DMGM: deformable-mechanism based cervical cancer staging via MRI multi-sequence.

Objective.This study aims to leverage a deep learning approach, specifically a deformable convolutional layer, for staging cervical cancer using multi-sequence MRI images. This is in response to the challenges doctors face in simultaneously identifying multiple sequences, a task that computer-aided diagnosis systems can potentially improve due to their vast information storage capabilities.Approach.To address the challenge of limited sample sizes, we introduce a sequence enhancement strategy to diversify samples and mitigate overfitting. We propose a novel deformable ConvLSTM module that integrates a deformable mechanism with ConvLSTM, enabling the model to adapt to data with varying structures. Furthermore, we introduce the deformable multi-sequence guidance model (DMGM) as an auxiliary diagnostic tool for cervical cancer staging.Main results.Through extensive testing, including comparative and ablation studies, we validate the effectiveness of the deformable ConvLSTM module and the DMGM. Our findings highlight the model's ability to adapt to the deformation mechanism and address the challenges in cervical cancer tumor staging, thereby overcoming the overfitting issue and ensuring the synchronization of asynchronous scan sequences. The research also utilized the multi-modal data from BraTS 2019 as an external test dataset to validate the effectiveness of the proposed methodology presented in this study.Significance.The DMGM represents the first deep learning model to analyze multiple MRI sequences for cervical cancer, demonstrating strong generalization capabilities and effective staging in small dataset scenarios. This has significant implications for both deep learning applications and medical diagnostics. The source code will be made available subsequently.

All-aqueous droplets-templated tailorable core-shell alginate microspheres for constructing vascularized intestinal mucosain vitromodels.

Recently,in vitromodels of intestinal mucosa have become important tools for drug screening and studying the physiology and pathology of the intestine. These models enable the examination of cellular behavior in diseased states or in reaction to alterations in the microenvironment, potentially serving as alternatives to animal models. One of the major challenges in constructing physiologically relevantin vitromodels of intestinal mucosa is the creation of three-dimensional microstructures that accurately mimic the integration of intestinal epithelium and vascularized stroma. Here, core-shell alginate (Alg) microspheres were generated to create the compartmentalized extracellular matrix microenvironment needed to simulate the epithelial and vascularized stromal compartments of the intestinal mucosa. We demonstrated that NIH-3T3 and human umbilical vein endothelial cells embedded in the core of the microspheres can proliferate and develop a vascular network, while human colorectal adenocarcinoma cells (Caco-2) can form an epithelial monolayer in the shell. Compared to Caco-2 monolayer encapsulated within the shell, the presence of the vascularized stroma enhances their proliferation and functionality. As such, our core-shell Alg microspheres provide a valuable method for generatingin vitromodels of vascularized intestinal mucosa with epithelial and vascularized stroma arranged in a spatially relevant manner and demonstrating near-physiological functionality.

Analysis of antiplatelet therapy adherence in patients with ischemic cerebral stroke.

BACKGROUND: The related factors affecting the adherence of ischemic cerebral stroke (ICS) patients to antiplatelet therapy have attracted much attention. METHODS: Patients with ICS (confirmed by CT or MRI) were enrolled from January 2020 to July 2021. The demographic data were retrospectively investigated and analyzed. The adherence calculation was as follows: Adherence = number of tablets taken/number of tablets needed to be taken. Adherence < 100% was defined as nonadherence. Severe nonadherence is defined as adherence ≤ 75%. RESULTS: A total of 229 patients with ICS were enrolled. We found no significant difference in the proportion of patients with nonadherence, while the proportion of severe nonadherence in the aspirin group was significantly higher (p < .001). Multivariable analysis indicated that medical insurance (odds ratio [OR] = 0.071, p < .001) and regular exercise (OR = 0.438, p = .015) were independent factors associated with adherence. In addition, only medical insurance (OR = 5.475, p < .001) and aspirin treatment (OR = 0.228, p < .001) were independent risk factors associated with severe nonadherence. We therefore constructed a nomogram plot and a model as follows: Adherence risk score = 3 × medical insurance + regular exercise. Patients were divided into low-risk and high-risk groups for adherence based on the median model score. A total of 13.3% of patients in the low-risk group were nonadherent patients compared with 53.4% in the high-risk group (p < .001). Similarly, 8.4% of patients in the low-risk group had severe nonadherence compared with 19.9% in the high-risk group (p = .022). Moreover, in low-risk patients, no significant difference was observed. In patients with high risk, aspirin-treated patients showed significantly decreased adherence compared with the other two groups. CONCLUSION: Medical insurance and regular exercise were independent factors for antiplatelet therapy adherence. For patients with high model scores, timely intervention is necessary.

