Transfer_Learning
-
From scratch to silver: Creating trustworthy training data for patent-SDG classification using Large Language Models
2025-09-11 09:44:16Grazia Sveva Ascione, Nicol\`o TamagnoneAbstract
Classifying patents by their relevance to the UN Sustainable Development Goals (SDGs) is crucial for tracking how innovation addresses global challenges. However, the absence of a large, labeled dataset limits the use of supervised learning. Existing methods, such as keyword searches, transfer learning, and citation-based heuristics, lack scalability and generalizability. This paper frames patent-to-SDG classification as a weak supervision problem, using citations from patents to SDG-tagged scientific publications (NPL citations) as a noisy initial signal. To address its sparsity and noise, we develop a composite labeling function (LF) that uses large language models (LLMs) to extract structured concepts, namely functions, solutions, and applications, from patents and SDG papers based on a patent ontology. Cross-domain similarity scores are computed and combined using a rank-based retrieval approach. The LF is calibrated via a custom positive-only loss that aligns with known NPL-SDG links without penalizing discovery of new SDG associations. The result is a silver-standard, soft multi-label dataset mapping patents to SDGs, enabling the training of effective multi-label regression models. We validate our approach through two complementary strategies: (1) internal validation against held-out NPL-based labels, where our method outperforms several baselines including transformer-based models, and zero-shot LLM; and (2) external validation using network modularity in patent citation, co-inventor, and co-applicant graphs, where our labels reveal greater thematic, cognitive, and organizational coherence than traditional technological classifications. These results show that weak supervision and semantic alignment can enhance SDG classification at scale.
Abstract (translated)
根据联合国可持续发展目标（SDGs）的相关性对专利进行分类对于跟踪创新如何应对全球挑战至关重要。然而，缺乏大规模、标注的数据集限制了监督学习的应用。现有的方法，如关键词搜索、迁移学习和基于引用的启发式算法，在可扩展性和泛化能力方面存在不足。本文将专利到SDG的分类视为一个弱监督问题，并利用专利对带有SDG标签的科学出版物（非专利文献或NPL）的引用作为初始信号，尽管这些信号可能有噪声。 为了应对其稀疏性和噪音，我们开发了一个复合标记函数（LF），该函数使用大型语言模型（LLMs）根据专利语义学从专利和SDG论文中提取结构化的概念，例如功能、解决方案和应用。通过基于排名的检索方法计算跨域相似度分数，并将其组合起来。我们的标记函数通过一个定制的仅正标签损失进行校准，该损失与已知的NPL-SDG链接保持一致，同时不会惩罚新SDG关联的发现。 结果是一个银标准、软多标签数据集，将专利映射到SDGs，使有效的多标签回归模型得以训练。我们通过两种互补的方法验证了我们的方法：（1）内部验证，即与保留的NPL基线标签进行比较，在这种方法中，我们的方法优于包括基于变压器的模型和零样本LLM在内的多种基准；以及（2）外部验证使用专利引用、共同发明人和联合申请人网络中的模块性，其中我们的标签显示了比传统技术分类更高的主题、认知和组织一致性。这些结果表明，弱监督和语义对齐可以在大规模上增强SDG分类的准确性与有效性。
URL
https://arxiv.org/abs/2509.09303
PDF
-
Multi-Label Transfer Learning in Non-Stationary Data Streams
2025-09-09 23:01:20Honghui Du, Leandro Minku, Aonghus Lawlor, Huiyu ZhouAbstract
Label concepts in multi-label data streams often experience drift in non-stationary environments, either independently or in relation to other labels. Transferring knowledge between related labels can accelerate adaptation, yet research on multi-label transfer learning for data streams remains limited. To address this, we propose two novel transfer learning methods: BR-MARLENE leverages knowledge from different labels in both source and target streams for multi-label classification; BRPW-MARLENE builds on this by explicitly modelling and transferring pairwise label dependencies to enhance learning performance. Comprehensive experiments show that both methods outperform state-of-the-art multi-label stream approaches in non-stationary environments, demonstrating the effectiveness of inter-label knowledge transfer for improved predictive performance.
Abstract (translated)
在多标签数据流中，标签概念往往会在非平稳环境中经历漂移，这种漂移可能是独立发生的，也可能是与其他标签相关的。利用相关标签之间的知识转移可以加速适应过程，但针对数据流的多标签迁移学习研究仍然有限。为了解决这一问题，我们提出了两种新颖的迁移学习方法：BR-MARLENE 方法从源和目标数据流中的不同标签中借用知识进行多标签分类；而 BRPW-MARLENE 则在此基础上进一步显式地建模并转移成对标签依赖性，以增强学习性能。全面的实验表明，在非平稳环境中，这两种方法均优于最先进的多标签流处理方法，证明了标签间知识迁移在提高预测性能方面的有效性。
URL
https://arxiv.org/abs/2509.08181
PDF
-
MARLINE: Multi-Source Mapping Transfer Learning for Non-Stationary Environments
2025-09-09 22:51:31Honghui Du, Leandro Minku, Huiyu ZhouAbstract
Concept drift is a major problem in online learning due to its impact on the predictive performance of data stream mining systems. Recent studies have started exploring data streams from different sources as a strategy to tackle concept drift in a given target domain. These approaches make the assumption that at least one of the source models represents a concept similar to the target concept, which may not hold in many real-world scenarios. In this paper, we propose a novel approach called Multi-source mApping with tRansfer LearnIng for Non-stationary Environments (MARLINE). MARLINE can benefit from knowledge from multiple data sources in non-stationary environments even when source and target concepts do not match. This is achieved by projecting the target concept to the space of each source concept, enabling multiple source sub-classifiers to contribute towards the prediction of the target concept as part of an ensemble. Experiments on several synthetic and real-world datasets show that MARLINE was more accurate than several state-of-the-art data stream learning approaches.
