Modulation Features Research Articles

REDef-DETR: real-time and efficient DETR for industrial surface defect detection

Abstract Industrial surface defect detection is an important part of industrial production, which aims to identify and detecting various defects on the surface of product to ensure quality and meet customer requirements. With the development of deep learning and image processing technologies, the surface defect detection methods based on computer vision has become the mainstream method. However, the prevalent convolutional neural network-based defect detection methods also have many problems. For example, these methods rely on post-processing of Non-Maximum Suppression and have poor detection ability for small targets, which affects the speed and accuracy of surface defect detection in industrial scenarios. Therefore, we propose a novel DEtection TRansformer-based surface defect detection method. Firstly, we propose a Multi-scale Contextual Information Dilated module and fuse it into the backbone. The module is mainly composed of large kernel convolutions, which aims to expand the receptive field of the model, thus reducing the leakage rate of the model. Moreover, we design an efficient encoder which mainly contains two important modules, namely feature enhancement based on cascaded group attention module and efficient feature fusion module based on content-aware. The former module effectively enhances the high-level semantic information extracted by the backbone, thus enabling the model to better interpret features, and it can improve the problem of high computational cost of transformer encoder, thus increasing the detection speed. The latter module performs multi-scale feature fusion across the feature information of various scales, thus improving the detection accuracy of the model for small-size defects. Experimental results show that the proposed method achieves 80.6%mAP and 80.3FPS on NEU-DET, and 98.0%mAP and 79.4FPS on PCB-DET. Our proposed method exhibits excellent detection performance and achieves real-time and efficient surface defect detection capability to meet the needs of industrial surface defect detection.

Collaborative leadership to empower nurses to implement ABCDE emergency nursing in an emergency department in Nepal

PurposeEmergency medicine can save lives and in 2018 the World Health Assembly passed resolution 72.16 ensuring the role of emergency care in all health systems. With a continued global shortage of emergency physicians, with many low-medium-income countries (LMIC) still to develop emergency medicine as a speciality, the role of emergency nurses is critical to deliver the WHO Emergency Care System Framework (WHO, 2018). Emergency medicine doctors play a critical role in collaborating with nurses, in emergency medicine where nurses are often the first clinicians are often the first clinicians to interact with patients in emergency care settings, making up the majority of health-care professionals in LMIC (Mamalelala, 2024). Yet emergency nursing has yet to become established in Nepal, where nurses are often recruited to emergency departments, without having received any training in emergency or critical care treatment and management. The purpose of this paper is to outline a collaborative leadership approach to co-design an airway, breathing, circulation, disability, exposure (ABCDE) structured approach to an emergency nursing training module designed for nurses to feel empowered in the emergency department and to report on its findings.Design/methodology/approachThis study draws upon mixed methodology research, enrolling 30 nurses (n = 30) from an emergency department in a tertiary hospital in Nepal through three stages of the project: Stage 1: training module co-design, collaborative leadership exploring the rationale for a training module and core features of design based on the ABCDE of emergency medicine; Stage 2: quantitative data were collected to assess baseline pre- and post-intervention knowledge and follow-up knowledge assessment at 30 and 45 days; Stage 3: qualitative data were collected with 24/30 (80%) nurses to evaluate the impact and application of the nurses ABCDE learning 7 months post-training. The qualitative survey was undertaken using online Google Forms.FindingsNurses were fully engaged in the co-design and collaboration of the development of an ABCDE training module which was delivered over 3 h. Full engagement was secured from all nurses in the department, and there were statistically significant advances in ABCDE emergency knowledge from the baseline, however, this knowledge began to decrease at 30 and 45 days. A follow-up qualitative survey was distributed to nurses seven months after training with an 80% return rate, which reported a range of examples of how nurses were continuing to apply their learning in practice.Originality/valueThis training module for emergency nurses was designed collaboratively from the “bottom up” in a tertiary hospital in Nepal, recognising the need to develop emergency nursing in the emergency department. The data revealed promising findings, while knowledge decreased from the post-training questionnaire, qualitative evidence revealed significant changes in practice, with the greatest reported change in the management of the airway. While this training module has made a difference in the quality of care provided, there is a need for a country-wide strategy in this area otherwise it is likely that such an initiative will only be developed by hospitals at a local level (Lecky, 2014). Education and training initiatives for nurses that focus on an evidence-based approach to clinical practice can bridge the workforce gap in the short term, however, the Government of Nepal must decide on establishing a recognised post-graduate sub-specialty in emergency nursing, the duration of training, who should be trained and what curriculum should be followed (Lecky, 2014).