An Efficient Federated Learning Framework for Machinery Fault Diagnosis With Improved Model Aggregation and Local Model Training.

Due to device operating environment limitations and data privacy protection, it is frequently difficult to obtain sufficient high-quality labeled data from devices, resulting in an insufficient generalization ability of fault diagnosis model. Therefore, a high-performance federated learning framework is proposed in this work, which makes improvements in the procedure of model aggregation and local model training. In the model aggregation of central server, an optimization aggregation strategy in which forgetting Kalman filter (FKF) is combined with cubic exponential smoothing (CES) is proposed to improve the efficiency of federated learning. In the local model training of multiclient, a deep learning network combined with multiscale convolution, attention mechanism, and multistage residual connection is proposed, which is able to fully extract multiclient data features simultaneously. Meanwhile, experiments on two machinery fault datasets show that the proposed framework is capable of achieving high accuracy and strong generalization of fault diagnosis on the premise of protecting data privacy in actual industrial situations.

Adaptive Iterative Learning Control for a Class of Nonlinear Strict-Feedback Systems With Unknown State Delays.

In this article, an adaptive iterative learning control scheme is presented for a class of nonlinear parametric strict-feedback systems with unknown state delays, aiming to achieve the point-wise tracking of desired trajectory in a finite interval. The appropriate Lyapunov-Krasovskii functions are established to compensate the influence of time-delay uncertainties on the control systems. As the main features, the proposed approach integrates the command filter into the backstepping procedure to avoid the differential explosion problem that may occur with the increase of system order, and introduces the hyperbolic tangent functions into the learning controller to handle the singularity problem thus maintaining the continuity of input signal. The results of theoretical analysis and numerical simulation demonstrate that the tracking errors at the entire period will converge to a compact set along the iteration axis. Compared with the existing works, the proposed control scheme is promising to manifest the better performance and practicability owing to the learning mechanism, the dynamic model, as well as the implementation of controller.

Phylogenomic analyses and chromosome ploidy identification reveal multiple cryptic species in Allium sikkimense complex (Amaryllidaceae).

Polyploidization is a process that typically leads to instantaneous reproductive isolation and has, therefore, been considered as one of the major evolutionary forces in the species-rich Hengduan Mountains (HM), yet this topic remains poorly studied in the region. Allium sikkimense and its relatives (about eight species) compose a natural diploid-polyploid complex with the highest diversity in the HM and adjacent areas. A combination of nuclear ribosomal DNA (nrDNA), plastome, transcriptome, and ploidy identification through chromosome counting and flow cytometry is employed to reconstruct the phylogenetic relationships in this complex and to investigate the frequency and the evolutionary significance of polyploidy in the complex. The plastome failed to resolve the phylogenetic relationships of the different species in the A. sikkimense complex, and the phylogenetic tree based on nrDNA also has limited resolution. However, our study reveals a well-resolved phylogenetic framework for species in the A. sikkimense complex using more than 1,000 orthologous genes from the transcriptome data. Previously recognized morphospecies A. sikkimense are non-monophyletic and comprise at least two independently evolved lineages (i.e., cryptic species), each forming a clade with different diploid species in this complex. The embedded pattern of octoploid A. jichouense and tetraploid A. sp. nov. within different polyploid samples of A. sikkimense supports a possible scenario of budding speciation (via niche divergence). Furthermore, our results reveal that co-occurring species in the A. sikkimense complex usually have different ploidy levels, suggesting that polyploidy is an important process for reproductive isolation of sympatric Allium species. Phylogenetic network analyses suggested that the phylogenetic relationships of the A. sikkimense complex, allowing for reticulation events, always fit the dataset better than a simple bifurcating tree. In addition, the included or exserted filaments, which have long been used to delimit species, are highly unreliable taxonomically due to their extensive parallel and convergent evolution.

Microstructure and Properties of Hollow Octet Nickel Lattice Materials.

