distribution, tumor segmentation still faces major challenges. Taking cues from the deep supervision and adversarial learning, we have devised a cascade-based methodology incorporating multi-scale adversarial learning and difficult-region supervision learning in this study to tackle these challenges.Approach.Overall, the method adheres to a coarse-to-fine strategy, first roughly locating the target region, and then refining the target object with multi-stage cascaded binary segmentation which converts complex multi-class segmentation problems into multiple simpler binary segmentation problems. In addition, a multi-scale adversarial learning difficult supervised UNet (MSALDS-UNet) is proposed as our model for fine-segmentation, which applies multiple discriminators along the decoding path of the segmentation network to implement multi-scale adversarial learning, thereby enhancing the accuracy of network segmentation. Meanwhile, in MSALDS-UNet, we introduce a difficult region supervision loss to effectively utilize structural information for segmenting difficult-to-distinguish areas, such as blurry boundary areas.Main results.A thorough validation of three independent public databases (KiTS21, MSD's Brain and Pancreas datasets) shows that our model achieves satisfactory results for tumor segmentation in terms of key evaluation metrics including dice similarity coefficient, Jaccard similarity coefficient, and HD95.Significance.This paper introduces a cascade approach that combines multi-scale adversarial learning and difficult supervision to achieve precise tumor segmentation. It confirms that the combination can improve the segmentation performance, especially for small objects (our codes are publicly availabled onhttps://zhengshenhai.github.io/).

CTNet: Contrastive Transformer Network for Polyp Segmentation.

Segmenting polyps from colonoscopy images is very important in clinical practice since it provides valuable information for colorectal cancer. However, polyp segmentation remains a challenging task as polyps have camouflage properties and vary greatly in size. Although many polyp segmentation methods have been recently proposed and produced remarkable results, most of them cannot yield stable results due to the lack of features with distinguishing properties and those with high-level semantic details. Therefore, we proposed a novel polyp segmentation framework called contrastive Transformer network (CTNet), with three key components of contrastive Transformer backbone, self-multiscale interaction module (SMIM), and collection information module (CIM), which has excellent learning and generalization abilities. The long-range dependence and highly structured feature map space obtained by CTNet through contrastive Transformer can effectively localize polyps with camouflage properties. CTNet benefits from the multiscale information and high-resolution feature maps with high-level semantic obtained by SMIM and CIM, respectively, and thus can obtain accurate segmentation results for polyps of different sizes. Without bells and whistles, CTNet yields significant gains of 2.3%, 3.7%, 3.7%, 18.2%, and 10.1% over classical method PraNet on Kvasir-SEG, CVC-ClinicDB, Endoscene, ETIS-LaribPolypDB, and CVC-ColonDB respectively. In addition, CTNet has advantages in camouflaged object detection and defect detection. The code is available at https://github.com/Fhujinwu/CTNet.

An elastic competitive and discriminative collaborative representation method for image classification.

Collaborative representation-based (CR) methods have become prevalent for pattern classification tasks, achieving formidable performance. Theoretically, we expect the learned class-specific representation of the correct class to be discriminative against others, with the representation of the correct class contributing dominantly in CR. However, most existing CR methods focus on improving discrimination while having a limited impact on enhancing the representation contribution of the correct category. In this work, we propose a novel CR approach for image classification called the elastic competitive and discriminative collaborative representation-based classifier (ECDCRC) to simultaneously strengthen representation contribution and discrimination of the correct class. The ECDCRC objective function penalizes two key terms by fully incorporating label information. The competitive term integrates the nearest subspace representation with corresponding elastic factors into the model, allowing each class to have varying competition intensities based on similarity with the query sample. This enhances the representation contribution of the correct class in CR. To further improve discrimination, the discriminative term introduces an elastic factor as a weight in the model to represent the gap between the query sample and the representation of each class. Moreover, instead of focusing on representation coefficients, the designed ECDCRC weights associated with representation components directly relate to the representation of each class, enabling more direct and precise discrimination improvement. Concurrently, sparsity is also enhanced through the two terms, further boosting model performance. Additionally, we propose a robust ECDCRC (R-ECDCRC) to handle image classification with noise. Extensive experiments on seven public databases demonstrate the proposed method's superior performance over related state-of-the-art CR methods.