SiamDCFF: Dynamic Cascade Feature Fusion for Vision Tracking.

Establishing an accurate and robust feature fusion mechanism is key to enhancing the tracking performance of single-object trackers based on a Siamese network. However, the output features of the depth-wise cross-correlation feature fusion module in fully convolutional trackers based on Siamese networks cannot establish global dependencies on the feature maps of a search area. This paper proposes a dynamic cascade feature fusion (DCFF) module by introducing a local feature guidance (LFG) module and dynamic attention modules (DAMs) after the depth-wise cross-correlation module to enhance the global dependency modeling capability during the feature fusion process. In this paper, a set of verification experiments is designed to investigate whether establishing global dependencies for the features output by the depth-wise cross-correlation operation can significantly improve the performance of fully convolutional trackers based on a Siamese network, providing experimental support for rational design of the structure of a dynamic cascade feature fusion module. Secondly, we integrate the dynamic cascade feature fusion module into the tracking framework based on a Siamese network, propose SiamDCFF, and evaluate it using public datasets. Compared with the baseline model, SiamDCFF demonstrated significant improvements.

Research on driving technology of radio‐over‐fiber (ROF) compact optical transmitter module

AbstractIn response to the problems of communication capacity and spectrum resource constraints, radio over fiber (ROF) technology has gained widespread adoption, and the quality and performance of the optical transmitter module also directly affect the transmission of the link. This paper presents the design of a direct modulation optical transmitter module in the frequency band ranging from 2.6 kHz to 206.8 MHz using OPA690 broadband voltage feedback op‐amp. The module features a tunable bias voltage range from 0 to 5 V to accommodate different threshold voltages for modulation power supply. It offers a gain of 12 dB and a signal‐to‐noise ratio exceeding 50 dB while demonstrating excellent stability, high signal purity, good spectral response characteristics, and compact size advantages. Based on this design, we further optimized the system circuit by adding a fourth‐order Butterworth filter structure, reducing the lower cut‐off frequency to about 500 Hz, increasing the gain by 6 dB, reducing the return loss to −26.91 dB, and the module's signal‐to‐noise ratio exceeding 90 dB. This optimization scheme improves the stability and signal purity of the optical transmitter module and improves the spectral response characteristics. In this paper, Multisim is used to analyze and verify the proposed optical transmitter circuit. In addition, a complete link experiment was carried out using an optical receiving module designed in the same laboratory to verify the feasibility of our proposed module.

Amplitude and frequency modulation of EEG predicts Intraventricular hemorrhage in preterm infants

BackgroundIntraventricular hemorrhage (IVH) is a common and significant complication in premature infants. While cranial ultrasound is the golden standard for IVH detection, it may not identify lesions until hours or days after occurring, which limits early intervention. Predicting IVH in premature infants would be highly advantageous. Recent studies have shown that EEG data’s amplitude and frequency modulation features could offer predictive insights for neurological diseases in adults. MethodsTo investigate the association between IVH and EEG monitoring, a retrospective case-control study was conducted in preterm infants. All infants underwent amplitude integrated EEG monitoring for at least 3 days after birth. The study included 20 cases who had an IVH diagnosed on cranial ultrasound and had a negative ultrasound 24 h earlier, and 20 matched controls without IVH. Amplitude and frequency modulation features were extracted from single-channel EEG data, and various machine learning algorithms were evaluated to create a predictive model. ResultsCases had an average gestational age and birth weight of 26.4 weeks and 965 g, respectively. The best-performing algorithm was adaptive boosting. EEG data from 24 h before IVH detection proved predictive with an area under the receiver operating characteristic curve of 93 %, an accuracy of 91 %, and a Kappa value of 0.85. The most informative features were the slow varying instantaneous frequency and amplitude in the Delta frequency band. ConclusionAmplitude and frequency modulation features obtained from single-channel EEG signals in extremely preterm infants show promise for predicting IVH occurrence within 24 h before detection on cranial ultrasound.