Abstract (translated)
概念漂移是在线学习中的一个主要问题，因为它会影响数据流挖掘系统的预测性能。最近的研究已经开始探索来自不同来源的数据流作为解决特定目标领域中概念漂移的策略。这些方法假设至少有一个源模型代表了与目标概念相似的概念，但这一假设在许多现实场景中可能并不成立。本文提出了一种名为多源映射和迁移学习用于非平稳环境（MARLINE）的新方法。即使当源概念与目标概念不匹配时，MARLINE也能从多个数据源的知识中受益，在非平稳环境中发挥作用。这是通过将目标概念投影到每个源概念的空间来实现的，从而使多个源子分类器能够作为集成的一部分共同预测目标概念。在几个合成和现实世界的数据集上的实验表明，MARLINE比几种最新的数据流学习方法更准确。
URL
https://arxiv.org/abs/2509.08176
PDF
-
Two-Stage Swarm Intelligence Ensemble Deep Transfer Learning for Vehicle Detection Using Unmanned Aerial Vehicles
2025-09-09 13:03:12Zeinab Ghasemi Darehnaei, Mohammad Shokouhifar, Hossein Yazdanjouei, S. M. J. Rastegar FatemiAbstract
This paper introduces SI-EDTL, a two-stage swarm intelligence ensemble deep transfer learning model for detecting multiple vehicles in UAV images. It combines three pre-trained Faster R-CNN feature extractor models (InceptionV3, ResNet50, GoogLeNet) with five transfer classifiers (KNN, SVM, MLP, C4.5, Naïve Bayes), resulting in 15 different base learners. These are aggregated via weighted averaging to classify regions as Car, Van, Truck, Bus, or background. Hyperparameters are optimized with the whale optimization algorithm to balance accuracy, precision, and recall. Implemented in MATLAB R2020b with parallel processing, SI-EDTL outperforms existing methods on the AU-AIR UAV dataset.
Abstract (translated)
这篇论文介绍了SI-EDTL，这是一种两阶段的群智能集成深度迁移学习模型，用于检测无人机图像中的多个车辆。该模型结合了三个预训练的Faster R-CNN特征提取器模型（InceptionV3、ResNet50和GoogLeNet）与五种迁移分类器（KNN、SVM、MLP、C4.5和朴素贝叶斯），共生成15个不同的基础学习器。这些基础学习器通过加权平均聚合，以将区域分类为小汽车、面包车、卡车、公共汽车或背景。该模型使用鲸鱼优化算法来调整超参数，从而在准确率、精确度和召回率之间取得平衡。SI-EDTL在MATLAB R2020b中实现，并采用并行处理，在AU-AIR无人机数据集上优于现有方法。
URL
https://arxiv.org/abs/2509.08026
PDF
-
Enhancing Privacy Preservation and Reducing Analysis Time with Federated Transfer Learning in Digital Twins-based Computed Tomography Scan Analysis
2025-09-09 04:54:08Avais Jan, Qasim Zia, Murray PattersonAbstract
The application of Digital Twin (DT) technology and Federated Learning (FL) has great potential to change the field of biomedical image analysis, particularly for Computed Tomography (CT) scans. This paper presents Federated Transfer Learning (FTL) as a new Digital Twin-based CT scan analysis paradigm. FTL uses pre-trained models and knowledge transfer between peer nodes to solve problems such as data privacy, limited computing resources, and data heterogeneity. The proposed framework allows real-time collaboration between cloud servers and Digital Twin-enabled CT scanners while protecting patient identity. We apply the FTL method to a heterogeneous CT scan dataset and assess model performance using convergence time, model accuracy, precision, recall, F1 score, and confusion matrix. It has been shown to perform better than conventional FL and Clustered Federated Learning (CFL) methods with better precision, accuracy, recall, and F1-score. The technique is beneficial in settings where the data is not independently and identically distributed (non-IID), and it offers reliable, efficient, and secure solutions for medical diagnosis. These findings highlight the possibility of using FTL to improve decision-making in digital twin-based CT scan analysis, secure and efficient medical image analysis, promote privacy, and open new possibilities for applying precision medicine and smart healthcare systems.
Abstract (translated)
数字孪生（DT）技术和联邦学习（FL）的应用在生物医学图像分析领域，特别是计算机断层扫描（CT）方面具有巨大的变革潜力。本文提出了一种基于数字孪生的CT扫描分析新范式——联合迁移学习（FTL）。该方法利用预训练模型和同级节点之间的知识转移来解决数据隐私、计算资源有限以及数据异质性等问题。所提出的框架允许云端服务器与具备数字孪生功能的CT扫描仪进行实时协作，同时保护患者身份信息。 我们将FTL方法应用于一个异构CT扫描数据集，并通过收敛时间、模型精度、精确度、召回率、F1分数和混淆矩阵来评估模型性能。结果显示，该方法在精度、准确性和召回率方面优于传统的FL和集群联邦学习（CFL）方法，并且具有更高的F1分数。 当数据分布不是独立同质时，此技术尤为有益。它提供了可靠、高效且安全的解决方案用于医学诊断。这些发现强调了使用FTL来改善基于数字孪生的CT扫描分析决策制定的可能性，以及在确保隐私和提高效率的同时进行安全高效的医疗图像分析，并为精准医疗和智慧医疗系统的应用开辟新的可能。
URL
https://arxiv.org/abs/2509.08018
PDF
-
Neurocognitive Modeling for Text Generation: Deep Learning Architecture for EEG Data
2025-09-08 20:32:41KhushiyantAbstract
Text generating capabilities have undergone a substantial transformation with the introduction of large language models (LLMs). Electroencephalography (EEG)-based text production is still difficult, though, because it requires a lot of data and processing power. This paper introduces a new method that combines the use of the Gemma 2B LLM with a classifier-LLM architecture to incorporate a Recurrent Neural Network (RNN) encoder. Our approach drastically lowers the amount of data and compute power needed while achieving performance close to that of cutting-edge methods. Notably, compared to current methodologies, our methodology delivers an overall performance improvement of 10%. The suggested architecture demonstrates the possibility of effective transfer learning for EEG-based text production, remaining strong and functional even in the face of data limits. This work highlights the potential of integrating LLMs with EEG decoding to improve assistive technologies and improve independence and communication for those with severe motor limitations. Our method pushes the limits of present capabilities and opens new paths for research and application in brain-computer interfaces by efficiently using the strengths of pre-trained language models. This makes EEG-based text production more accessible and efficient.