Shape-Scale Co-Awareness Network for 3D Brain Tumor Segmentation.

The accurate segmentation of brain tumor is significant in clinical practice. Convolutional Neural Network (CNN)-based methods have made great progress in brain tumor segmentation due to powerful local modeling ability. However, brain tumors are frequently pattern-agnostic, i.e. variable in shape, size and location, which can not be effectively matched by traditional CNN-based methods with local and regular receptive fields. To address the above issues, we propose a shape-scale co-awareness network (S2CA-Net) for brain tumor segmentation, which can efficiently learn shape-aware and scale-aware features simultaneously to enhance pattern-agnostic representations. Primarily, three key components are proposed to accomplish the co-awareness of shape and scale. The Local-Global Scale Mixer (LGSM) decouples the extraction of local and global context by adopting the CNN-Former parallel structure, which contributes to obtaining finer hierarchical features. The Multi-level Context Aggregator (MCA) enriches the scale diversity of input patches by modeling global features across multiple receptive fields. The Multi-Scale Attentive Deformable Convolution (MS-ADC) learns the target deformation based on the multiscale inputs, which motivates the network to enforce feature constraints both in terms of scale and shape for optimal feature matching. Overall, LGSM and MCA focus on enhancing the scale-awareness of the network to cope with the size and location variations, while MS-ADC focuses on capturing deformation information for optimal shape matching. Finally, their effective integration prompts the network to perceive variations in shape and scale simultaneously, which can robustly tackle the variations in patterns of brain tumors. The experimental results on BraTS 2019, BraTS 2020, MSD BTS Task and BraTS2023-MEN show that S2CA-Net has superior overall performance in accuracy and efficiency compared to other state-of-the-art methods. Code: https://github.com/jiangyu945/S2CA-Net.

Two-dimensional semiconductor transistors and integrated circuits for advanced technology nodes.

This Perspective aims to provide a concise survey of current progress and outlook future directions in high-performance transistors and integrated circuits (ICs) based on 2D semiconductors.

YTHDF3 modulates the Cbln1 level by recruiting BTG2 and is implicated in the impaired cognition of prenatal hypoxia offspring.

The "Fetal Origins of Adult Disease (FOAD)" hypothesis holds that adverse factors during pregnancy can increase the risk of chronic diseases in offspring. Here, we investigated the effects of prenatal hypoxia (PH) on brain structure and function in adult offspring and explored the role of the N6-methyladenosine (m6A) pathway. The results suggest that abnormal cognition in PH offspring may be related to the dysregulation of the m6A pathway, specifically increased levels of YTHDF3 in the hippocampus. YTHDF3 interacts with BTG2 and is involved in the decay of Cbln1 mRNA, leading to the down-regulation of Cbln1 expression. Deficiency of Cbln1 may contribute to abnormal synaptic function, which in turn causes cognitive impairment in PH offspring. This study provides a scientific clues for understanding the mechanisms of impaired cognition in PH offspring and provides a theoretical basis for the treatment of cognitive impairment in offspring exposed to PH.

CIFG-Net: Cross-level information fusion and guidance network for Polyp Segmentation.

Colorectal cancer is a common malignant tumor of the digestive tract. Most colorectal cancer is caused by colorectal polyp lesions. Timely detection and removal of colorectal polyps can substantially reduce the incidence of colorectal cancer. Accurate polyp segmentation can provide important polyp information that can aid in the early diagnosis and treatment of colorectal cancer. However, polyps of the same type can vary in texture, color, and even size. Furthermore, some polyps are similar in colour to the surrounding healthy tissue, which makes the boundary between the polyp and the surrounding area unclear. In order to overcome the issues of inaccurate polyp localization and unclear boundary segmentation, we propose a polyp segmentation network based on cross-level information fusion and guidance. We use a Transformer encoder to extract a more robust feature representation. In addition, to refine the processing of feature information from encoders, we propose the edge feature processing module (EFPM) and the cross-level information processing module (CIPM). EFPM is used to focus on the boundary information in polyp features. After processing each feature, EFPM can obtain clear and accurate polyp boundary features, which can mitigate unclear boundary segmentation. CIPM is used to aggregate and process multi-scale features transmitted by various encoder layers and to solve the problem of inaccurate polyp location by using multi-level features to obtain the location information of polyps. In order to better use the processed features to optimise our segmentation effect, we also propose an information guidance module (IGM) to integrate the processed features of EFPM and CIPM to obtain accurate positioning and segmentation of polyps. Through experiments on five public polyp datasets using six metrics, it was demonstrated that the proposed network has better robustness and more accurate segmentation effect. Compared with other advanced algorithms, CIFG-Net has superior performance. Code available at: https://github.com/zspnb/CIFG-Net.

