Ali Alkhathlan, Faris Alahmadi, Faris Kateb, Hend Al-Khalifa
{"title":"Constructing and evaluating ArabicStanceX: a social media dataset for Arabic stance detection.","authors":"Ali Alkhathlan, Faris Alahmadi, Faris Kateb, Hend Al-Khalifa","doi":"10.3389/frai.2025.1615800","DOIUrl":"https://doi.org/10.3389/frai.2025.1615800","url":null,"abstract":"<p><p>Arabic stance detection has attracted significant interest due to the growing importance of social media in shaping public opinion. However, the lack of comprehensive datasets has limited research progress in Arabic Natural Language Processing (NLP). To address this, we introduce ArabicStanceX, a novel and extensive Arabic stance detection dataset sourced from social media, comprising 14,477 tweets across 17 diverse topics. Utilizing the transformer-based MARBERTv2 model, we explore stance detection through Multi-Topic Single Model (MTSM) strategies, achieving a promising F1 score of 0.74 for detecting 'favor' and 'against' stances, and 0.67 overall. Our experiments highlight the model's capabilities and challenges, particularly in accurately classifying neutral stances and generalizing to unseen topics. Further investigations using zero-shot and few-shot learning demonstrate the model's adaptability to new contexts. This study significantly advances Arabic NLP, providing crucial resources and insights into stance detection methodologies and future research directions. The dataset is publicly available.</p>","PeriodicalId":33315,"journal":{"name":"Frontiers in Artificial Intelligence","volume":"8 ","pages":"1615800"},"PeriodicalIF":3.0,"publicationDate":"2025-06-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12222287/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144561343","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"On the construction of artificial general intelligence based on the correspondence between goals and means.","authors":"Pavel Prudkov","doi":"10.3389/frai.2025.1588726","DOIUrl":"10.3389/frai.2025.1588726","url":null,"abstract":"<p><p>Humans are goal-directed agents and intelligence is suggested to be a characteristic of such agents. AGI can be achieved following the principle of the goals-means correspondence that posits the necessary condition for achieving a goal is the correspondence between the goal and the means. The goals-means correspondence is used in all architectures underlying intelligent systems. There are two conventional architectures regarding how the correspondence can be established. One conventional architecture that is based on observations of animals, is intelligent agents whose goals, means, or criteria for its construction are determined jointly at the moment of the birth of an agent. The other conventional architecture that is based on the analysis of human actions, defines intelligent agents whose goals and means are constructed arbitrarily and independently from each other. The conventional architectures cannot explain human actions and thinking. Since the conventional architectures underlie all artificial intelligent systems these systems are insufficient to construct AGI. The formal analysis of architectures demonstrates that there is another architecture in that arbitrary goals and means are constructed jointly on the basis of the criterion of minimal construction costs. This architecture is suggested to underlie human goal-directed processes. The view on humans as goal-directed agents constructing goals and means jointly allows creating an AGI agent that is capable of functioning in real situations. Unlike conventional AI agents that have an unaltered structure, the structure of agents in the new architecture is alterable. The development of an AGI agent may be similar to human growth from an infant to an adult. A model including a simple agent based on the new architecture, is considered. In the model the agent wanders in a quadrangular field filled with various objects that stimulate the agent to move in several directions simultaneously, thus trapping the agent. However, changing its structure the agent constructs goal-directed processes; therefore it is capable of leaving traps.</p>","PeriodicalId":33315,"journal":{"name":"Frontiers in Artificial Intelligence","volume":"8 ","pages":"1588726"},"PeriodicalIF":3.0,"publicationDate":"2025-06-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12213789/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144555175","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Ayat S Hammad, Ali Tajammul, Ismail Dergaa, Maha Al-Asmakh