In this study, electroless nickel plating and electrodeposition were used to deposit thin films on the polymer lattice template prepared by 3D printing, then seven Octet hollow nickel lattice materials with different structural parameters were synthesized by etching process at the expense of the polymer backbone. The microstructure and properties of the Octet structure nickel lattice were characterized by X-ray diffraction, Electron backscattering diffraction and transmission electron microscopy. According to the results, the average grain size of the electrodeposition Ni lattice material was 429 nm, and (001) weak texture was found along the direction of the film deposition. The lattice deformation mode changed with the increase of the lattice length-to-diameter ratio, and it shifted from the lattice deformation layer-by-layer and the overall deformation to the shear deformation in the 45° direction. The strength, modulus and energy absorption properties of the Octet lattice increased with the density, and they were exponentially related to density. In the relative density range of 0.7~5%, Octet hollow Ni lattices with the same density conditions but different structural parameters showed similar compressive strength and elasticity modulus; the energy absorption capacity, however, was weakened as the length-to-diameter ratio increased.

Effect of Aggregate Size on the Axial Compressive Behavior of FRP-Confined Coral Aggregate Concrete.

Using locally available raw materials for preparing concrete, such as coral reefs, seawater, and sea sand, is conducive to compensating for the shortage of construction materials used on remote islands. Jacketing fiber-reinforced polymer (FRP), as passive confinement, is a practical approach to enhance the strength, ductility, and durability of such coral aggregate concrete (CAC). Rational and economical CAC structural design requires understanding the interactions between the CAC fracture process and FRP confinement. The coral aggregate size is the critical parameter of their interaction since it affects the crack propagation of CAC and FRP confinement efficiency. This study conducted axial compression tests on FRP-confined CAC cylinders with varying coral aggregate sizes and FRP confinement levels. The test results indicate that the coral aggregate sizes affected the unconfined CAC strength. In addition, the dilation behavior of FRP-confined CAC varied with aggregate sizes, showing that CAC with smaller coral aggregate featured a more uniform hoop strain distribution and larger FRP rupture strain. These coupling effects are epitomized by the variation in the transition stress on the stress-strain curve, which makes the existing stress-strain models not applicable for FRP-confined CAC. A modified stress-strain model is subsequently proposed. Finally, the practical and environmental implications of the present study are discussed.

Efficient Axillary Lymph Node Detection Via Two-stage Spatial-information-fusion-based CNN.

BACKGROUND AND OBJECTIVE: Preoperative imaging diagnosis of axillary lymph node (ALN) metastasis is particularly important for breast cancer patients. This paper focuses on developing non-invasive and automatic schemes for accurate localization and classification (metastasis prediction) of ALN via contrast-enhanced computed tomography (CECT) image and deep learning models. METHODS: Based on a two-stage strategy, a novel detection neural network is proposed, where the convolutional block attention module is utilized to extract spacial information and the bottleneck feature fusion module is designed for feature fusion in different scales. RESULTS: Owing to the two embedded modules, the proposed convolutional neural network (CNN) model outperforms Faster R-CNN, YOLOv3, and EfficientDet in the sense that the achieved mAP is 0.454, higher than 0.247, 0.335, and 0.329, respectively. In particular, considering the function of classification only, the proposed model reaches the best performance on most indices (accuracy of 0.968, positive predictive value of 0.972, negative predictive value of 0.966, specificity of 0.983), compared with the methods that have been frequently adopted to predict ALN. In addition, the proposed CNN model has the function of locating ALN, which is lacking in existing models. CONCLUSIONS: In this paper, a supervised deep learning method is proposed to detect ALN in CECT images. The positive effect of new added modules are verified, and the benefits of spatial information in ALN detection are confirmed. Further, the two subtasks called localization and classification are evaluated separately, where the proposed model achieves the best performance on most indices. The source code mentioned in this article will be released later.

Effect of proprioceptive neuromuscular facilitation technique on the treatment of frozen shoulder: a pilot randomized controlled trial.