Abstract (translated)
文本生成能力随着大型语言模型（LLM）的出现经历了重大变革。然而，基于脑电图（EEG）的文本生产仍然面临挑战，因为这需要大量的数据和计算资源。本文介绍了一种新的方法，该方法结合使用了Gemini 2B LLM与分类器-LLM架构，并引入了一个循环神经网络（RNN）编码器。我们的方法显著降低了所需的数据量和计算能力，同时实现了接近前沿技术的性能表现。值得注意的是，相比当前的方法，我们提出的方法总体性能提高了10%。 所提出的架构展示了基于EEG文本生成的有效迁移学习的可能性，在数据有限的情况下依然保持强大且功能完备。这项工作强调了将LLM与EEG解码相结合以改进辅助技术和提升严重运动受限人士的独立性和沟通能力的巨大潜力。我们的方法突破了现有技术的能力极限，通过充分利用预训练语言模型的优势，为脑机接口的研究和应用开辟了新的路径，从而使得基于EEG的文本生成更加可访问且高效。
URL
https://arxiv.org/abs/2509.07202
PDF
-
Leveraging Generic Foundation Models for Multimodal Surgical Data Analysis
2025-09-08 16:04:19Simon Pezold, J\'er\^ome A. Kurylec, Jan S. Liechti, Beat P. M\"uller, Jo\"el L. LavanchyAbstract
We investigate how both the adaptation of a generic foundation model via transfer learning and the integration of complementary modalities from the operating room (OR) can support surgical data science. To this end, we use V-JEPA as the single-modality foundation of a multimodal model for minimally invasive surgery support. We analyze how the model's downstream performance can benefit (a) from finetuning on unlabeled surgical video data and (b) from providing additional time-resolved data streams from the OR in a multimodal setup. In an in-house dataset of liver surgery videos, we analyze the tasks of predicting hospital length of stay and postoperative complications. In videos of the public HeiCo dataset, we analyze the task of surgical phase recognition. As a baseline, we apply pretrained V-JEPA to all tasks. We then finetune it on unlabeled, held-out videos to investigate its change in performance after domain adaptation. Following the idea of modular decision support networks, we integrate additional data streams from the OR by training a separate encoder to form a shared representation space with V-JEPA's embeddings. Our experiments show that finetuning on domain-specific data increases model performance. On the in-house data, integrating additional time-resolved data likewise benefits the model. On the HeiCo data, accuracy of the pretrained video-only, single-modality baseline setup is on par with the top-performing submissions of the EndoVis2017 challenge, while finetuning on domain-specific data increases accuracy further. Our results thus demonstrate how surgical data science can leverage public, generic foundation models. Likewise, they indicate the potential of domain adaptation and of integrating suitable complementary data streams from the OR. To support further research, we release our code and model weights at this https URL.
Abstract (translated)
我们研究了通过迁移学习调整通用基础模型以及整合来自手术室（OR）的互补模态如何支持外科数据科学。为此，我们将V-JEPA用作微创手术支持多模态模型的基础单模态模型。我们分析了该模型在以下两种情况下的下游性能如何受益：(a) 在未标注的手术视频数据上进行微调；以及(b) 通过提供来自OR的时间序列附加数据流，在多模态设置下运行。在一个内部肝脏手术视频数据集中，我们分析了预测住院时间长短和术后并发症的任务。在公共HeiCo数据集中的视频中，我们分析了识别外科阶段的任务。作为基线，我们将预训练的V-JEPA应用于所有任务。然后我们在未标注、保留的数据上对其进行微调，以研究其在领域适应后的性能变化。借鉴模块化决策支持网络的理念，通过训练一个单独的编码器来形成与V-JEPA嵌入共享的表示空间的方式，我们整合了来自OR的附加数据流。 我们的实验表明，在特定领域的数据上进行微调可以提高模型的表现。在内部数据集上，集成额外的时间序列数据同样有助于提升模型性能。而在HeiCo数据集上，预训练视频单一模态基线设置的准确性与EndoVis2017挑战赛中最佳提交结果相当，而针对特定领域数据进行微调可以进一步提高精度。 因此，我们的研究结果展示了外科数据科学如何利用公共、通用的基础模型。同时，它们表明了域适应和整合来自OR的合适互补数据流的潜在价值。为了支持后续的研究工作，我们在以下网址发布了我们的代码和模型权重：[此处应填写具体URL]。
URL
https://arxiv.org/abs/2509.06831
PDF
-
Pothole Detection and Recognition based on Transfer Learning
2025-09-08 14:40:16Mang Hu, Qianqian XiaAbstract
With the rapid development of computer vision and machine learning, automated methods for pothole detection and recognition based on image and video data have received significant attention. It is of great significance for social development to conduct an in-depth analysis of road images through feature extraction, thereby achieving automatic identification of the pothole condition in new images. Consequently, this is the main issue addressed in this study. Based on preprocessing techniques such as standardization, normalization, and data augmentation applied to the collected raw dataset, we continuously improved the network model based on experimental results. Ultimately, we constructed a deep learning feature extraction network ResNet50-EfficientNet-RegNet model based on transfer learning. This model exhibits high classification accuracy and computational efficiency. In terms of model evaluation, this study employed a comparative evaluation approach by comparing the performance of the proposed transfer learning model with other models, including Random Forest, MLP, SVM, and LightGBM. The comparison analysis was conducted based on metrics such as Accuracy, Recall, Precision, F1-score, and FPS, to assess the classification performance of the transfer learning model proposed in this paper. The results demonstrate that our model exhibits high performance in terms of recognition speed and accuracy, surpassing the performance of other models. Through careful parameter selection and model optimization, our transfer learning model achieved a classification accuracy of 97.78% (88/90) on the initial set of 90 test samples and 98.89% (890/900) on the expanded test set.
Abstract (translated)
随着计算机视觉和机器学习的迅速发展，基于图像和视频数据自动检测和识别路面坑洞的方法受到了广泛关注。通过特征提取对道路图像进行深入分析，从而实现新图像中路面状况的自动化识别，这对社会发展具有重要意义，这也是本研究的主要关注点。 在收集到的原始数据集基础上，我们应用了诸如标准化、归一化和数据增强等预处理技术，并根据实验结果不断改进网络模型。最终，基于迁移学习构建了一个深度学习特征提取网络ResNet50-EfficientNet-RegNet模型。该模型表现出高分类准确率和计算效率。 在模型评估方面，本研究采用了一种比较评价方法，将提出的迁移学习模型与随机森林、多层感知器(MLP)、支持向量机(SVM)以及LightGBM等其他模型进行性能对比分析，并基于准确性(Accuracy)、召回率(Recall)、精确度(Precision)、F1值(F1-score)和FPS(Frame Per Second，每秒帧数)等指标来评估本文提出的迁移学习模型的分类性能。实验结果表明，我们的模型在识别速度和准确率方面表现出色，优于其他模型的表现。 通过精心选择参数并优化模型，我们的迁移学习模型在最初的90个测试样本中达到了97.78%（88/90）的分类准确率，并且在扩展后的测试集(900个样本)上实现了高达98.89% (890/900) 的分类准确度。
URL
https://arxiv.org/abs/2509.06750
PDF
-
SAM$^{*}$: Task-Adaptive SAM with Physics-Guided Rewards
2025-09-08 13:51:20Kamyar Barakati, Utkarsh Pratiush, Sheryl L. Sanchez, Aditya Raghavan, Delia J. Milliron, Mahshid Ahmadi, Philip D. Rack, Sergei V. KalininAbstract
Image segmentation is a critical task in microscopy, essential for accurately analyzing and interpreting complex visual data. This task can be performed using custom models trained on domain-specific datasets, transfer learning from pre-trained models, or foundational models that offer broad applicability. However, foundational models often present a considerable number of non-transparent tuning parameters that require extensive manual optimization, limiting their usability for real-time streaming data analysis. Here, we introduce a reward function-based optimization to fine-tune foundational models and illustrate this approach for SAM (Segment Anything Model) framework by Meta. The reward functions can be constructed to represent the physics of the imaged system, including particle size distributions, geometries, and other criteria. By integrating a reward-driven optimization framework, we enhance SAM's adaptability and performance, leading to an optimized variant, SAM$^{*}$, that better aligns with the requirements of diverse segmentation tasks and particularly allows for real-time streaming data segmentation. We demonstrate the effectiveness of this approach in microscopy imaging, where precise segmentation is crucial for analyzing cellular structures, material interfaces, and nanoscale features.
