Innovative Architecture Research Articles

TIA-UNet: transformer-enhanced deep learning for adolescent idiopathic scoliosis spinal x-ray image segmentation

Abstract Adolescent Idiopathic Scoliosis (AIS) is a common spinal deformity where precise diagnosis is crucial for developing effective treatment strategies. Traditional manual x-ray image analysis is time-consuming and highly dependent on the operator’s expertise, thus constraining diagnostic efficiency and accuracy. This study aimed to develop an automated thoracolumbar spine segmentation method utilizing deep learning to enhance the efficiency and accuracy of AIS diagnosis. We introduced TIA-UNet, an innovative network architecture that combines Convolutional Neural Networks (CNN) with Transformer models. By integrating the IR Block and DA Block, TIA-UNet was optimized for feature extraction and multi-scale information fusion. The model underwent training and validation using our established Adolescent Scoliosis Medical Dataset (ASMD). TIA-UNet attained a Dice similarity coefficient of 90. 02%, a Mean Intersection over Union (MIoU) of 81. 96%, and a Hausdorff Distance (HD) of 4. 09, significantly surpassing current state-of-the-art methods such as UNet and TransUNet. Moreover, TIA-UNet exhibited superior computational efficiency regarding parameter count, inference time, and floating-point operations (FLOPs). As an automated medical image segmentation algorithm, TIA-UNet enhanced segmentation accuracy while maintaining high computational efficiency, demonstrating significant potential for clinical diagnostic applications. This study provides compelling evidence supporting the utilization of deep learning techniques in medical image analysis.

Exploring the foundations of ‘the architecture of innovation: how firms configure different types of complementarities in emerging ecosystems’

Exploring the foundations of ‘the architecture of innovation: how firms configure different types of complementarities in emerging ecosystems’

A plaintext-related and ciphertext feedback mechanism for medical image encryption based on a new one-dimensional chaotic system

Abstract In recent years, Plaintext-Related Image Encryption (PRIE) algorithms have been introduced, demonstrating a commendable level of plaintext sensitivity to resist chosen plaintext attack (CPA). However, these approaches suffer from several drawbacks, including inability to fully reconstruct the original image, limited practical value and excessive computational demands etc.. Moreover, the exponential expansion of medical data necessitates the formulation of more secure and efficient encryption algorithms. In this paper, firstly, a novel one-dimensional chaotic map, designated as 1D-SAM, which strikes an excellent balance between structural complexity and chaotic performance is proposed. The 1D-SAM achieve a larger chaotic range and an elevated Lyapunov exponent, signifying enhanced dynamical complexity. Subsequently, we devise a lightweight medical image encryption system leveraging the 1D-SAM and an innovative diffusion architecture, termed the plaintext-related and ciphertext feedback mechanism(PRCFM). This encryption system is a symmetric-key cryptosystem, eliminating the need for transmitting supplementary data beyond the secret keys to the recipient. Notably, the encrypted image maintains identical dimensions to its original counterpart and is fully recoverable. Complete simulation experiments were conducted on a personal computer equipped with MATLAB R2021a, OS Windows 11, 2.60GHz CPU and 16GB RAM. The experimental results indicate that our encryption system, employing a single permutation-diffusion round, efficiently encrypts a 512×512 image in approximately 0.2854 seconds. Leveraging the advantages of the PRCFM, our approach demonstrates superior plaintext sensitivity, achieving an average number of pixels changing rate (NPCR) of 99.6051 and a unified average changed intensity (UACI) of 33.4452. In summary, our work addresses key limitations of contemporary encryption frameworks, exhibiting acceptable performance in both encryption speed and security strength.

FusionGCN: Multi-focus image fusion using superpixel features generation GCN and pixel-level feature reconstruction CNN

In recent years, convolutional neural networks have demonstrated significant advancements in the domain of computer vision, effectively addressing numerous previously challenging issues. An increasing number of researchers are focusing their investigations on this field, proposing innovative network architectures. However, many existing networks necessitate intricate module designs and a substantial number of parameters to achieve satisfactory fusion outcomes, which poses challenges for lightweight devices with constrained computational resources. To mitigate this concern, the present study introduces a novel methodology that integrates block segmentation with pixel optimization. Specifically, we initially employ graph convolutional networks to execute flexible convolutions on large-scale, irregular regions generated through superpixel clustering, thereby achieving coarse segmentation at the block level. Subsequently, we utilize parallel lightweight convolutional networks to provide pixel-level guidance, ultimately resulting in a more accurate decision map. Furthermore, to leverage the strengths of both networks and facilitate the optimization of feature generation from the graph convolutional network for non-Euclidean data, we meticulously design a superpixel-based graph decoder alongside a pixel-based convolutional neural network extraction block to enhance feature acquisition and propagation. In comparison to numerous state-of-the-art methodologies, our approach demonstrates commendable competitiveness in both qualitative and quantitative analyses, as well as in efficiency evaluations. The code can be downloaded at https://github.com/ouyangbaicai/FusionGCN.