BACKGROUND AND OBJECTIVES: Frozen shoulder is a common painful disease of the shoulder joint characterized by structural changes in the shoulder joint, restricting both active and passive shoulder joint activities. Proprioceptive neuromuscular facilitation (PNF) effectively improved and maintained the range of motion; however, it is not clear whether it can improve the shoulder joint structure in patients with frozen shoulder. This pilot study used magnetic resonance imaging (MRI) observation to assess the improvement of the local structure of the shoulder joint upon PNF treatment to elucidate a target based on structure for the treatment of frozen shoulder. MATERIALS AND METHODS: Forty-eight patients with frozen shoulder were randomly divided into the traditional manual therapy group and the PNF technique group. Changes in the thicknesses of the coracohumeral ligament (CHL) and capsule in axillary recess (CAR) of the shoulder joint were observed via MRI upon admission and at 4 weeks after treatment. A visual analog scale (VAS) and passive shoulder range of motion (ROM) at abduction, anteflexion and external rotation position were used to evaluate the improvement of shoulder joint pain and function in the initial, mid-term, and discharge of the two groups of patients. RESULTS: The primary outcome results shown that the PNF joint mobilization significantly reduced the thickness of the CHL (p = 0.0217) and CAR (p = 0.0133). Compared with simple joint mobilization, The mid-term and discharge rehabilitation assessment results showed that PNF has a better effect on shoulder pain. At the mid-term evaluation, the ROM of the PNF group was significantly better than that of the Control group in the three directions (p < 0.05). CONCLUSION: As an adjunctive therapy, PNF can improve the shoulder joint structure of patients with frozen shoulder and is an effective treatment strategy for frozen shoulder.

Spatial Iterative Learning Control for Robotic Path Learning.

A spatial iterative learning control (sILC) method is proposed for a robot to learn a desired path in an unknown environment. When interacting with the environment, the robot initially starts with a predefined trajectory so an interaction force is generated. By assuming that the environment is subjected to fixed spatial constraints, a learning law is proposed to update the robot's reference trajectory so that a desired interaction force is achieved. Different from existing iterative learning control methods in the literature, this method does not require repeating the interaction with the environment in time, which relaxes the assumption of the environment and thus addresses the limits of the existing methods. With the rigorous convergence analysis, simulation and experimental results in two applications of surface exploration and teaching by demonstration illustrate the significance and feasibility of the proposed method.

Shenfu Injection Protects Brain Injury in Rats with Cardiac Arrest through Nogo/NgR Pathway.

The effect of Shenfu injection on brain injury after cardiac arrest (CA) and cardiopulmonary resuscitation (CPR) along with the underlying mechanism of axonal regeneration was explored. CA/CPR model in rats was established for subsequent experiments. A total of 160 rats were randomly divided into sham group, model group, conventional western medicine (CWM) group, Shenfu group, and antagonist group (n = 32 per group). After 3 hours, 24 hours, 3 days, and 7 days of drug administration, the modified Neurological Severity Score tests were performed. The ultrastructure of the brain and hippocampus was observed by electron microscopy. Real-time quantitative polymerase chain reaction (PCR), western blotting, and immunohistochemistry were used to detect Nogo receptor (NgR) expression in the hippocampus and cerebral cortex, and Nogo-NgR expression in CA/CPR model. Neurological deficits in the model group were severe at 3 hours, 24 hours, 3 days, and 7 days after the recovery of natural circulation, whereas the neurological deficits in CWM, antagonist, and Shenfu group were relatively mild. The ultrastructure of neuronal cells in Shenfu group had relatively complete cell membranes and more vesicles than those in the model group. The results of PCR and western blotting showed lower messenger ribonucleic acid and protein expression of NgR in Shenfu group than the model group and CWM group. Immunohistochemical examination indicated a reduction of Nogo-NgR expression in Shenfu group and antagonist group. Our results suggested that Shenfu injection reduced brain injury by attenuating Nogo-NgR signaling pathway and promoting axonal regeneration.

Prediction of Metastasis in the Axillary Lymph Nodes of Patients With Breast Cancer: A Radiomics Method Based on Contrast-Enhanced Computed Tomography.

BACKGROUND: The use of traditional techniques to evaluate breast cancer is restricted by the subjective nature of assessment, variation across radiologists, and limited data. Radiomics may predict axillary lymph node metastasis (ALNM) of breast cancer more accurately. PURPOSE: The aim was to evaluate the diagnostic performance of a radiomics model based on ALNs themselves that used contrast-enhanced computed tomography (CECT) to detect ALNM of breast cancer. METHODS: We retrospectively enrolled 402 patients with breast cancer confirmed by pathology from January 2016 to October 2019. Three hundred and ninety-six features were extracted for all patients from axial CECT images of 825 ALNs using Artificial Intelligent Kit software (GE Medical Systems, Version V3.1.0.R). Next, the radiomics model was trained, validated, and tested for predicting ALNM in breast cancer by using a support vector machine algorithm. Finally, the performance of the radiomics model was evaluated in terms of its classification accuracy and the value of the area under the curve (AUC). RESULTS: The radiomics model yielded the best classification accuracy of 89.1% and the highest AUC of 0.92 (95% CI: 0.91-0.93, p=0.002) for discriminating ALNM in breast cancer in the validation cohorts. In the testing cohorts, the model also demonstrated better performance, with an accuracy of 88.5% and an AUC of 0.94 (95% CI: 0.93-0.95, p=0.005) for predicting ALNM in breast cancer. CONCLUSION: The radiomics model based on CECT images can be used to predict ALNM in breast cancer and has significant potential in clinical noninvasive diagnosis and in the prediction of breast cancer metastasis.