Abstract (translated)
图像分割是显微镜分析中的关键任务，对于准确解析和解释复杂的视觉数据至关重要。这项任务可以通过使用针对特定领域训练的自定义模型、从预训练模型迁移学习或使用具有广泛适用性的基础模型来完成。然而，基础模型通常包含大量不透明的调优参数，这些参数需要进行大量的手动优化，从而限制了它们在实时流数据分析中的实用性。 在这里，我们介绍了一种基于奖励函数的优化方法，用于微调基础模型，并通过Meta的SAM（Segment Anything Model）框架来说明这种方法。奖励函数可以构造为表示成像系统的物理特性，包括粒子尺寸分布、几何形状和其他标准。通过集成驱动奖励的优化框架，我们可以增强SAM的适应性和性能，从而产生一种优化变体SAM$^{*}$，使其更符合各种分割任务的需求，并特别允许实时流数据分割。 我们在显微镜成像中展示了这种方法的有效性，在这种应用中，精确的分割对于分析细胞结构、材料界面和纳米级特征至关重要。
URL
https://arxiv.org/abs/2509.07047
PDF
-
QualityFM: a Multimodal Physiological Signal Foundation Model with Self-Distillation for Signal Quality Challenges in Critically Ill Patients
2025-09-08 10:20:56Zongheng Guo, Tao Chen, Manuela FerrarioAbstract
Photoplethysmogram (PPG) and electrocardiogram (ECG) are commonly recorded in intesive care unit (ICU) and operating room (OR). However, the high incidence of poor, incomplete, and inconsistent signal quality, can lead to false alarms or diagnostic inaccuracies. The methods explored so far suffer from limited generalizability, reliance on extensive labeled data, and poor cross-task transferability. To overcome these challenges, we introduce QualityFM, a novel multimodal foundation model for these physiological signals, designed to acquire a general-purpose understanding of signal quality. Our model is pre-trained on an large-scale dataset comprising over 21 million 30-second waveforms and 179,757 hours of data. Our approach involves a dual-track architecture that processes paired physiological signals of differing quality, leveraging a self-distillation strategy where an encoder for high-quality signals is used to guide the training of an encoder for low-quality signals. To efficiently handle long sequential signals and capture essential local quasi-periodic patterns, we integrate a windowed sparse attention mechanism within our Transformer-based model. Furthermore, a composite loss function, which combines direct distillation loss on encoder outputs with indirect reconstruction loss based on power and phase spectra, ensures the preservation of frequency-domain characteristics of the signals. We pre-train three models with varying parameter counts (9.6 M to 319 M) and demonstrate their efficacy and practical value through transfer learning on three distinct clinical tasks: false alarm of ventricular tachycardia detection, the identification of atrial fibrillation and the estimation of arterial blood pressure (ABP) from PPG and ECG signals.
Abstract (translated)
心电图（ECG）和光电容积描记图（PPG）是重症监护病房（ICU）和手术室（OR）中常见的记录生理信号。然而，这些信号往往出现质量差、不完整或一致性不佳的问题，这可能导致误报或诊断错误。迄今为止探索的方法存在泛化能力有限、依赖大量标注数据以及跨任务迁移性能较差等问题。为解决这些问题，我们引入了QualityFM，这是一种新型的多模态基础模型，专门针对生理信号的质量理解进行了设计。我们的模型在包含超过2100万份30秒波形和179,757小时数据的大规模数据集上进行了预训练。 我们的方法采用了一种双轨架构，用于处理质量不同的成对生理信号，并利用自蒸馏策略，即通过高质量信号的编码器来指导低质量信号的编码器的训练。为了高效地处理长序列信号并捕捉重要的局部准周期模式，我们在基于Transformer模型中集成了窗口稀疏注意力机制。 此外，我们采用了一种复合损失函数，该函数结合了直接在编码器输出上的蒸馏损失以及基于功率和相位谱的间接重建损失，以确保信号频域特性的保留。我们使用不同参数量（从9.6M到319M）预训练了三个模型，并通过三种不同的临床任务的迁移学习展示了其有效性和实用性：室性心动过速误报检测、心房颤动识别以及根据PPG和ECG信号估计动脉血压（ABP）。
URL
https://arxiv.org/abs/2509.06516
PDF
-
AI-Based Applied Innovation for Fracture Detection in X-rays Using Custom CNN and Transfer Learning Models
2025-09-07 22:30:25Amna Hassan, Ilsa Afzaal, Nouman Muneeb, Aneeqa Batool, Hamail NoorAbstract
Bone fractures present a major global health challenge, often resulting in pain, reduced mobility, and productivity loss, particularly in low-resource settings where access to expert radiology services is limited. Conventional imaging methods suffer from high costs, radiation exposure, and dependency on specialized interpretation. To address this, we developed an AI-based solution for automated fracture detection from X-ray images using a custom Convolutional Neural Network (CNN) and benchmarked it against transfer learning models including EfficientNetB0, MobileNetV2, and ResNet50. Training was conducted on the publicly available FracAtlas dataset, comprising 4,083 anonymized musculoskeletal radiographs. The custom CNN achieved 95.96% accuracy, 0.94 precision, 0.88 recall, and an F1-score of 0.91 on the FracAtlas dataset. Although transfer learning models (EfficientNetB0, MobileNetV2, ResNet50) performed poorly in this specific setup, these results should be interpreted in light of class imbalance and data set limitations. This work highlights the promise of lightweight CNNs for detecting fractures in X-rays and underscores the importance of fair benchmarking, diverse datasets, and external validation for clinical translation
Abstract (translated)
骨骼骨折是全球健康的一大挑战，常常导致疼痛、活动能力下降和生产力损失，尤其是在资源有限的环境中，那里获取专家放射学服务的机会较少。传统影像方法存在成本高、辐射暴露以及依赖专业解读的问题。为解决这些问题，我们开发了一种基于人工智能（AI）的方法，利用自定义卷积神经网络(CNN)从X射线图像中自动检测骨折，并将其与包括EfficientNetB0、MobileNetV2和ResNet50在内的迁移学习模型进行了对比测试。训练是在公开可用的FracAtlas数据集上进行的，该数据集包含4,083张匿名化的肌肉骨骼放射影像。自定义CNN在FracAtlas数据集上的准确率为95.96%，精确度为0.94，召回率（灵敏度）为0.88，F1分数达到0.91。尽管迁移学习模型在此特定设置中的表现不佳，但这些结果应该考虑到类别不平衡和数据集限制的影响来解读。这项工作突显了轻量级CNN在X射线骨折检测方面的潜力，并强调了公平基准测试、多样化数据集以及外部验证对于临床应用的重要性。
URL
https://arxiv.org/abs/2509.06228