Transparent and Mechanically Robust Janus Nanofiber Membranes for Open Wound Healing and Monitoring.

The electrospun nanofiber membrane has demonstrated great potential for wound management due to its porous structure, large surface area, mechanical strength, and barrier properties. However, there is a need to develop transparent bioactive nanofibers with strong mechanical properties to facilitate the monitoring of the healing process. In this study, we present an electrospinning-based method for creating transparent (∼80-90%), strong (∼11-13 MPa), and Janus nanofiber membranes. The innovative square pattern architecture of the membrane includes a thin hydrophobic polycaprolactone layer on top of a layer of hydrophilic ethylene-vinyl alcohol nanofiber, which enables the absorption of excess biofluid from the wound and exhibits Janus wettability for water. Furthermore, incorporating 5% chitosan into the composition of the nanofibers accelerates the healing process through its antioxidant properties and antimicrobial activity against various bacteria, including drug-resistant strains. The developed membrane also demonstrates skin-repairing function, quick blood clotting (around 145 ± 12 s), and biocompatibility with keratinocyte (≥90%), as well as in vitro quick cell migration (∼24 h). With a tensile strength of 11-13 MPa, the membrane effectively adheres to the knee joint even after running 4 km. These optimal properties of the electrospun nanofiber membrane make it suitable for effective wound management and inspection of the healing process, without the need for frequent dressing changes.

A Review of Hybrid Search Architectures: Integrating Keyword and Semantic Search to Optimize Query Relevance

Abstract—The advent of semantic search alongside the tra- ditional robustness of keyword search systems has established an intriguing dichotomy within the landscape of information retrieval. While keyword-based methods excel in terms of speci- ficity, semantic models offer contextual richness and intent aware- ness. However, each paradigm faces challenges in harnessing the strengths of the other. This paper investigates a hybrid search model that integrates both keyword and semantic search techniques, aiming to optimize relevance, interpretability, and user experience in complex information-seeking scenarios. We present an innovative hybrid architecture, conduct a comprehen- sive evaluation of its performance, and discuss its implications for information retrieval in specialized domains such as healthcare and education. Index Terms—Hybrid Search, Keyword Search, Semantic Search, Information Retrieval, Natural Language Processing, Search Architectures

A High Scalability Memory NoC with Shared-Inside Hierarchical-Groupings for Triplet-Based Many-Core Architecture

Innovative processor architecture designs are shifting towards Many-Core Architectures (MCAs) to meet the future demands of high-performance computing as the limits of Moore’s Law have almost been reached. Many-core processors utilize shared memory hierarchies to achieve high-speed memory systems, improving memory access efficiency. However, as the number of cores multiplies, the scalability of this system is significantly constrained by the increased proportion of long-distance and Non-Uniform Memory Access (NUMA). Improving the scalability of MCAs is crucial for achieving large/super-scale general-purpose many-core processors. This work proposes a high scalability memory Network-on-Chip (NoC) for Triplet-Based Many-Core Architecture (TriBA), named TriBA-mNoC. TriBA-mNoC maintains a consistent core-to-core spacing as the network scale increases, effectively preventing increased long-distance memory access latency. Moreover, it leverages an inherent advantage of shared-inside hierarchical-groupings, alleviating common NUMA issues in the NoC design. Evaluations of static network characteristics show that TriBA-mNoC outperforms most classical NoCs in network diameter, average distance, and cost. TriBA-mNoC can be integrated with TriBA in the same silicon die with a tile-like floorplan, forming a novel NoC called TriBA-NoC, which can combine the strengths of both networks to maximize the architecture performance. We evaluated the memory access performance and scalability of TriBA-NoC using the mathematical evaluation models and actual simulations with real traffic (PARSEC 3.0 and SPLASH-2) at different network scales. The mathematical evaluation results indicate that TriBA-NoC achieves an aggregate speedup of approximately 3x compared with 2D-Mesh for a similar number of cores. Furthermore, TriBA-NoC’s single-core speedup efficiency remains stable as the number of cores increases under the same cache hit ratio, while 2D-Mesh experiences a rapid decline, highlighting TriBA-NoC’s exceptional scalability. Finally, the actual traffic simulation results show that TriBA-NoC achieves an average memory access latency and time reduction of 25.90% − 40.50% and 5.61% − 31.69% respectively, compared with 2D-Mesh.