A broadband modulator based on graphene/black phosphorus heterostructure with enhanced modulation depth

We theoretically present a broadband modulator based on graphene/black phosphorus heterostructure which can work over a large waveband from visible (VIS) to mid-infrared (MIR) regions. By utilizing the angle dependence of black phosphorus, surface plasmon polaritons (SPP) modulation can be achieved in VIS regime, while the wavelength is tuned within the near-infrared (NIR) or MIR regions, the enhanced modulation depth can be achieved by few-layer graphene films. Results show that the proposed plasmonic modulator exhibits a broad waveband from 400 nm to 3 μm. In addition, this proposed modulator features high modulation depth (MD), low insertion loss (IL), large 3-dB modulation bandwidth and small power consumption from VIS to MIR regions. Our work may extend the operation waveband of opto-electro devices based on the hybridized 2D materials and would promote their potential future applications.

Research on unbalanced characteristics of axial-varying eccentric rotor based on signal demodulation

In the presence of non-uniform loads, rotor misalignment emerges as a prevalent fault in permanent magnet synchronous motors (PMSMs), leading to the manifestation of an axial-varying eccentric air–gap. Considering the three-dimensional (3D) nature of magnetic flux under axial-varying eccentric fault, 3D finite element (FE) analysis method was used in this paper. Then, the electromagnetism simulations were conducted to calculate the magnetic flux density and unbalanced magnetic pull (UMP). With the signal demodulation algorithm based on principal component analysis (DPCA), the frequency modulation of various signals from simulation and experiment was extracted. Furthermore, based on the modulation features, the relevance among magnetic flux density, UMP, and vibration acceleration was established with shaft frequency as the link. The coupling of magnetic field and vibration level shown in results was analysis, which provide important theoretical and engineering references for the safe operation of PMSM system and the exact identification of rotor tilting faults.

Special Features of Self-Access English Learning Pandemic Modules

In the Covid-19 pandemic, the Indonesian Ministry of Education and Culture developed special English self-access learning modules for grade IX students. These modules are intended to facilitate the students in studying English at home by themselves. There are special features in these modules that are especially provided for the students. These special features are worth to be investigated. The aim of this research is to explain the special features provided by the writers in the self-access English learning modules. This research is a content analysis research. It is found that there are five special features in these modules. These special features include 1) detailed objectives of each unit, 2) roles of the teacher and parents table, 3) Indonesian translation for the instructions, 4) graded scaffolding activities, and 5) detailed answer keys. Hopefully, this research would give new insights upon self-access English learning materials especially for special nationwide conditions.

663-P: Acceptability and Preferences for Features of a Family Module in an Artificial Intelligence–Enabled Empower Mobile App to Support Type 2 Diabetes Self-Management—A Formative Study

663-P: Acceptability and Preferences for Features of a Family Module in an Artificial Intelligence–Enabled Empower Mobile App to Support Type 2 Diabetes Self-Management—A Formative Study

CONTEXTUALIZED MODULE IN LAGUNA FESTIVAL DANCES IN ENHANCING STUDENT LEARNING COMPETENCIES AND PERFORMANCE

The main goal of this study was to examine effectiveness of the contextualized module in laguna festival dances in enhancing student learning competencies and performance. Particularly, this would determine the level of validity of the component and features of the contextualized module in festival dances. Also, this would like to identify the student’s learning competencies and performance. Moreover, it sought to prove the significant differences on student’s performance in formative and summative as well as the effect of component and features of festival dances on student’s competencies. This study used a descriptive method in collecting and gathering data. To gather data, the researcher utilized a self-made questionnaire and served as the primary tool distributed among 80 Grade 7 students. Statistical analyses included calculating the mean, standard deviation, frequency, percentage, t-test and Pearson correlation coefficient to assess hypotheses. From the data analysis of the study, it was found that the physical education students perceived the components of contextualized module in festival dances was very high. Also, students perceived the features of module in festival dances as very high and the students rated very high in level of student’s learning competencies. The students showed fairy satisfactory on their level of performance in terms formative while the students showed excellent performance on their summative assessment. Moreover, it was evident that the student’s performance showed significant difference from formative evaluation to summative evaluations. the objectives contextualized learning module showed importance in shaping two of the student’s competency namely cultural empathy and sensitivity as well as well as the cross-cultural collaboration while the majority of the components and features showed no significant. Based on the research finding, it can be concluded that the performance of student’s showed significant differences between formative and summative test results thus signified the rejection of hypothesis as the student’s performance in written examination increased after being exposed in the curriculum delivery through the utilization of contextualized module. The contextualized module showed no significant effects of variables. Since majority of the variables of the components and features of contextualized modules showed no significant effect in student competency thus signified acceptance of the hypothesis. The students enhanced their positive attitude in consideration of other’s ideas, opinion, and practices as they shared expertise in their academic requirement upon understanding the lesson goals and outcome of the discussion. The following recommendations were forwarded: The school may utilize the contextualized module and may practice the use of formative and summative assessment constantly. Also, The teachers may design contextualized learning module, design culminating activities and expose learners to different contextualized learning materials to maintain excellent performance in physical education. The future researchers are advised to study more cultural dances in the different municipality of laguna and its effects to student’s holistic development.