{"title":"Machine learning applications in the analysis of sedentary behavior and associated health risks.","authors":"Ayat S Hammad, Ali Tajammul, Ismail Dergaa, Maha Al-Asmakh","doi":"10.3389/frai.2025.1538807","DOIUrl":"10.3389/frai.2025.1538807","url":null,"abstract":"<p><strong>Background: </strong>The rapid advancement of technology has brought numerous benefits to public health but has also contributed to a rise in sedentary lifestyles, linked to various health issues. As prolonged inactivity becomes a growing public health concern, researchers are increasingly utilizing machine learning (ML) techniques to examine and understand these patterns. ML offers powerful tools for analyzing large datasets and identifying trends in physical activity and inactivity, generating insights that can support effective interventions.</p><p><strong>Objectives: </strong>This review aims to: (i) examine the role of ML in analyzing sedentary patterns, (ii) explore how different ML techniques can be optimized to improve the accuracy of predicting sedentary behavior, and (iii) assess strategies to enhance the effectiveness of ML algorithms.</p><p><strong>Methods: </strong>A comprehensive search was conducted in PubMed and Scopus, targeting peer-reviewed articles published between 2004 and 2024. The search included the subject terms \"sedentary behavior,\" \"sedentary lifestyle health,\" and \"machine learning sedentary lifestyle,\" combined with the keywords \"physical inactivity\" and \"diseases\" using Boolean operators (AND, OR). Articles were included if they addressed the health impacts of sedentary behavior or employed ML techniques for its analysis. Exclusion criteria involved studies older than 20 years or lacking direct relevance. After screening 33 core articles and identifying 13 more through citation tracking, 46 articles were included in the final review.</p><p><strong>Results: </strong>This narrative review describes the characteristics of sedentary behavior, associated health risks, and the applications of ML in this context. Based on the reviewed literature, sedentary behavior was consistently associated with cardiovascular disease, metabolic disorders, and mental health conditions. The review highlights the utility of various ML approaches in classifying activity levels and significantly improving the prediction of sedentary behavior, offering a promising approach to address this widespread health issue.</p><p><strong>Conclusion: </strong>ML algorithms, including supervised and unsupervised models, show great potential in accurately detecting and predicting sedentary behavior. When integrated with wearable sensor data and validated in real-world settings, these models can enhance the scalability and precision of AI-driven interventions. Such advancements support personalized health strategies and could help lower healthcare costs linked to physical inactivity, ultimately improving public health outcomes.</p>","PeriodicalId":33315,"journal":{"name":"Frontiers in Artificial Intelligence","volume":"8 ","pages":"1538807"},"PeriodicalIF":3.0,"publicationDate":"2025-06-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12213918/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144555158","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
José Manuel Brotons-Martínez, José María Cámara-Zapata
{"title":"The evaluation of performance for agroecological greenhouse tomato strategies by the CRITIC-OWA model.","authors":"José Manuel Brotons-Martínez, José María Cámara-Zapata","doi":"10.3389/frai.2025.1599334","DOIUrl":"10.3389/frai.2025.1599334","url":null,"abstract":"<p><strong>Introduction: </strong>Modern agriculture must begin to use production strategies that are increasingly sustainable. To help in decision-making, the present work analyzes the sustainability of greenhouse tomato production with different agroecological strategies: shading (conventional fixed mesh and mobile photovoltaic shading), grafting and deficit irrigation, based on economic, social, and environmental criteria.</p><p><strong>Methods: </strong>For the ranking of the different strategies, the use of an extension of the CRiteria Importance Through Inter-criteria Correlation (CRITIC) is proposed, in which the correlation between the criteria is obtained through the Pearson-OWA, where the aggregation of the quadratic differences between criteria is carried out considering the attitudinal character of the decision-maker, that is, using Ordered Weighted Averaging (OWA), in addition to induced variables, with the Induced Probabilistic OWA CRITIC (IPOWA CRITIC). Three extensions are considered based on this model depending on the way the multicriteria score is calculated: i) the ranking is carried out on the relative score (S) of each alternative (IPOWA-S-CRITIC), ii) on the weighting vector (W) (IPOWA-W-CRITIC), or iii) on both (IPOWA-S-W-CRITIC).</p><p><strong>Results: </strong>The results of the classifications conducted indicate that the use of mobile photovoltaic mesh is a sustainable production strategy, due to its effect on production and quality of the crop, CO2 fixation, and irrigation water savings.</p><p><strong>Discussion: </strong>The use of mobile photovoltaic shades is compatible with tomato cultivation in a greenhouse if the management of the installation is performed considering the needs of the plants in most of the rankings.</p>","PeriodicalId":33315,"journal":{"name":"Frontiers in Artificial Intelligence","volume":"8 ","pages":"1599334"},"PeriodicalIF":3.0,"publicationDate":"2025-06-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12213796/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144555177","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Guoyao Shen, Yancheng Zhu, Mengyu Li, Ryan McNaughton, Hernan Jara, Sean B Andersson, Chad W Farris, Stephan Anderson, Xin Zhang