Microstructural properties of Asian hornet nest paper-like materials: Preliminary step towards biomimicry materials for civil engineering applications

This paper presents a comprehensive examination of the microstructure and mineralogy of the paper-like material found in Asian hornet nests (referred to as AHN P-LM) and the nearby tree leaves, utilising scanning electron microscopy with energy dispersive spectroscopy (SEM-EDS) and X-ray diffraction (XRD). The analysis conducted through SEM-EDS demonstrates that AHN P-LM is primarily composed of plant fragments, with slight traces of inorganic substances (such as CaOx). The thread-like morphologies observed in the SEM analysis, which are linked to the secretion of hornets, warrant attention due to their significance in understanding the microstructure of AHN P-LM. The presence of nitrogen (N) in the EDS analysis of AHN P-LM, in contrast to its absence in analysed leaves, strongly implies that the nitrogen originates from hornet saliva. Additionally, SEM-EDS analysis revealed the impact of chemical composition variations on colour differences within AHN P-LM. Furthermore, X-ray diffraction analysis on ash samples from AHN P-LM confirmed the presence of minerals such as quartz, lime, phlogopite, and microcline. These findings about mineral composition align closely with results obtained from SEM-EDS analyses, presenting various aspects related to understanding the structure and compositional makeup of AHN P-LM. Biomimicry principles can be applied to draw inspiration from the AHN P-LM for replication in the civil engineering field, facilitating innovative solutions. Understanding hornet behaviour and nest material composition is essential. Future outlook includes utilising these materials in wood construction, 3D printing for structural components, shelters, infrastructure repair, green building, and architectural innovation.

From One-dimensional to Multidimensional Map Neural Networks

In the contemporary landscape, artificial neural networks transcend their traditional role of function approximation and have found diverse applications in fields such as image classification, machine translation, speech recognition, and natural language processing. In some datasets, traditional architectures exhibit low test and training accuracy, coupled with high loss and prolonged training times. This study aims to introduce innovative neural network architectures that outperform conventional models. The research presents a novel framework integrating one-dimensional and multidimensional map neural networks, consisting of three distinct architectures: 1D-Map, 2D-Map, and 3D-Map. A systematic performance comparison with traditional models is conducted following the implementation of these architectures. The evaluation spans four datasets, encompassing the domains of heat treatment of electroless Ni-P nano coatings, letter recognition, combined cycle power plants, and Seoul bike sharing demand. In the heat treatment dataset, the proposed 3D-Map architecture reached 0.055 higher average test accuracy than traditional MLP architecture. In the letter recognition dataset, the test accuracy of 3D-Map architecture was 0.0523 higher than the test accuracy of the LSTM architecture. In the combined cycle power plant dataset, the test accuracy of the 3D-Map architecture was 0.0612 more than the test accuracy of the MLP architecture. In the Seoul bike-sharing demand dataset, the test accuracy of the 2D-Map architecture was 0.0696 higher than the test accuracy of the LSTM architecture. The study's findings underscore the consistently superior performance of the proposed architectures compared to their traditional counterparts.

An Innovative Honeypot Architecture for Detecting and Mitigating Hardware Trojans in IoT Devices

The exponential growth and widespread adoption of Internet of Things (IoT) devices have introduced many vulnerabilities. Attackers frequently exploit these flaws, necessitating advanced technological approaches to protect against emerging cyber threats. This paper introduces a novel approach utilizing hardware honeypots as an additional defensive layer against hardware vulnerabilities, particularly hardware Trojans (HTs). HTs pose significant risks to the security of modern integrated circuits (ICs), potentially causing operational failures, denial of service, or data leakage through intentional modifications. The proposed system was implemented on a Raspberry Pi and tested on an emulated HT circuit using a Field-Programmable Gate Array (FPGA). This approach leverages hardware honeypots to detect and mitigate HTs in the IoT devices. The results demonstrate that the system effectively detects and mitigates HTs without imposing additional complexity on the IoT devices. The Trojan-agnostic solution offers full customization to meet specific security needs, providing a flexible and robust layer of security. These findings provide valuable insights into enhancing the security of IoT devices against hardware-based cyber threats, thereby contributing to the overall resilience of IoT networks. This innovative approach offers a promising solution to address the growing security challenges in IoT environments.