PCDC: Prototype-assisted dual-contrastive learning with depthwise separable convolutional neural network for few-shot fault diagnosis of permanent magnet synchronous motors under new operating conditions

Abstract The fault diagnosis of permanent magnet synchronous motor is of vital importance in industrial fields to ensure user safety and minimize economic losses from accidents. However, recent fault diagnosis methods, particularly the methods using deep learning, require a massive amount of labeled data, which may not be available in industrial fields. Few-shot learning has been recently applied in fault diagnosis for rotary machineries, to alleviate the data deficiency and/or to enable unseen fault diagnosis. However, two major obstacles still remain, specifically: a) the limited ability of the models to be generalized for use under new operating conditions and b) insufficient discriminative features to precisely diagnose fault types. To address these limitations, this study proposes a Prototype-assisted dual-Contrastive learning with Depthwise separable Convolutional neural network (PCDC) for few-shot fault diagnosis for permanent magnet synchronous motors under new working conditions. Operation-robust fault features are extracted to reinforce generalization of PCDC under new operating conditions by extracting fault-induced amplitude and frequency modulation features and by eliminating the influence of operating conditions from the motor stator current signals. Prototype-assisted dual-contrastive learning is proposed to clearly distinguish the fault categories even when the fault features are similar to each other by learning both local- and global-similarity features, which increases the instance-discrimination ability while alleviating an overfitting issue. Experimental results show that the proposed PCDC outperforms the comparison models in few-shot fault diagnosis tasks under new operating conditions.

Modelling the Influence of the Tube Support Plate on the Eddy Current Testing of the Steam Generator: Sensitivity Study & Validation

Eddy current testing (ECT) of Steam Generator (SG) tubes is part of the maintenance program of Nuclear Power Plants. Tube Support Plates (TSP) are usually only considered as extraneous signals for tube inspection. However, for long-term operation of SGs, there is a safety concern related to the build-up of deposits in the TSPs flow holes. The clogging-up of the TSPs flow holes affect the safe thermal-hydraulic operation of SGs. The deposits come from the feedwater train and are mostly iron oxides (magnetite) with a weak electrical conductivity and magnetite permeability.The ability of ECT probes to detect and characterize TSPs deposits is therefore of great interest. The most common and industrial ECT technique, the bobbin coil, averages the surrounding electromagnetic field over 360°, whereas the geometry of tri-foiled and quadri-foiled TSPs, and thus of clogging, is rather complex and non- axisymmetric, hence motivating the evaluation of the performance of conventional ECT rotating probes used for SGs tubes inspection. Although simulation is a powerful tool to support such a study, it requires dedicated models to be made available to NDT experts, and CEA develops ECT physical models in CIVA for this purpose. In addition to standard fast CIVA modules based on 3D semi-analytical or 2D numerical calculations restricted to canonical or axisymmetric parts, respectively, CEA develops a module dedicated to the simulation of SG tube inspection based on a 3D numerical simulation. This module allows tube deformation such as ovalization, bending or tube expansion. It also allows the addition of external objects such as anti-vibration bars, different geometries of tube support plates and now their clogging by deposits. This model is used to study various influent parameters and to perform benchmarks in the framework of a scientific collaboration between the CEA and the IRSN. Here, the specificity of the TSPs and their clogging justifies qualifying the simulation model before evaluating the influence of material and geometry properties of the deposit as well as the performance of the inspection techniques. We present the main features of the simulation module in CIVA, its experimental validation, and the main trends in the distortion of the simulated signal as a function of variation in the clogging description.