A Symmetrical Siamese Network Framework With Contrastive Learning for Pose-Robust Face Recognition.

Face recognition has achieved remarkable success owing to the development of deep learning. However, most of existing face recognition models perform poorly against pose variations. We argue that, it is primarily caused by pose-based long-tailed data - imbalanced distribution of training samples between profile faces and near-frontal faces. Additionally, self-occlusion and nonlinear warping of facial textures caused by large pose variations also increase the difficulty in learning discriminative features of profile faces. In this study, we propose a novel framework called Symmetrical Siamese Network (SSN), which can simultaneously overcome the limitation of pose-based long-tailed data and pose-invariant features learning. Specifically, two sub-modules are proposed in the SSN, i.e., Feature-Consistence Learning sub-Net (FCLN) and Identity-Consistence Learning sub-Net (ICLN). For FCLN, the inputs are all face images on training dataset. Inspired by the contrastive learning, we simulate pose variations of faces and constrain the model to focus on the consistent areas between the original face image and its corresponding virtual pose face images. For ICLN, only profile images are used as inputs, and we propose to adopt Identity Consistence Loss to minimize the intra-class feature variation across different poses. The collaborative learning of two sub-modules guarantees that the parameters of network are updated in a relatively equal probability between near-frontal face images and profile images, so that the pose-based long-tailed problem can be effectively addressed. The proposed SSN shows comparable results over the state-of-the-art methods on several public datasets. In this study, LightCNN is selected as the backbone of SSN, and existing popular networks also can be used into our framework for pose-robust face recognition.

Multiple herbicide resistance in a Cyperus difformis population in rice field from China.

Herbicide resistance is rapidly emerging in Cyperus difformis in rice fields across China. The response of a C. difformis population GX-35 was tested against five acetolactate synthase (ALS)-inhibiting herbicides, auxin herbicide MCPA and photosynthesis II (PSII)-inhibitor bentazone. Population GX-35 evolved multiple resistance to ALS-inhibiting herbicides (penoxsulam, bispyribac­sodium, pyrazosulfuron-ethyl, halosulfuron-methly and imazapic) and auxin herbicide MCPA, with resistance levels of 140-, 1253-, 578-, 18-, 13-, and 21-fold, respectively, compared to the susceptible population. In this population, ALS gene expression was similar to that of the susceptible population. However, an Asp376Glu mutation in ALS gene was observed, leading to reduced inhibition of in-vitro ALS activities by five ALS-inhibiting herbicides. Furthermore, CYP71D8, CYP77A3, CYP78A5 and three ABC transporter genes (cluster-14412.23067, cluster-14412.25321, and cluster-14412.24716) over-expressed in absence of penoxsulam. On the other hand, an UGT73C1 and an ABC transporter (cluster-14412.25038) were induced by penoxsulam. Additionally, both over-expression and induction were observed for CYP74, CYP71A1, UGT88A1 and an ABC transporter (cluster-14412.21723). The GX-35 population has indeed evolved multiple herbicide resistance in China. Therefore, a diverse range of weed control tactics should be implemented in rice field.

Establishing a prognostic model with ferroptosis-related long non-coding RNAs in bladder cancer.