Integrated Pockels Modulators on Silicon Photonics Platform

AbstractElectro‐optic (EO) modulators are essential components in various fields, including optical communication, free‐space communication, microwave photonics, sensing, and light detection and ranging. The EO modulation enables the fast conversion of electric signals into optical signals, facilitating the precise manipulation of light. With advancements in fabrication processing techniques, next‐generation integrated EO modulators have demonstrated substantial improvements in modulation efficiency, bandwidth, and footprint. Here, the latest research progress in integrated EO modulation, focusing on the principle of the Pockels effect, key modulation metrics, novel EO thin‐film material platforms, and innovative device architectures is overviewed. Finally, it is evaluated different schemes and provide perspectives on future trends in developing integrated EO modulators, highlighting both the advantages and challenges of integrated EO modulation, including waveguide and electrode engineering, integrated methods, and other applications for large‐scale photonic integrated circuits.

Mechanical equipment fault diagnosis method based on improved deep residual shrinkage network.

Fault diagnosis of mechanical equipment can effectively reduce property losses and casualties. Bearing vibration signals, as one of the effective sources of diagnostic information, are often overwhelmed by substantial environmental noise. To address this issue, we present a fault diagnosis method, CCSDRSN, which exhibits strong noise resistance. This method enhances the soft threshold function in the traditional deep residual shrinkage network, allowing it to extract useful information from the fault signal to the maximum extent, thus significantly improving diagnostic accuracy. Additionally, we have developed a novel activation function that can nonlinearly transform the time frequency map across multiple dimensions and the entire region. In pursuit of network optimization and parameter reduction, we have strategically incorporated depthwise separable convolutions, effectively replacing conventional convolutional layers. This architectural innovation streamlines the network. By verifying the effectiveness of the proposed method using Case Western Reserve University datasets, the results demonstrate that the proposed method not only possesses strong noise resistance in high noise environments but also achieves high diagnostic accuracy and good generalization performance under different load conditions.

Neural differentiable modeling with diffusion-based super-resolution for two-dimensional spatiotemporal turbulence

Simulating spatiotemporal turbulence with high fidelity remains a cornerstone challenge in computational fluid dynamics (CFD) due to its intricate multiscale nature and prohibitive computational demands. Traditional approaches typically employ closure models, which attempt to represent small-scale features in an unresolved manner. However, these methods often sacrifice accuracy and lose high-frequency/wavenumber information, especially in scenarios involving complex flow physics. In this paper, we introduce an innovative neural differentiable modeling framework designed to enhance the predictability and efficiency of spatiotemporal turbulence simulations. Our approach features differentiable hybrid modeling techniques that seamlessly integrate deep neural networks with numerical PDE solvers within a differentiable programming framework, synergizing deep learning with physics-based CFD modeling. Specifically, a hybrid differentiable neural solver is constructed on a coarser grid to capture large-scale turbulent phenomena, followed by the application of a Bayesian conditional diffusion model that generates small-scale turbulence conditioned on large-scale flow predictions. Two innovative hybrid architecture designs are studied, and their performance is evaluated through comparative analysis against conventional large eddy simulation techniques with physics-based subgrid-scale closures and purely data-driven neural solvers. The findings underscore the potential of the neural differentiable modeling framework to significantly enhance the accuracy and computational efficiency of turbulence simulations. This study not only demonstrates the efficacy of merging deep learning with physics-based numerical solvers but also sets a new precedent for advanced CFD modeling techniques, highlighting the transformative impact of differentiable programming in scientific computing.