PDF
-
A Fine-Grained Attention and Geometric Correspondence Model for Musculoskeletal Risk Classification in Athletes Using Multimodal Visual and Skeletal Features
2025-09-07 04:09:06Md. Abdur Rahman, Mohaimenul Azam Khan Raiaan, Tamanna Shermin, Md Rafiqul Islam, Mukhtar Hussain, Sami AzamAbstract
Musculoskeletal disorders pose significant risks to athletes, and assessing risk early is important for prevention. However, most existing methods are designed for controlled settings and fail to reliably assess risk in complex environments due to their reliance on a single type of data. This research proposes ViSK-GAT (Visual-Skeletal Geometric Attention Transformer), a novel multimodal deep learning framework designed to classify musculoskeletal risk using visual and skeletal coordinate-based features. In addition, a custom multimodal dataset is constructed by combining visual data and skeletal coordinates for risk assessment. Each sample is labeled into eight risk categories based on the Rapid Entire Body Assessment system. ViSK-GAT combines a Residual Block with a Lightweight Transformer Block to learn spatial and temporal dependencies jointly. It incorporates two novel modules: the Fine-Grained Attention Module (FGAM), which enables precise inter-modal feature refinement through cross-attention between visual and skeletal inputs, and the Multimodal Geometric Correspondence Module (MGCM), which enhances cross-modal coherence by aligning image features with coordinate-based representations. ViSK-GAT achieved strong performance with validation and test accuracies of 93.55\% and 93.89\%, respectively; a precision of 93.86\%; an F1 score of 93.85\%; and Cohen's Kappa and Matthews Correlation Coefficient of 93\%. The regression results also indicated a low Root Mean Square Error of the predicted probability distribution of 0.1205 and a corresponding Mean Absolute Error of 0.0156. Compared to nine popular transfer learning backbones, ViSK-GAT consistently outperformed previous methods. The ViSK-GAT model advances artificial intelligence implementation and application, transforming musculoskeletal risk classification and enabling impactful early interventions in sports.
Abstract (translated)
肌肉骨骼疾病对运动员构成了重大风险，及早评估这些风险对于预防至关重要。然而，大多数现有的方法设计用于受控环境，并且由于依赖单一类型的数据而无法可靠地在复杂环境中进行风险评估。这项研究提出了一种新的多模态深度学习框架——ViSK-GAT（视觉-骨骼几何注意力变换器），该框架旨在使用视觉和基于坐标的骨骼特征来分类肌肉骨骼风险。此外，还构建了一个自定义的多模态数据集，通过结合视觉数据和骨骼坐标来进行风险评估。每个样本根据快速全身评估系统被标记为八个风险类别。 ViSK-GAT 结合了残差块（Residual Block）与轻量级变换器块（Lightweight Transformer Block），以共同学习空间和时间依赖关系。它还包含了两个新颖的模块：细粒度注意力模块（FGAM），该模块通过视觉和骨骼输入之间的交叉注意来实现精确的跨模态特征细化；以及多模态几何对应模块（MGCM），该模块通过将图像特征与基于坐标的表示对齐，以增强跨模式的一致性。 ViSK-GAT 在验证集和测试集上分别实现了 93.55% 和 93.89% 的准确性；精度为 93.86%；F1 分数为 93.85%；并且 Cohen's Kappa 值和 Matthews 相关系数分别为 93%。回归结果也表明，预测概率分布的均方根误差（Root Mean Square Error）仅为 0.1205，相应的平均绝对误差（Mean Absolute Error）为 0.0156。 与九种流行的迁移学习骨干网络相比，ViSK-GAT 在各个指标上始终优于以前的方法。这一模型的进步推动了人工智能在肌肉骨骼风险分类中的实施和应用，并且通过早期干预使运动领域的成果具有深远的影响。
URL
https://arxiv.org/abs/2509.05913
PDF
-
Brain Tumor Detection Through Diverse CNN Architectures in IoT Healthcare Industries: Fast R-CNN, U-Net, Transfer Learning-Based CNN, and Fully Connected CNN
2025-09-06 20:03:51Mohsen Asghari Ilani, Yaser M. BanadAbstract
Artificial intelligence (AI)-powered deep learning has advanced brain tumor diagnosis in Internet of Things (IoT)-healthcare systems, achieving high accuracy with large datasets. Brain health is critical to human life, and accurate diagnosis is essential for effective treatment. Magnetic Resonance Imaging (MRI) provides key data for brain tumor detection, serving as a major source of big data for AI-driven image classification. In this study, we classified glioma, meningioma, and pituitary tumors from MRI images using Region-based Convolutional Neural Network (R-CNN) and UNet architectures. We also applied Convolutional Neural Networks (CNN) and CNN-based transfer learning models such as Inception-V3, EfficientNetB4, and VGG19. Model performance was assessed using F-score, recall, precision, and accuracy. The Fast R-CNN achieved the best results with 99% accuracy, 98.5% F-score, 99.5% Area Under the Curve (AUC), 99.4% recall, and 98.5% precision. Combining R-CNN, UNet, and transfer learning enables earlier diagnosis and more effective treatment in IoT-healthcare systems, improving patient outcomes. IoT devices such as wearable monitors and smart imaging systems continuously collect real-time data, which AI algorithms analyze to provide immediate insights for timely interventions and personalized care. For external cohort cross-dataset validation, EfficientNetB2 achieved the strongest performance among fine-tuned EfficientNet models, with 92.11% precision, 92.11% recall/sensitivity, 95.96% specificity, 92.02% F1-score, and 92.23% accuracy. These findings underscore the robustness and reliability of AI models in handling diverse datasets, reinforcing their potential to enhance brain tumor classification and patient care in IoT healthcare environments.