SDD-YOLO: A Lightweight, High-Generalization Methodology for Real-Time Detection of Strip Surface Defects

Flat-rolled steel sheets are one of the major products of the metal industry. Strip steel’s production quality is crucial for the economic and safety aspects of humanity. Addressing the challenges of identifying the surface defects of strip steel in real production environments and low detection efficiency, this study presents an approach for strip defect detection based on YOLOv5s, termed SDD-YOLO. Initially, this study designs the Convolution-GhostNet Hybrid module (CGH) and Multi-Convolution Feature Fusion block (MCFF), effectively reducing computational complexity and enhancing feature extraction efficiency. Subsequently, CARAFE is employed to replace bilinear interpolation upsampling to improve image feature utilization; finally, the Bidirectional Feature Pyramid Network (BiFPN) is introduced to enhance the model’s adaptability to targets of different scales. Experimental results demonstrate that, compared to the baseline YOLOv5s, this method achieves a 6.3% increase in mAP50, reaching 76.1% on the Northeastern University Surface Defect Database for Detection (NEU-DET), with parameters and FLOPs of only 3.4MB and 6.4G, respectively, and FPS reaching 121, effectively identifying six types of defects such as Crazing and Inclusion. Furthermore, under the conditions of strong exposure, insufficient brightness, and the addition of Gaussian noise, the model’s mAP50 still exceeds 70%, demonstrating the model’s strong robustness. In conclusion, the proposed SDD-YOLO in this study features high accuracy, efficiency, and lightweight characteristics, making it applicable in actual production to enhance strip steel production quality and efficiency.

The amplitude modulation bispectrum: A weak modulation features extracting method for bearing fault diagnosis

As a key component of the industrial equipment, the failure of rolling bearings could cause equipment to malfunction, resulting in serious losses. However, the bearing fault information is often disturbed by the operation of other parts in the industrial equipment. A key issue is the recognition of weak information in the fault diagnosis of rolling bearings. The amplitude modulation bispectrum proposed in this paper is a rolling bearing fault diagnosis method for weak modulation feature extraction. Aiming at the problem that the bearing fault information is weak and difficult to extract, the method reconfigures the amplitude of the signal in the frequency domain to adjust the proportions of different components in the signal effectively and highlight the fault characteristic information. Based on the use of advanced demodulation tool to deal with the complex modulation components in the signal, an index, named bispectrum signal-to-noise ratio, to evaluate the fault information quantitatively for two-dimensional data is also proposed. The index helps to optimize the bispectrum demodulation results and make the valid information clearer. The effectiveness of this method in rolling bearing fault diagnosis is confirmed using simulation and experimental signals, and comparison with other methods has demonstrated the superiority of this method.

Comparative analyses of gene networks mediating cancer metastatic potentials across lineage types.

Studies have identified genes and molecular pathways regulating cancer metastasis. However, it remains largely unknown whether metastatic potentials of cancer cells from different lineage types are driven by the same or different gene networks. Here, we aim to address this question through integrative analyses of 493 human cancer cells' transcriptomic profiles and their metastatic potentials in vivo. Using an unsupervised approach and considering both gene coexpression and protein-protein interaction networks, we identify different gene networks associated with various biological pathways (i.e. inflammation, cell cycle, and RNA translation), the expression of which are correlated with metastatic potentials across subsets of lineage types. By developing a regularized random forest regression model, we show that the combination of the gene module features expressed in the native cancer cells can predict their metastatic potentials with an overall Pearson correlation coefficient of 0.90. By analyzing transcriptomic profile data from cancer patients, we show that these networks are conserved in vivo and contribute to cancer aggressiveness. The intrinsic expression levels of these networks are correlated with drug sensitivity. Altogether, our study provides novel comparative insights into cancer cells' intrinsic gene networks mediating metastatic potentials across different lineage types, and our results can potentially be useful for designing personalized treatments for metastatic cancers.

Detection of Road Crack Images Based on Multistage Feature Fusion and a Texture Awareness Method.

Structural health monitoring for roads is an important task that supports inspection of transportation infrastructure. This paper explores deep learning techniques for crack detection in road images and proposes an automatic pixel-level semantic road crack image segmentation method based on a Swin transformer. This method employs Swin-T as the backbone network to extract feature information from crack images at various levels and utilizes the texture unit to extract the texture and edge characteristic information of cracks. The refinement attention module (RAM) and panoramic feature module (PFM) then merge these diverse features, ultimately refining the segmentation results. This method is called FetNet. We collect four public real-world datasets and conduct extensive experiments, comparing FetNet with various deep-learning methods. FetNet achieves the highest precision of 90.4%, a recall of 85.3%, an F1 score of 87.9%, and a mean intersection over union of 78.6% on the Crack500 dataset. The experimental results show that the FetNet approach surpasses other advanced models in terms of crack segmentation accuracy and exhibits excellent generalizability for use in complex scenes.