Mapping Data to Concepts: Enhancing Quantum Neural Network Transparency with Concept-Driven Quantum Neural Networks

We introduce the concept-driven quantum neural network (CD-QNN), an innovative architecture designed to enhance the interpretability of quantum neural networks (QNNs). CD-QNN merges the representational capabilities of QNNs with the transparency of self-explanatory models by mapping input data into a human-understandable concept space and making decisions based on these concepts. The algorithmic design of CD-QNN is comprehensively analyzed, detailing the roles of the concept generator, feature extractor, and feature integrator in improving and balancing model expressivity and interpretability. Experimental results demonstrate that CD-QNN maintains high predictive accuracy while offering clear and meaningful explanations of its decision-making process. This paradigm shift in QNN design underscores the growing importance of interpretability in quantum artificial intelligence, positioning CD-QNN and its derivative technologies as pivotal in advancing reliable and interpretable quantum intelligent systems for future research and applications.

Hyperspectral Image Classification Algorithm for Forest Analysis Based on a Group-Sensitive Selective Perceptual Transformer

Substantial advancements have been achieved in hyperspectral image (HSI) classification through contemporary deep learning techniques. Nevertheless, the incorporation of an excessive number of irrelevant tokens in large-scale remote sensing data results in inefficient long-range modeling. To overcome this hurdle, this study introduces the Group-Sensitive Selective Perception Transformer (GSAT) framework, which builds upon the Vision Transformer (ViT) to enhance HSI classification outcomes. The innovation of the GSAT architecture is primarily evident in several key aspects. Firstly, the GSAT incorporates a Group-Sensitive Pixel Group Mapping (PGM) module, which organizes pixels into distinct groups. This allows the global self-attention mechanism to function within these groupings, effectively capturing local interdependencies within spectral channels. This grouping tactic not only boosts the model’s spatial awareness but also lessens computational complexity, enhancing overall efficiency. Secondly, the GSAT addresses the detrimental effects of superfluous tokens on model efficacy by introducing the Sensitivity Selection Framework (SSF) module. This module selectively identifies the most pertinent tokens for classification purposes, thereby minimizing distractions from extraneous information and bolstering the model’s representational strength. Furthermore, the SSF refines local representation through multi-scale feature selection, enabling the model to more effectively encapsulate feature data across various scales. Additionally, the GSAT architecture adeptly represents both global and local features of HSI data by merging global self-attention with local feature extraction. This integration strategy not only elevates classification precision but also enhances the model’s versatility in navigating complex scenes, particularly in urban mapping scenarios where it significantly outclasses previous deep learning methods. The advent of the GSAT architecture not only rectifies the inefficiencies of traditional deep learning approaches in processing extensive remote sensing imagery but also markededly enhances the performance of HSI classification tasks through the deployment of group-sensitive and selective perception mechanisms. It presents a novel viewpoint within the domain of hyperspectral image classification and is poised to propel further advancements in the field. Empirical testing on six standard HSI datasets confirms the superior performance of the proposed GSAT method in HSI classification, especially within urban mapping contexts, where it exceeds the capabilities of prior deep learning techniques. In essence, the GSAT architecture markedly refines HSI classification by pioneering group-sensitive pixel group mapping and selective perception mechanisms, heralding a significant breakthrough in hyperspectral image processing.

Pt/ZnO and Pt/few-layer graphene/ZnO Schottky devices with Al ohmic contacts using Atlas simulation and machine learning

The search for highly efficient photodetectors, driven by various applications ranging from environmental monitoring to communication systems and imaging, continues to drive research into novel materials and innovative device architectures. This paper offers an in-depth comparative analysis to optimize the performance of two types of Schottky ultraviolet (UV) photodetectors: platinum (Pt)/zinc oxide (ZnO) and Pt/few-layer graphene (FLG)/ZnO, both featuring aluminum (Al) ohmic contacts. The study systematically examines and compares the electrical and optical characteristics of these two photodetector configurations using the Silvaco TCAD simulation tool and machine learning regression models. The research outcomes demonstrate that the proposed photodetector configurations exhibit remarkable improvements, showcasing their substantial potential for superior performance in UV applications. The Pt/ZnO photodetector demonstrates a dark current density of 8.2 × 10−11 pA/cm2, a photocurrent density of 0.26 μA/cm2, a 3-dB cut-off frequency of 85.4 GHz, an external quantum efficiency of 90.41%, and an external photocurrent responsivity of 0.26A/W, at a bias of −1.0 V. On the other hand, Pt/FLG/ZnO photodetector demonstrates near zero dark current density, a photocurrent density of 0.2 μA/cm2, a 3-dB cut-off frequency of 2.44 THz, an external quantum efficiency of 68.52%, and an external photocurrent responsivity of 0.2 A/W at a bias of −1.0 V. Furthermore, a comprehensive comparative analysis of various machine-learning regression models is conducted, validating the simulation findings and providing a predictive framework for optimizing the photodetector's performance. Each machine-learning regression model is evaluated by getting root mean squared error and R2 values across different test set sizes to assess their accuracy in predicting the photodetector's characteristics. This study underscores the promising role of cutting-edge materials and computational techniques in advancing the development of next-generation optoelectronic devices with enhanced capabilities and performance.