Abstract (translated)
基于人工智能（AI）的深度学习在物联网（IoT）医疗系统中的脑肿瘤诊断方面取得了显著进展，利用大规模数据集实现了高精度。大脑健康对人类生命至关重要，准确的诊断对于有效的治疗是必不可少的。磁共振成像（MRI）提供了脑肿瘤检测的关键数据，并且作为人工智能驱动图像分类的大数据的主要来源。在这项研究中，我们使用Region-based Convolutional Neural Network (R-CNN) 和 UNet 架构从MRI影像中对胶质瘤、脑膜瘤和垂体瘤进行了分类。同时，我们也应用了基于卷积神经网络（CNN）的模型以及诸如Inception-V3, EfficientNetB4 和 VGG19 的迁移学习模型。通过F值、召回率、准确率等指标来评估模型性能。Fast R-CNN 模型取得了最佳效果，其精度达到了99%，F值为98.5%，曲线下面积（AUC）为99.5%，召回率为99.4%，准确率为98.5%。 将R-CNN、UNet和迁移学习结合起来可以实现在物联网医疗系统中更早的诊断和更有成效的治疗，从而改善患者预后。物联网设备如可穿戴监测器和智能成像系统会持续收集实时数据，AI算法分析这些数据以提供即时洞察并进行及时干预和个人化护理。 在外部队列跨数据集验证方面，经过微调后的EfficientNetB2模型表现出色，在精度、召回率/灵敏度、特异性、F1得分和准确度上分别达到了92.11%、92.11%、95.96%、92.02% 和 92.23%。这些研究结果强调了AI模型处理多样化数据集的稳健性和可靠性，进一步强化了它们在物联网医疗环境中增强脑肿瘤分类和患者护理方面的潜力。
URL
https://arxiv.org/abs/2509.05821
PDF
-
A Survey of the State-of-the-Art in Conversational Question Answering Systems
2025-09-06 13:38:03Manoj Madushanka Perera, Adnan Mahmood, Kasun Eranda Wijethilake, Fahmida Islam, Maryam Tahermazandarani, Quan Z. ShengAbstract
Conversational Question Answering (ConvQA) systems have emerged as a pivotal area within Natural Language Processing (NLP) by driving advancements that enable machines to engage in dynamic and context-aware conversations. These capabilities are increasingly being applied across various domains, i.e., customer support, education, legal, and healthcare where maintaining a coherent and relevant conversation is essential. Building on recent advancements, this survey provides a comprehensive analysis of the state-of-the-art in ConvQA. This survey begins by examining the core components of ConvQA systems, i.e., history selection, question understanding, and answer prediction, highlighting their interplay in ensuring coherence and relevance in multi-turn conversations. It further investigates the use of advanced machine learning techniques, including but not limited to, reinforcement learning, contrastive learning, and transfer learning to improve ConvQA accuracy and efficiency. The pivotal role of large language models, i.e., RoBERTa, GPT-4, Gemini 2.0 Flash, Mistral 7B, and LLaMA 3, is also explored, thereby showcasing their impact through data scalability and architectural advancements. Additionally, this survey presents a comprehensive analysis of key ConvQA datasets and concludes by outlining open research directions. Overall, this work offers a comprehensive overview of the ConvQA landscape and provides valuable insights to guide future advancements in the field.
Abstract (translated)
对话式问答（ConvQA）系统在自然语言处理（NLP）领域中已成为一个至关重要的研究方向，推动了机器进行动态和上下文感知的会话能力的发展。这种技术正在被广泛应用于各个领域，例如客户服务、教育、法律及医疗保健等，这些领域需要维持连贯且相关性的对话内容。基于近期的研究进展，本文综述提供了关于ConvQA最新技术水平的全面分析。 此综述首先审视了ConvQA系统的核心组成部分——历史信息选择、问题理解和答案预测，并强调它们在确保多轮会话中连贯性和相关性方面的相互作用。此外，还探讨了包括强化学习、对比学习和迁移学习在内的高级机器学习技术的应用，以提高ConvQA的准确性和效率。综述还探索了大型语言模型（如RoBERTa, GPT-4, Gemini 2.0 Flash, Mistral 7B 和 LLaMA 3）在其数据规模扩展及架构改进方面所发挥的关键作用，并展示了它们在ConvQA中的影响。此外，本文还对关键的ConvQA数据集进行了全面分析，并总结了当前的研究方向。 总的来说，这项工作提供了关于ConvQA领域现状的一个全面概述，并为未来该领域的进一步发展提供了宝贵的见解和指导。
URL
https://arxiv.org/abs/2509.05716
PDF
-
An Analysis of Layer-Freezing Strategies for Enhanced Transfer Learning in YOLO Architectures
2025-09-05 20:39:43Andrzej D. Dobrzycki, Ana M. Bernardos, Jos\'e R. CasarAbstract
The You Only Look Once (YOLO) architecture is crucial for real-time object detection. However, deploying it in resource-constrained environments such as unmanned aerial vehicles (UAVs) requires efficient transfer learning. Although layer freezing is a common technique, the specific impact of various freezing configurations on contemporary YOLOv8 and YOLOv10 architectures remains unexplored, particularly with regard to the interplay between freezing depth, dataset characteristics, and training dynamics. This research addresses this gap by presenting a detailed analysis of layer-freezing strategies. We systematically investigate multiple freezing configurations across YOLOv8 and YOLOv10 variants using four challenging datasets that represent critical infrastructure monitoring. Our methodology integrates a gradient behavior analysis (L2 norm) and visual explanations (Grad-CAM) to provide deeper insights into training dynamics under different freezing strategies. Our results reveal that there is no universal optimal freezing strategy but, rather, one that depends on the properties of the data. For example, freezing the backbone is effective for preserving general-purpose features, while a shallower freeze is better suited to handling extreme class imbalance. These configurations reduce graphics processing unit (GPU) memory consumption by up to 28% compared to full fine-tuning and, in some cases, achieve mean average precision (mAP@50) scores that surpass those of full fine-tuning. Gradient analysis corroborates these findings, showing distinct convergence patterns for moderately frozen models. Ultimately, this work provides empirical findings and practical guidelines for selecting freezing strategies. It offers a practical, evidence-based approach to balanced transfer learning for object detection in scenarios with limited resources.