Axillary lymph node metastasis prediction by contrast-enhanced computed tomography images for breast cancer patients based on deep learning.

When doctors use contrast-enhanced computed tomography (CECT) images to predict the metastasis of axillary lymph nodes (ALN) for breast cancer patients, the prediction performance could be degraded by subjective factors such as experience, psychological factors, and degree of fatigue. This study aims to exploit efficient deep learning schemes to predict the metastasis of ALN automatically via CECT images. A new construction called deformable sampling module (DSM) was meticulously designed as a plug-and-play sampling module in the proposed deformable attention VGG19 (DA-VGG19). A dataset of 800 samples labeled from 800 CECT images of 401 breast cancer patients retrospectively enrolled in the last three years was adopted to train, validate, and test the deep convolutional neural network models. By comparing the accuracy, positive predictive value, negative predictive value, sensitivity and specificity indices, the performance of the proposed model is analyzed in detail. The best-performing DA-VGG19 model achieved an accuracy of 0.9088, which is higher than that of other classification neural networks. As such, the proposed intelligent diagnosis algorithm can provide doctors with daily diagnostic assistance and advice and reduce the workload of doctors. The source code mentioned in this article will be released later.

LeanNet: An Efficient Convolutional Neural Network for Digital Number Recognition in Industrial Products.

The remarkable success of convolutional neural networks (CNNs) in computer vision tasks is shown in large-scale datasets and high-performance computing platforms. However, it is infeasible to deploy large CNNs on resource constrained platforms, such as embedded devices, on account of the huge overhead. To recognize the label numbers of industrial black material product and deploy deep CNNs in real-world applications, this research uses an efficient method to simultaneously (a) reduce the network model size and (b) lower the amount of calculation without compromising accuracy. More specifically, the method is implemented by pruning channels and corresponding filters that are identified as having a trivial effect on the output accuracy. In this paper, we prune VGG-16 to obtain a compact network called LeanNet, which gives a 25× reduction in model size and a 4.5× reduction in float point operations (FLOPs), while the accuracy on our dataset is close to the original accuracy by retraining the network. Besides, we also find that LeanNet could achieve better performance on reductions in model size and computation compared to some lightweight networks like MobileNet and SqueezeNet, which are widely used in engineering applications. This research has good application value in the field of industrial production.

Three Highly Sensitive and High-Throughput Serological Approaches for Detecting Dickeya dadantii in Sweet Potato.

Sweet potato stem and root rot is an important bacterial disease and often causes serious economic losses to sweet potato. Development of rapid and sensitive detection methods is crucial for diagnosis and management of this disease in field. Here, we report the production of four hybridoma cell lines (25C4, 16C10, 9B1, and 9H10) using Dickeya dadantii strain FY1710 as an immunogen. Monoclonal antibodies (MAbs) produced by these four hybridoma cell lines were highly specific and sensitive for D. dadantii detection. Indirect enzyme-linked immunosorbent assay (indirect-ELISA) results showed that the four MAbs 25C4, 16C10, 9B1, and 9H10 could detect D. dadantii in suspensions diluted to 4.89 × 104, 4.89 × 104, 9.78 × 104, and 9.78 × 104 CFU/ml, respectively. Furthermore, all four MAbs can react strongly and specifically with all four D. dadantii strains used in this study, not with the other seven tested bacterial strains. Using these four MAbs, three different serological approaches, triple-antibody sandwich enzyme-linked immunosorbent assay (TAS-ELISA), dot-ELISA, and tissue-print-ELISA, were developed for detection of D. dadantii in crude extracts prepared from field-collected sweet potato plants. Among these three methods, TAS-ELISA and dot-ELISA were used to detect D. dadantii in suspensions diluted up to 1.23 × 104 and 1.17 × 106 CFU/ml, respectively, or in sweet potato crude extracts diluted up to 1:3,840 and 1:1,920 (wt/vol, grams per milliliter), respectively. Surprisingly, both TAS-ELISA and dot-ELISA serological approaches were more sensitive than the conventional PCR. Analyses using field-collected sweet potato samples showed that the newly developed TAS-ELISA, dot-ELISA, or tissue-print-ELISA were reliable in detecting D. dadantii in sweet potato tissues. Thus, the three serological approaches were highly valuable for diagnosis of stem and root rot in sweet potato production.