Background: Ferroptosis is a distinct form of cell death that has the potential to supersede the drug resistance that is commonly observed with current chemotherapeutic agents. As a result, ferroptosis presents a new and innovative therapeutic pathway for cancer treatment. The current understanding regarding the expression of genes associated with ferroptosis in bladder cancer (BLCA) and their prognostic implications remains unclear. Consequently, this study aimed to examine the potential prognostic value of ferroptosis-associated long non-coding RNAs (lncRNAs) in BLCA. Methods: The Cancer Genome Atlas (TCGA) was accessed to download RNA sequencing data and clinicopathological features of BLCA while accessing the FerrDb database to download ferroptosis-associated genes. The study calculated risk scores for ferroptosis-associated lncRNAs, and subsequently divided patients with BLCA into two groups, namely high- and low-risk, on the basis of the median risk score. Moreover, Kaplan-Meier (K-M) curves, Cox regression analysis, and column plots were utilized for evaluating the risk score prognostic value. Subsequently, the involvement of ferroptosis-associated mRNA, N6-methyladenosine (m6A) mRNA status, and immune responses was investigated for BLCA prognosis. Results: Thirty-six lncRNAs were identified to be differently expressed and linked to the prognosis of BLCA. The findings from the K-M curve analysis indicated a significant association between a high-risk lncRNA profile and poor BLCA prognosis. The area under curve (AUC) value of the receiver operating characteristic (ROC) curve was 0.810. The risk assessment model exhibited superior performance in predicting prognosis for BLCA compared to conventional clinicopathological features. Conclusions: Thirty-six lncRNAs were found to be linked to ferroptosis for the prognosis of patients with BLCA, and these results may provide new insights for treating BLCA.

Autism spectrum disorder diagnosis based on deep unrolling-based spatial constraint representation.

Accurate diagnosis of autism spectrum disorder (ASD) is crucial for effective treatment and prognosis. Functional brain networks (FBNs) constructed from functional magnetic resonance imaging (fMRI) have become a popular tool for ASD diagnosis. However, existing model-driven approaches used to construct FBNs lack the ability to capture potential non-linear relationships between data and labels. Moreover, most existing studies treat the FBNs construction and disease classification as separate steps, leading to large inter-subject variability in the estimated FBNs and reducing the statistical power of subsequent group comparison. To address these limitations, we propose a new approach to FBNs construction called the deep unrolling-based spatial constraint representation (DUSCR) model and integrate it with a convolutional classifier to create an end-to-end framework for ASD recognition. Specifically, the model spatial constraint representation (SCR) is solved using a proximal gradient descent algorithm, and we unroll it into deep networks using the deep unrolling algorithm. Classification is then performed using a convolutional prototype learning model. We evaluated the effectiveness of the proposed method on the ABIDE I dataset and observed a significant improvement in model performance and classification accuracy. The resting state fMRI images are preprocessed into time series data and 3D coordinates of each region of interest. The data are fed into the DUSCR model, a model for building functional brain networks using deep learning instead of traditional models, that we propose, and then the outputs are fed into the convolutional classifier with prototype learning to determine whether the patient has ASD disease.

Association of short-term nitrogen dioxide exposure with hospitalization for urolithiasis in Xinxiang, China: a time series study.

Urolithiasis accounts for the highest incidence of all urologic-associated hospitalizations. However, few studies have explored the effect of nitrogen dioxide (NO2) on hospitalizations for urolithiasis. We included 5956 patients with urolithiasis, collected daily meteorological and air pollution data between 2016 and 2021, and analyzed the associations between air pollutants and hospitalization, length of the hospital stay, and hospitalization costs attributable to urolithiasis. NO2 exposure was associated with an increased risk of hospitalization for urinary tract stones. For each 10-µg/m3 increase and 1-day lag of NO2, the maximum daily effect on the risk of hospitalization for urolithiasis was 1.020 (95% confidence interval [CI]: 1.001-1.039), and the cumulative effect peaked on lag day 4 (relative risk [RR]: 1.061; 95% CI: 1.003-1.122). Attribution scores and quantitative analysis revealed that the mean number of hospital days and mean hospital costs were 16 days and 21,164.39 RMB, respectively. Up to 5.75% of all urolithiasis hospitalizations were estimated to be attributable to NO2, and the cost of NO2-related urolithiasis hospitalizations reached approximately 3,430,000 RMB. Stratified analysis showed that NO2 had a more sensitive impact on urolithiasis hospitalizations in women and in those aged ≥65 years. Notably, men and those younger than 65 years of age (exclude people aged 65) incurred more costs for urolithiasis hospitalizations. In the population level, the association between NO2 and risk of urolithiasis hospitalization was more pronounced during the warm season. NO2 can increase hospitalizations for urolithiasis for Xinxiang City residents, and there is a cumulative lag effect. Focusing on air pollution may have practical significance in terms of the prevention and control of urolithiasis.