Abstract (translated)
仅此一次看（YOLO）架构对于实时目标检测至关重要。然而，在资源受限的环境中（如无人驾驶飞行器[UAV]）部署它需要高效的迁移学习。尽管冻结层是一种常见的技术，但不同冻结配置对当代YOLOv8和YOLOv10架构的具体影响尚未得到充分研究，尤其是在冻结深度、数据集特征以及训练动态之间的相互作用方面。这项研究通过详细分析层冻结策略来填补这一空白。我们系统地调查了YOLOv8和YOLOv10变体中多种冻结配置，并使用四个具有挑战性的代表关键基础设施监控的数据集进行测试。我们的方法结合了梯度行为分析（L2范数）和视觉解释（Grad-CAM），以在不同冻结策略下提供对训练动态的更深入了解。研究结果表明，没有通用的最佳冻结策略，而是根据数据属性有所不同。例如，冻结骨干网络有助于保持一般特征的有效性，而浅层冻结则更适合处理极端类别不平衡的情况。这些配置与全面微调相比可减少多达28%的图形处理器（GPU）内存消耗，并在某些情况下实现超过全面微调的平均精度均值（mAP@50）分数。梯度分析证实了这一发现，展示了适度冻结模型的独特收敛模式。最终，这项工作提供了经验性发现和实用指南来选择冻结策略。它为资源有限场景中的目标检测平衡迁移学习提供了一种基于证据的实际方法。
URL
https://arxiv.org/abs/2509.05490
PDF
-
Leveraging Transfer Learning and Mobile-enabled Convolutional Neural Networks for Improved Arabic Handwritten Character Recognition
2025-09-05 11:28:53Mohsine El Khayati, Ayyad Maafiri, Yassine Himeur, Hamzah Ali Alkhazaleh, Shadi Atalla, Wathiq MansoorAbstract
The study explores the integration of transfer learning (TL) with mobile-enabled convolutional neural networks (MbNets) to enhance Arabic Handwritten Character Recognition (AHCR). Addressing challenges like extensive computational requirements and dataset scarcity, this research evaluates three TL strategies--full fine-tuning, partial fine-tuning, and training from scratch--using four lightweight MbNets: MobileNet, SqueezeNet, MnasNet, and ShuffleNet. Experiments were conducted on three benchmark datasets: AHCD, HIJJA, and IFHCDB. MobileNet emerged as the top-performing model, consistently achieving superior accuracy, robustness, and efficiency, with ShuffleNet excelling in generalization, particularly under full fine-tuning. The IFHCDB dataset yielded the highest results, with 99% accuracy using MnasNet under full fine-tuning, highlighting its suitability for robust character recognition. The AHCD dataset achieved competitive accuracy (97%) with ShuffleNet, while HIJJA posed significant challenges due to its variability, achieving a peak accuracy of 92% with ShuffleNet. Notably, full fine-tuning demonstrated the best overall performance, balancing accuracy and convergence speed, while partial fine-tuning underperformed across metrics. These findings underscore the potential of combining TL and MbNets for resource-efficient AHCR, paving the way for further optimizations and broader applications. Future work will explore architectural modifications, in-depth dataset feature analysis, data augmentation, and advanced sensitivity analysis to enhance model robustness and generalizability.
Abstract (translated)
这项研究探讨了将迁移学习（TL）与移动设备支持的卷积神经网络（MbNets）结合，以增强阿拉伯手写字符识别（AHCR）。为了解决诸如计算需求量大和数据集稀缺等挑战，本研究评估了三种迁移学习策略——完全微调、部分微调以及从头开始训练，使用四种轻量级的MbNet模型：MobileNet、SqueezeNet、MnasNet 和 ShuffleNet。实验在三个基准数据集上进行：AHCD、HIJJA 和 IFHCDB。 MobileNet 凸显为表现最佳的模型，在精度、鲁棒性和效率方面持续表现出色，而ShuffleNet 在完全微调的情况下，尤其在泛化性能方面表现出色。IFHCDB 数据集取得了最高结果，使用 MnasNet 进行完全微调达到了99% 的准确率，突显其对于字符识别的稳健性。AHCD 数据集利用 ShuffleNet 达到了有竞争力的 97% 准确率，而 HIJJA 数据集由于其变异性带来了重大挑战，在使用 ShuffleNet 的情况下达到最高的 92% 准确率。 值得注意的是，完全微调策略展示了最佳的整体性能，很好地平衡了精度和收敛速度，相比之下部分微调在各项指标中表现较差。这些发现强调了将迁移学习与MbNets 结合应用于资源高效的AHCR的潜力，并为进一步优化及更广泛的应用铺平道路。未来的研究工作将进一步探索架构改进、深入的数据集特征分析、数据增强以及高级敏感性分析，以提高模型的鲁棒性和泛化能力。
URL
https://arxiv.org/abs/2509.05019
PDF
-
Interpretable Deep Transfer Learning for Breast Ultrasound Cancer Detection: A Multi-Dataset Study
2025-09-05 11:03:15Mohammad Abbadi, Yassine Himeur, Shadi Atalla, Wathiq MansoorAbstract
Breast cancer remains a leading cause of cancer-related mortality among women worldwide. Ultrasound imaging, widely used due to its safety and cost-effectiveness, plays a key role in early detection, especially in patients with dense breast tissue. This paper presents a comprehensive study on the application of machine learning and deep learning techniques for breast cancer classification using ultrasound images. Using datasets such as BUSI, BUS-BRA, and BrEaST-Lesions USG, we evaluate classical machine learning models (SVM, KNN) and deep convolutional neural networks (ResNet-18, EfficientNet-B0, GoogLeNet). Experimental results show that ResNet-18 achieves the highest accuracy (99.7%) and perfect sensitivity for malignant lesions. Classical ML models, though outperformed by CNNs, achieve competitive performance when enhanced with deep feature extraction. Grad-CAM visualizations further improve model transparency by highlighting diagnostically relevant image regions. These findings support the integration of AI-based diagnostic tools into clinical workflows and demonstrate the feasibility of deploying high-performing, interpretable systems for ultrasound-based breast cancer detection.