Multiple Convolutional Recurrent Neural Networks for Fault Identification and Performance Degradation Evaluation of High-Speed Train Bogie.

As an important part of high-speed train (HST), the mechanical performance of bogies imposes a direct impact on the safety and reliability of HST. It is a fact that, regardless of the potential mechanical performance degradation status, most existing fault diagnosis methods focus only on the identification of bogie fault types. However, for application scenarios such as auxiliary maintenance, identifying the performance degradation of bogie is critical in determining a particular maintenance strategy. In this article, by considering the intrinsic link between fault type and performance degradation of bogie, a novel multiple convolutional recurrent neural network (M-CRNN) that consists of two CRNN frameworks is proposed for simultaneous diagnosis of fault type and performance degradation state. Specifically, the CRNN framework 1 is designed to detect the fault types of the bogie. Meanwhile, CRNN framework 2, which is formed by CRNN Framework 1 and an RNN module, is adopted to further extract the features of fault performance degradation. It is worth highlighting that M-CRNN extends the structure of traditional neural networks and makes full use of the temporal correlation of performance degradation and model fault types. The effectiveness of the proposed M-CRNN algorithm is tested via the HST model CRH380A at different running speeds, including 160, 200, and 220 km/h. The overall accuracy of M-CRNN, i.e., the product of the accuracies for identifying the fault types and evaluating the fault performance degradation, is beyond 94.6% in all cases. This clearly demonstrates the potential applicability of the proposed method for multiple fault diagnosis tasks of HST bogie system.

An Active Repetitive Learning Control Method for Lateral Suspension Systems of High-Speed Trains.

This article presents a novel perspective to improve the ride quality of high-speed trains (HSTs), namely, by virtue of the periodicity of lateral dynamics to suppress the lateral vibration of HST. To resolve the contradiction between the complex HST model and the effective controller design, a simplified three-degrees-of-freedom (3-DOF) quarter-vehicle model is first employed for controller design, while a 17-DOF full-vehicle model is built for efficiency verification, where periodic and random track irregularities are considered, respectively. An active repetitive learning control (RLC) method is proposed to achieve the periodic tracking control, where the learning convergence is proved rigorously in a Lyapunov way. The configuration of RLC-based lateral suspensions is economical in the sense that only four actuators and six sensors are needed. It is verified by simulation that, compared with the dynamic matrix controller, the proposed RLC controller has greatly reduced the lateral vibration of a vehicle body, especially the lateral acceleration in the frequency range of (0, 3] Hz to which human body is strongly sensitive.

Radiomics model of magnetic resonance imaging for predicting pathological grading and lymph node metastases of extrahepatic cholangiocarcinoma.

The aim of this study was to evaluate diagnostic performance of radiomics models of MRI in the detection of differentiation degree (DD) and lymph node metastases (LNM) of extrahepatic cholangiocarcinoma (ECC). We retrospectively enrolled 100 patients with ECC confirmed by pathology from January 2011 to December 2018. Three hundred radiomics features were extracted from each region of interest using MaZda software. Next, the radiomics model was developed by incorporating the optimal radiomics signatures and ADC values of tumors to predict DD (model A) and LNM (model B) of ECC, respectively, through the random forest algorithm. After which, the performance of the radiomics models were further evaluated. The model A showed better performance in both training and testing cohorts to discriminate high and medium-low differentiation groups of ECC, with an average AUC of 0.78 and 0.80, respectively. The model B also yielded the good average AUC of 0.80 and 0.90 to predict the LNM of ECC in training and testing cohorts. The radiomics models based on MRI performed well in predicting DD and LNM of ECC and have significant potential in clinical noninvasive diagnosis and in the prediction of ECC.