Impact of pyroxasulfone on sugarcane rhizosphere microbiome and functioning during field degradation.

Pyroxasulfone (PYR) is a widely used herbicide, but its effects on non-target organisms, particularly microorganisms, are largely unknown. Herein, we investigated the effects of various doses of PYR on the sugarcane rhizosphere microbiome by using amplicon sequencing of rRNA genes and quantitative PCR techniques. Correlation analyses indicated that several bacterial phyla (Verrucomicrobia and Rhodothermaeota) and genera (Streptomyces and Ignavibacteria) strongly responded to PYR application. Additionally, we found that both bacterial diversity and composition were significantly altered after 30 days, indicating a prolonged effect of the herbicide. Moreover, co-occurrence analyses of the bacterial community showed that the network complexity was significantly decreased by PYR at day 45. Furthermore, FAPROTAX analysis suggested that some functions with implications for carbon cycling groups were significantly altered after 30 days. Overall, we provide the first indications that PYR may not pose a significant risk for altering microbial communities in the short term (less than 30 days). However, its potential negative effects on bacterial communities in the middle and late stages of degradation deserve further attention. To our knowledge, this is the first study to provide insight into the effects of PYR on the rhizosphere microbiome, providing an extended basis for future risk assessments.

MyoPS: A benchmark of myocardial pathology segmentation combining three-sequence cardiac magnetic resonance images.

Assessment of myocardial viability is essential in diagnosis and treatment management of patients suffering from myocardial infarction, and classification of pathology on the myocardium is the key to this assessment. This work defines a new task of medical image analysis, i.e., to perform myocardial pathology segmentation (MyoPS) combining three-sequence cardiac magnetic resonance (CMR) images, which was first proposed in the MyoPS challenge, in conjunction with MICCAI 2020. Note that MyoPS refers to both myocardial pathology segmentation and the challenge in this paper. The challenge provided 45 paired and pre-aligned CMR images, allowing algorithms to combine the complementary information from the three CMR sequences for pathology segmentation. In this article, we provide details of the challenge, survey the works from fifteen participants and interpret their methods according to five aspects, i.e., preprocessing, data augmentation, learning strategy, model architecture and post-processing. In addition, we analyze the results with respect to different factors, in order to examine the key obstacles and explore the potential of solutions, as well as to provide a benchmark for future research. The average Dice scores of submitted algorithms were 0.614±0.231 and 0.644±0.153 for myocardial scars and edema, respectively. We conclude that while promising results have been reported, the research is still in the early stage, and more in-depth exploration is needed before a successful application to the clinics. MyoPS data and evaluation tool continue to be publicly available upon registration via its homepage (www.sdspeople.fudan.edu.cn/zhuangxiahai/0/myops20/).

An in-memory computing architecture based on a duplex two-dimensional material structure for in situ machine learning.

The growing computational demand in artificial intelligence calls for hardware solutions that are capable of in situ machine learning, where both training and inference are performed by edge computation. This not only requires extremely energy-efficient architecture (such as in-memory computing) but also memory hardware with tunable properties to simultaneously meet the demand for training and inference. Here we report a duplex device structure based on a ferroelectric field-effect transistor and an atomically thin MoS2 channel, and realize a universal in-memory computing architecture for in situ learning. By exploiting the tunability of the ferroelectric energy landscape, the duplex building block demonstrates an overall excellent performance in endurance (>1013), retention (>10 years), speed (4.8 ns) and energy consumption (22.7 fJ bit-1 µm-2). We implemented a hardware neural network using arrays of two-transistors-one-duplex ferroelectric field-effect transistor cells and achieved 99.86% accuracy in a nonlinear localization task with in situ trained weights. Simulations show that the proposed device architecture could achieve the same level of performance as a graphics processing unit under notably improved energy efficiency. Our device core can be combined with silicon circuitry through three-dimensional heterogeneous integration to give a hardware solution towards general edge intelligence.