Abstract (translated)
乳腺癌仍然是全球女性癌症死亡的主要原因之一。超声成像由于其安全性及成本效益，被广泛用于早期检测，特别是在具有致密乳腺组织的患者中发挥关键作用。本文介绍了一项关于使用机器学习和深度学习技术对乳腺癌进行分类的综合研究，通过使用包括BUSI、BUS-BRA和BrEaST-Lesions USG等数据集，评估了传统机器学习模型（如支持向量机SVM和K近邻算法KNN）以及深度卷积神经网络（ResNet-18、EfficientNet-B0、GoogLeNet）。实验结果表明，ResNet-18取得了最高的准确率（99.7%），并且对恶性肿瘤的敏感性达到了完美。尽管传统的机器学习模型在性能上被CNNs超越，但通过深度特征提取增强后，它们仍能获得有竞争力的表现。Grad-CAM可视化技术进一步提升了模型的透明度，通过突出显示具有诊断相关性的图像区域来提高解释能力。这些发现支持将基于AI的诊断工具整合到临床工作流程中，并展示了部署高性能且易于解释的系统用于超声波乳腺癌检测的可行性。
URL
https://arxiv.org/abs/2509.05004
PDF
-
SpiderNets: Estimating Fear Ratings of Spider-Related Images with Vision Models
2025-09-05 08:10:40Dominik Pegler, David Steyrl, Mengfan Zhang, Alexander Karner, Jozsef Arato, Frank Scharnowski, Filip MelinscakAbstract
Advances in computer vision have opened new avenues for clinical applications, particularly in computerized exposure therapy where visual stimuli can be dynamically adjusted based on patient responses. As a critical step toward such adaptive systems, we investigated whether pretrained computer vision models can accurately predict fear levels from spider-related images. We adapted three diverse models using transfer learning to predict human fear ratings (on a 0-100 scale) from a standardized dataset of 313 images. The models were evaluated using cross-validation, achieving an average mean absolute error (MAE) between 10.1 and 11.0. Our learning curve analysis revealed that reducing the dataset size significantly harmed performance, though further increases yielded no substantial gains. Explainability assessments showed the models' predictions were based on spider-related features. A category-wise error analysis further identified visual conditions associated with higher errors (e.g., distant views and artificial/painted spiders). These findings demonstrate the potential of explainable computer vision models in predicting fear ratings, highlighting the importance of both model explainability and a sufficient dataset size for developing effective emotion-aware therapeutic technologies.
Abstract (translated)
计算机视觉的进步为临床应用开辟了新的途径，尤其是在计算机化暴露疗法中，可以根据患者的反应动态调整视觉刺激。为了向此类自适应系统迈进的关键一步，我们研究了预训练的计算机视觉模型是否能够准确地从与蜘蛛相关的图像中预测恐惧水平。我们使用迁移学习方法改编了三种不同的模型，以根据标准化的313张图片的数据集预测人类对恐惧程度（在0到100分制上的评分）。通过交叉验证评估这些模型后发现，它们达到了平均绝对误差（MAE）为10.1至11.0的成绩。学习曲线分析表明，减少数据集大小显著影响了性能，但进一步增加数据量却没有带来实质性的改进。 在可解释性评估中，我们发现模型的预测主要基于与蜘蛛相关的特征，并且通过按类别划分错误分析后，我们还确定了一些视觉条件会导致更高的误差（例如，远距离视图和人工/画出来的蜘蛛）。这些结果证明了具有透明度的计算机视觉模型在预测恐惧评分方面的潜力，并强调了对于开发有效的感知情感的技术而言，模型可解释性和足够大的数据集的重要性。
URL
https://arxiv.org/abs/2509.04889
PDF
-
Crossing the Species Divide: Transfer Learning from Speech to Animal Sounds
2025-09-04 12:39:05Jules Cauzinille, Marius Miron, Olivier Pietquin, Masato Hagiwara, Ricard Marxer, Arnaud Rey, Benoit FavreAbstract
Self-supervised speech models have demonstrated impressive performance in speech processing, but their effectiveness on non-speech data remains underexplored. We study the transfer learning capabilities of such models on bioacoustic detection and classification tasks. We show that models such as HuBERT, WavLM, and XEUS can generate rich latent representations of animal sounds across taxa. We analyze the models properties with linear probing on time-averaged representations. We then extend the approach to account for the effect of time-wise information with other downstream architectures. Finally, we study the implication of frequency range and noise on performance. Notably, our results are competitive with fine-tuned bioacoustic pre-trained models and show the impact of noise-robust pre-training setups. These findings highlight the potential of speech-based self-supervised learning as an efficient framework for advancing bioacoustic research.
Abstract (translated)
自监督语音模型在语音处理中表现出令人印象深刻的表现，但它们在非语音数据上的有效性仍需进一步探索。我们研究了这些模型在生物声学检测和分类任务中的迁移学习能力。研究表明，如HuBERT、WavLM和XEUS等模型能够生成跨物种的动物声音的丰富潜在表示。我们通过时间平均表示进行线性探测来分析模型特性，并随后扩展方法以考虑时序信息的影响及其对下游架构性能的作用。最后，我们研究了频率范围和噪声对性能的影响。值得注意的是，我们的结果与经过生物声学预训练并微调后的模型表现相当，并显示了鲁棒去噪预训练设置的重要性。这些发现突显了基于语音的自监督学习作为推进生物声学研究的有效框架的巨大潜力。
URL
https://arxiv.org/abs/2509.04166
PDF
-
Chest X-ray Pneumothorax Segmentation Using EfficientNet-B4 Transfer Learning in a U-Net Architecture
2025-09-04 07:21:37Alvaro Aranibar Roque, Helga SebastianAbstract
Pneumothorax, the abnormal accumulation of air in the pleural space, can be life-threatening if undetected. Chest X-rays are the first-line diagnostic tool, but small cases may be subtle. We propose an automated deep-learning pipeline using a U-Net with an EfficientNet-B4 encoder to segment pneumothorax regions. Trained on the SIIM-ACR dataset with data augmentation and a combined binary cross-entropy plus Dice loss, the model achieved an IoU of 0.7008 and Dice score of 0.8241 on the independent PTX-498 dataset. These results demonstrate that the model can accurately localize pneumothoraces and support radiologists.
Abstract (translated)
气胸是指肺部周围积聚了异常的空气，这种情况如果未被发现可能会危及生命。胸部 X 光片是首选诊断工具，但较小的情况可能难以察觉。我们提出了一种使用 U-Net 结构和 EfficientNet-B4 编码器的自动化深度学习流程，用于分割气胸区域。该模型在 SIIM-ACR 数据集上进行训练，并采用了数据增强及结合二元交叉熵与 Dice 损失函数的组合方法。经过训练后，该模型在独立的 PTX-498 数据集中达到了 0.7008 的交并比（IoU）和 0.8241 的 Dice 分数。这些结果表明，该模型能够准确地定位气胸，并且可以支持放射科医生的工作。
URL
https://arxiv.org/abs/2509.03950
PDF