{"title":"Regularization by neural style transfer for MRI field-transfer reconstruction with limited data.","authors":"Guoyao Shen, Yancheng Zhu, Mengyu Li, Ryan McNaughton, Hernan Jara, Sean B Andersson, Chad W Farris, Stephan Anderson, Xin Zhang","doi":"10.3389/frai.2025.1579251","DOIUrl":"10.3389/frai.2025.1579251","url":null,"abstract":"<p><p>Recent advances in MRI reconstruction have demonstrated remarkable success through deep learning-based models. However, most existing methods rely heavily on large-scale, task-specific datasets, making reconstruction in data-limited settings a critical yet underexplored challenge. While regularization by denoising (RED) leverages denoisers as priors for reconstruction, we propose Regularization by Neural Style Transfer (RNST), a novel framework that integrates a neural style transfer (NST) engine with a denoiser to enable magnetic field-transfer reconstruction. RNST generates high-field-quality images from low-field inputs without requiring paired training data, leveraging style priors to address limited-data settings. Our experiment results demonstrate RNST's ability to reconstruct high-quality images across diverse anatomical planes (axial, coronal, sagittal) and noise levels, achieving superior clarity, contrast, and structural fidelity compared to lower-field references. Crucially, RNST maintains robustness even when style and content images lack exact alignment, broadening its applicability in clinical environments where precise reference matches are unavailable. By combining the strengths of NST and denoising, RNST offers a scalable, data-efficient solution for MRI field-transfer reconstruction, demonstrating significant potential for resource-limited settings.</p>","PeriodicalId":33315,"journal":{"name":"Frontiers in Artificial Intelligence","volume":"8 ","pages":"1579251"},"PeriodicalIF":3.0,"publicationDate":"2025-06-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12213525/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144555176","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"VMDU-net: a dual encoder multi-scale fusion network for polyp segmentation with Vision Mamba and Cross-Shape Transformer integration.","authors":"Peng Li, Jianhua Ding, Chia S Lim","doi":"10.3389/frai.2025.1557508","DOIUrl":"10.3389/frai.2025.1557508","url":null,"abstract":"<p><strong>Introduction: </strong>Rectal cancer often originates from polyps. Early detection and timely removal of polyps are crucial for preventing colorectal cancer and inhibiting its progression to malignancy. While polyp segmentation algorithms are essential for aiding polyp removal, they face significant challenges due to the diverse shapes, unclear boundaries, and varying sizes of polyps. Additionally, capturing long-range dependencies remains difficult, with many existing algorithms struggling to converge effectively, limiting their practical application.</p><p><strong>Methods: </strong>To address these challenges, we propose a novel Dual Encoder Multi-Scale Feature Fusion Network, termed VMDU-Net. This architecture employs two parallel encoders: one incorporates Vision Mamba modules, and the other integrates a custom-designed Cross-Shape Transformer. To enhance semantic understanding of polyp morphology and boundaries, we design a Mamba-Transformer-Merge (MTM) module that performs attention-weighted fusion across spatial and channel dimensions. Furthermore, Depthwise Separable Convolutions are introduced to facilitate multi-scale feature extraction and improve convergence efficiency by leveraging the inductive bias of convolution.</p><p><strong>Results: </strong>Extensive experiments were conducted on five widely-used polyp segmentation datasets. The results show that VMDU-Net significantly outperforms existing state-of-the-art methods, especially in terms of segmentation accuracy and boundary detail preservation. Notably, the model achieved a Dice score of 0.934 on the Kvasir-SEG dataset and 0.951 on the CVC-ClinicDB dataset.</p><p><strong>Discussion: </strong>The proposed VMDU-Net effectively addresses key challenges in polyp segmentation by leveraging complementary strengths of Transformer-based and Mamba-based modules. Its strong performance across multiple datasets highlights its potential for practical clinical application in early colorectal cancer prevention.</p><p><strong>Code availability: </strong>The source code is publicly available at: https://github.com/sulayman-lee0212/VMDUNet/tree/4a8b95804178511fa5798af4a7d98fd6e6b1ebf7.</p>","PeriodicalId":33315,"journal":{"name":"Frontiers in Artificial Intelligence","volume":"8 ","pages":"1557508"},"PeriodicalIF":3.0,"publicationDate":"2025-06-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12213873/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144555178","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Florian Peter Busch, Matej Zečević, Kristian Kersting, Devendra Singh Dhami