The Impact and Future of Generative Pre-trained Transformers: A Study of GPT in Enhancing Business and Technology

With the development of artificial intelligence (AI), natural language processing (NIP) has clearly become the key to enhancing AI capabilities, especially through the deployment of GPT. This paper explores the major advances and related commercial applications of GPT, as it is at the cutting edge of AI technology with far-reaching capabilities in terms of text generation, language understanding, and human-machine interaction. Developed by OpenAI, GPT uses innovative transformer architecture to make breakthroughs in a variety of language tasks, marking a major evolution in the potential applications of AI in various industries. This research explores the integration of GPT in sectors, highlighting how its complex text processing capabilities can be utilized to improve efficiency and innovative service delivery. The study highlights the transformative impact of GPT on business efficiency, strategic decision-making, and customer engagement, thanks to its scalable and adaptable AI framework. Finally, the document reflects the need for a strong ethical framework and regulatory supervision to guide the deployment of GPT technologies. As the Global Trade Agreement evolves, it is committed to shaping the future of global industry by promoting innovation while calling for a balanced approach to ethical considerations and social impacts.

Andrey Kryachkov. The man who created the city

Andrey Dmitrievich Kryachkov is an outstanding architect, the founder of the Siberian architectural school, whose creative path is closely connected with the foundation of the future administrative and cultural capital of Siberia. The author of most of the iconic historical architectural objects of modern Novosibirsk, Kryachkov largely predetermined the aesthetic code of the young Siberian city at that time, combining traditions and innovations in regional architecture of the first half of the 20th century. The architecture of Kryachkov in the modern appearance of Novosibirsk continues to play a significant city-forming aesthetic role.

DEEP LEARNING APPROACH FOR ASPECT CATEGORY DETECTION: A BIBLIOMETRIC ANALYSIS

This article presents a quantitative and qualitative assessment of current research trends by conducting a bibliometric analysis of the sentiment analysis literature from 2020 to March 2024 using the Scopus database. Our focus is on the review of scientific documents, the arrangement of subject categories, the research trend in aspect category detection, the top 10 scholars that write the most number articles in aspect category detection and keyword trends. Our research shows that specialists in computer science, engineering, mathematics, medicine, decision science, material science, social sciences, business and management accounting, energy, and health are the most common topic groups in this industry. A study of keywords shows that terms like “BERT” and “deep learning” are frequently used together. This highlights the use of sophisticated models like BERT in this field and suggests a tendency to use innovative architecture to achieve better results. Conversely, although terms such as “sentiment analysis” and “aspect-based sentiment analysis” have modest frequency, their link strengths indicate a significant correlation with the main theme, emphasising the relationship between aspect category detection and sentiment analysis in research projects. we also provide the deep-learning technique used by the author for aspect category detection.

Selective and multi-scale fusion Mamba for medical image segmentation

Given the high variability in the morphology and size of lesion areas in medical images, accurate medical image segmentation requires both precise positioning of global contours and careful processing of local boundaries. This emphasizes the importance of fusing multi-scale local and global features. However, existing CNN and Transformer-based models are often limited by high parameter counts and complex calculations, making it difficult to efficiently integrate these features. To address this challenge, we proposed two innovative optimization architectures: Selective Fusion Mamba (SF-Mamba) and Multi-Scale Fusion Mamba (MF-Mamba). SF-Mamba can flexibly and dynamically adjust the fusion strategy of local and global features according to the characteristics of the lesions, effectively handling segmentation tasks with variable morphology. MF-Mamba enhances the model’s segmentation ability for lesions of different sizes by capturing global information across scales. Based on these two structures, we constructed a lightweight SMM-UNet model, which not only significantly reduces the computational burden (with only 0.038M parameters) but also demonstrates excellent generalization ability and can efficiently adapt to various types of medical images. Extensive tests on the ISIC2017, ISIC2018, and BUSI public datasets show that SMM-UNet achieves excellent segmentation performance with an extremely low parameter cost.