Should we care about pay ratios? Dealing with ISSP questions on actual and ideal wages from a comparative perspective

ABSTRACT Questions on actual and ideal wages of various occupations are distinctive features of ISSP Social Inequality modules which can be used to quantify respondents’ perceived and ideal wage inequality, by computing pay ratios ex-post. However, recent research has raised harsh criticism of pay ratios, emphasizing that they suffer from anchoring bias, ratio bias, and, more generally, low validity. This article investigates what adjustments can be made to improve the validity of pay ratios for comparative analyses. It is found that country comparisons are very sensitive to adjustments and have major limitations. Nevertheless, as attitudes expressed by word-based items do not measure the extent of inequalities that respondents are willing to tolerate, further research is needed to find out ways to fix problems with pay ratios at the level of data collection rather than of data analysis. The article concludes with some suggestions regarding data collection.

On the nonlinear energy interactions in vibro-impacting cantilever

The nonlinear energy interactions are an intriguing aspect of nonlinear oscillators, such as vibro-impacting cantilevers (VICs). Despite extensive research, the domain of energy interactions in nonlinear oscillators still requires further exploration. The motivation arises from the complex and captivating dynamics as well as the vast array of potential applications. This paper presents a comprehensive investigation of various energy interaction mechanisms within the resonance realms of the first and second mode resonance frequencies of harmonically base-excited VICs. These energy interactions are of paramount importance as they give rise to the captivating response of the VIC, characterized by multi- and quasi-periodicity, energy amplification, combination resonances, amplitude modulation, and other intriguing features. This paper introduces the phenomenon termed “Harmonic Resonance Energy Stimulation (HRES)” and extensively investigates its role in generating the secondary jump in the frequency response. Moreover, the present study discovers and reports the non-resonant energy interactions and resulting modulating, quasi-periodic response in the VIC. Various illustrative plots obtained from the theoretical investigation, relying on the mathematical model, have been utilized to facilitate a comprehensive investigation. The results hold immense significance as they contribute valuable knowledge to the nonlinear dynamics of VIC, which is of great importance in research advancements and practical applications.

Using cumulant analysis to recognize digital types of radio signal

Problem statement. Recognition of types of modulation of radio signals is an important task solved in radio engineering systems for various purposes. In radio monitoring tools, information about the type of modulation is used to identify unauthorized sources of radio emission and measure their parameters, in electronic warfare systems - when choosing a more effective type of interference, etc. The article substantiates the relevance of improving methods for recognizing digital types of modulation. It is shown that mixed cumulants of higher orders can be used as modulation features for recognizing types of modulation. Analytical expressions are given for calculating mixed cumulants up to the tenth order inclusive through the values of mixed moments and the value of mixed cumulants for radio signals with the most common types of digital modulation. Recommendations have been developed for the synthesis of algorithms for recognizing the type of modulation based on cumulant analysis. Goal. Obtain mathematical expressions for calculating mixed cumulants up to the tenth order inclusive and determine the values of mixed cumulants to recognize the most common digital types of modulation. Results. A general methodology has been developed for the synthesis of algorithms for recognizing the type of modulation based on cumulant analysis. Practical significance. Application of the developed approach in the design of radio-electronic systems makes it possible to obtain a high probability of correct recognition of digital types of modulation with a relatively low design complexity.

NTSM: a non-salient target segmentation model for oral mucosal diseases

BackgroundOral mucosal diseases are similar to the surrounding normal tissues, i.e., their many non-salient features, which poses a challenge for accurate segmentation lesions. Additionally, high-precision large models generate too many parameters, which puts pressure on storage and makes it difficult to deploy on portable devices.MethodsTo address these issues, we design a non-salient target segmentation model (NTSM) to improve segmentation performance while reducing the number of parameters. The NTSM includes a difference association (DA) module and multiple feature hierarchy pyramid attention (FHPA) modules. The DA module enhances feature differences at different levels to learn local context information and extend the segmentation mask to potentially similar areas. It also learns logical semantic relationship information through different receptive fields to determine the actual lesions and further elevates the segmentation performance of non-salient lesions. The FHPA module extracts pathological information from different views by performing the hadamard product attention (HPA) operation on input features, which reduces the number of parameters.ResultsThe experimental results on the oral mucosal diseases (OMD) dataset and international skin imaging collaboration (ISIC) dataset demonstrate that our model outperforms existing state-of-the-art methods. Compared with the nnU-Net backbone, our model has 43.20% fewer parameters while still achieving a 3.14% increase in the Dice score.ConclusionsOur model has high segmentation accuracy on non-salient areas of oral mucosal diseases and can effectively reduce resource consumption.