{"title":"Elucidating linear programs by neural encodings.","authors":"Florian Peter Busch, Matej Zečević, Kristian Kersting, Devendra Singh Dhami","doi":"10.3389/frai.2025.1549085","DOIUrl":"https://doi.org/10.3389/frai.2025.1549085","url":null,"abstract":"<p><p>Linear Programs (LPs) are one of the major building blocks of AI and have championed recent strides in differentiable optimizers for learning systems. While efficient solvers exist for even high-dimensional LPs, explaining their solutions has not received much attention yet, as explainable artificial intelligence (XAI) has mostly focused on deep learning models. LPs are mostly considered white-box and thus assumed simple to explain, but we argue that they are not easy to understand in terms of relationships between inputs and outputs. To mitigate this rather non-explainability of LPs we show how to adapt attribution methods by encoding LPs in a neural fashion. The encoding functions consider aspects such as the feasibility of the decision space, the cost attached to each input, and the distance to special points of interest. Using a variety of LPs, including a very large-scale LP with 10k dimensions, we demonstrate the usefulness of explanation methods using our neural LP encodings, although the attribution methods Saliency and LIME are indistinguishable for low perturbation levels. In essence, we demonstrate that LPs can and should be explained, which can be achieved by representing an LP as a neural network.</p>","PeriodicalId":33315,"journal":{"name":"Frontiers in Artificial Intelligence","volume":"8 ","pages":"1549085"},"PeriodicalIF":3.0,"publicationDate":"2025-06-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12213677/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144555155","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Medical reasoning in LLMs: an in-depth analysis of DeepSeek R1.","authors":"Birger Moëll, Fredrik Sand Aronsson, Sanian Akbar","doi":"10.3389/frai.2025.1616145","DOIUrl":"10.3389/frai.2025.1616145","url":null,"abstract":"<p><strong>Introduction: </strong>The integration of large language models (LLMs) into healthcare holds immense promise, but also raises critical challenges, particularly regarding the interpretability and reliability of their reasoning processes. While models like DeepSeek R1-which incorporates explicit reasoning steps-show promise in enhancing performance and explainability, their alignment with domain-specific expert reasoning remains understudied.</p><p><strong>Methods: </strong>This paper evaluates the medical reasoning capabilities of DeepSeek R1, comparing its outputs to the reasoning patterns of medical domain experts.</p><p><strong>Results: </strong>Through qualitative and quantitative analyses of 100 diverse clinical cases from the MedQA dataset, we demonstrate that DeepSeek R1 achieves 93% diagnostic accuracy and shows patterns of medical reasoning. Analysis of the seven error cases revealed several recurring errors: anchoring bias, difficulty integrating conflicting data, limited consideration of alternative diagnoses, overthinking, incomplete knowledge, and prioritizing definitive treatment over crucial intermediate steps.</p><p><strong>Discussion: </strong>These findings highlight areas for improvement in LLM reasoning for medical applications. Notably the length of reasoning was important with longer responses having a higher probability for error. The marked disparity in reasoning length suggests that extended explanations may signal uncertainty or reflect attempts to rationalize incorrect conclusions. Shorter responses (e.g., under 5,000 characters) were strongly associated with accuracy, providing a practical threshold for assessing confidence in model-generated answers. Beyond observed reasoning errors, the LLM demonstrated sound clinical judgment by systematically evaluating patient information, forming a differential diagnosis, and selecting appropriate treatment based on established guidelines, drug efficacy, resistance patterns, and patient-specific factors. This ability to integrate complex information and apply clinical knowledge highlights the potential of LLMs for supporting medical decision-making through artificial medical reasoning.</p>","PeriodicalId":33315,"journal":{"name":"Frontiers in Artificial Intelligence","volume":"8 ","pages":"1616145"},"PeriodicalIF":3.0,"publicationDate":"2025-06-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12213874/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144555174","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"European sovereign debt control through reinforcement learning.","authors":"Tato Khundadze, Willi Semmler","doi":"10.3389/frai.2025.1569395","DOIUrl":"https://doi.org/10.3389/frai.2025.1569395","url":null,"abstract":"<p><p>The resilience of economic systems depends mainly on coordination among key stakeholders during macroeconomic or external shocks, while a lack of coordination can lead to financial and economic crises. The paper builds on the experience of global and regional shocks, such as the Eurozone crises of 2009-2012 and the economic disruption resulting from COVID-19, starting in 2020. The paper demonstrates the importance of cooperation in monetary and fiscal policies during emergencies to address macroeconomic non-resilience, particularly focusing on public debt management. The Euro area is chosen as the sample for testing the models presented in the paper, given that its resilience is heavily dependent on cooperation among different actors within the region. The shocks affecting nations within the European Union are asymmetric, and the responses to these shocks require coordination, considering heterogeneous economic structures, levels of economic development, and policies. We develop a macroeconomic modeling framework to simulate fiscal and monetary policy interactions under a cooperative regime. The approach builds on earlier nonlinear control models and incorporates modern reinforcement learning techniques. Specifically, we implement the Soft Actor-Critic algorithm to optimize policy responses across key variables including inflation, interest rates, output gaps, public debt, and government net lending. We demonstrate that the Soft Actor-Critic algorithm provides comparable or, in some cases, better solutions to multi-objective macroeconomic optimization problems, in comparison to Nonlinear Model Predictive Control (NMPC) algorithm.</p>","PeriodicalId":33315,"journal":{"name":"Frontiers in Artificial Intelligence","volume":"8 ","pages":"1569395"},"PeriodicalIF":3.0,"publicationDate":"2025-06-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12213571/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144555156","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"HMA-Net: a hybrid mixer framework with multihead attention for breast ultrasound image segmentation.","authors":"Soumya Sara Koshy, L Jani Anbarasi","doi":"10.3389/frai.2025.1572433","DOIUrl":"10.3389/frai.2025.1572433","url":null,"abstract":"<p><strong>Introduction: </strong>Breast cancer is a severe illness predominantly affecting women, and in most cases, it leads to loss of life if left undetected. Early detection can significantly reduce the mortality rate associated with breast cancer. Ultrasound imaging has been widely used for effectively detecting the disease, and segmenting breast ultrasound images aid in the identification and localization of tumors, thereby enhancing disease detection accuracy. Numerous computer-aided methods have been proposed for the segmentation of breast ultrasound images.</p><p><strong>Methods: </strong>A deep learning-based architecture utilizing a ConvMixer-based encoder and ConvNeXT-based decoder coupled with convolution-enhanced multihead attention has been proposed for segmenting breast ultrasound images. The enhanced ConvMixer modules utilize spatial filtering and channel-wise integration to efficiently capture local and global contextual features, enhancing feature relevance and thus increasing segmentation accuracy through dynamic channel recalibration and residual connections. The bottleneck with the attention mechanism enhances segmentation by utilizing multihead attention to capture long-range dependencies, thus enabling the model to focus on relevant features across distinct regions. The enhanced ConvNeXT modules with squeeze and excitation utilize depthwise convolution for efficient spatial filtering, layer normalization for stabilizing training, and residual connections to ensure the preservation of relevant features for accurate segmentation. A combined loss function, integrating binary cross entropy and dice loss, is used to train the model.</p><p><strong>Results: </strong>The proposed model has an exceptional performance in segmenting intricate structures, as confirmed by comprehensive experiments conducted on two datasets, namely the breast ultrasound image dataset (BUSI) dataset and the BrEaST dataset of breast ultrasound images. The model achieved a Jaccard index of 98.04% and 94.84% and a Dice similarity coefficient of 99.01% and 97.35% on the BUSI and BrEaST datasets, respectively.</p><p><strong>Discussion: </strong>The ConvMixer and ConvNeXT modules are integrated with convolution-enhanced multihead attention, which enhances the model's ability to capture local and global contextual information. The strong performance of the model on the BUSI and BrEaST datasets demonstrates the robustness and generalization capability of the model.</p>","PeriodicalId":33315,"journal":{"name":"Frontiers in Artificial Intelligence","volume":"8 ","pages":"1572433"},"PeriodicalIF":3.0,"publicationDate":"2025-06-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12213868/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144555157","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}