Local Model Research Articles

Predicting treatment response in multicenter non-small cell lung cancer patients based on federated learning

BackgroundMulticenter non-small cell lung cancer (NSCLC) patient data is information-rich. However, its direct integration becomes exceptionally challenging due to constraints involving different healthcare organizations and regulations. Traditional centralized machine learning methods require centralizing these sensitive medical data for training, posing risks of patient privacy leakage and data security issues. In this context, federated learning (FL) has attracted much attention as a distributed machine learning framework. It effectively addresses this contradiction by preserving data locally, conducting local model training, and aggregating model parameters. This approach enables the utilization of multicenter data with maximum benefit while ensuring privacy safeguards. Based on pre-radiotherapy planning target volume images of NSCLC patients, a multicenter treatment response prediction model is designed by FL for predicting the probability of remission of NSCLC patients. This approach ensures medical data privacy, high prediction accuracy and computing efficiency, offering valuable insights for clinical decision-making.MethodsWe retrospectively collected CT images from 245 NSCLC patients undergoing chemotherapy and radiotherapy (CRT) in four Chinese hospitals. In a simulation environment, we compared the performance of the centralized deep learning (DL) model with that of the FL model using data from two sites. Additionally, due to the unavailability of data from one hospital, we established a real-world FL model using data from three sites. Assessments were conducted using measures such as accuracy, receiver operating characteristic curve, and confusion matrices.ResultsThe model’s prediction performance obtained using FL methods outperforms that of traditional centralized learning methods. In the comparative experiment, the DL model achieves an AUC of 0.718/0.695, while the FL model demonstrates an AUC of 0.725/0.689, with real-world FL model achieving an AUC of 0.698/0.672.ConclusionsWe demonstrate that the performance of a FL predictive model, developed by combining convolutional neural networks (CNNs) with data from multiple medical centers, is comparable to that of a traditional DL model obtained through centralized training. It can efficiently predict CRT treatment response in NSCLC patients while preserving privacy.

Dynamo action driven by precessional turbulence.

We reveal and analyze an efficient magnetic dynamo action due to precession-driven hydrodynamic turbulence in the local model of a precessional flow, focusing on the kinematic stage of this dynamo. The growth rate of the magnetic field monotonically increases with the Poincaré number Po, characterizing precession strength, and the magnetic Prandtl number Pm, equal to the ratio of viscosity to resistivity, for the considered ranges of these parameters. The critical Po_{c} for the dynamo onset decreases with increasing Pm. To understand the scale-by-scale evolution (growth) of the precession dynamo and its driving processes, we perform spectral analysis by calculating the spectra of magnetic energy and of different terms in the induction equationin Fourier space. To this end, we decompose the velocity field of precession-driven turbulence into two-dimensional (2D) vortical and three-dimensional (3D) inertial wave modes. It is shown that the dynamo operates across a broad range of scales and exhibits a remarkable transition from a primarily vortex-driven regime at lower Po to a more complex regime at higher Po where it is driven jointly by vortices, inertial waves, and the shear of the background precessional flow. Vortices and shear drive the dynamo mostly at large scales comparable to the flow system size, and at intermediate scales, while at smaller scales it is mainly driven by inertial waves. This study can be important not only for understanding the magnetic dynamo action in precession-driven flows, but also in a general context of flows where vortices emerge and govern the flow dynamics and evolution.

An Audio Copy-Move Forgery Localization Model by CNN-Based Spectral Analysis

In audio copy-move forgery forensics, existing traditional methods typically first segment audio into voiced and silent segments, then compute the similarity between voiced segments to detect and locate forged segments. However, audio collected in noisy environments is difficult to segment and manually set, and heuristic similarity thresholds lack robustness. Existing deep learning methods extract features from audio and then use neural networks for binary classification, lacking the ability to locate forged segments. Therefore, for locating audio copy-move forgery segments, we have improved deep learning methods and proposed a robust localization model by CNN-based spectral analysis. In the localization model, the Feature Extraction Module extracts deep features from Mel-spectrograms, while the Correlation Detection Module automatically decides on the correlation between these deep features. Finally, the Mask Decoding Module visually locates the forged segments. Experimental results show that compared to existing methods, the localization model improves the detection accuracy of audio copy-move forgery by 3.0–6.8%and improves the average detection accuracy of forged audio with post-processing attacks such as noise, filtering, resampling, and MP3 compression by over 7.0%.

Enhancing IoT Healthcare with Federated Learning and Variational Autoencoder.

The growth of IoT healthcare is aimed at providing efficient services to patients by utilizing data from local hospitals. However, privacy concerns can impede data sharing among third parties. Federated learning offers a solution by enabling the training of neural networks while maintaining the privacy of the data. To integrate federated learning into IoT healthcare, hospitals must be part of the network to jointly train a global central model on the server. Local hospitals can train the global model using their patient datasets and send the trained localized models to the server. These localized models are then aggregated to enhance the global model training process. The aggregation of local models dramatically influences the performance of global training, mainly due to the heterogeneous nature of patient data. Existing solutions to address this issue are iterative, slow, and susceptible to convergence. We propose two novel approaches that form groups efficiently and assign the aggregation weightage considering essential parameters vital for global training. Specifically, our method utilizes an autoencoder to extract features and learn the divergence between the latent representations of patient data to form groups, facilitating more efficient handling of heterogeneity. Additionally, we propose another novel aggregation process that utilizes several factors, including extracted features of patient data, to maximize performance further. Our proposed approaches for group formation and aggregation weighting outperform existing conventional methods. Notably, significant results are obtained, one of which shows that our proposed method achieves 20.8% higher accuracy and 7% lower loss reduction compared to the conventional methods.

Improved acoustic source localization method for crack identification in structures

Acoustic Emission (AE) monitoring is considered one of the popular non-destructive testing (NDT) methodologies that have been used to predict and identify the location of damage in critical civil infrastructure. In this paper, an improved AE crack localization method is proposed by integrating the empirical mode decomposition (EMD)-based signal decomposition method with a source localization model. Unlike the conventional AE method, the proposed method is free of the use of bulk AE parameters such as counts, rise time, signal strength, and energy. First, EMD is used to minimize the presence of noise in the recorded AE waveforms and extract the key AE components. Then, key AE events are located using the source localization model to localize the crack in concrete structures. The performance of the proposed method is validated experimentally on small and large-scale concrete beams, where the damage is induced using progressive static load testing. In particular, the large-scale beams are designed for flexural and shear mode failure to evaluate the performance of the proposed method under various types of damage. Finally, the results of the proposed method are compared with the traditional method that uses raw AE waveforms and the method that uses bandpass-filtered AE waveforms. The results show higher crack location accuracy of the proposed method than the other methods, which makes it a suitable approach as a crack localization technology.

MCNet: Multivariate long-term time series forecasting with local and global context modeling

Time series data typically exhibit various intra-sequence and inter-sequence correlations, resulting in intricate, intertwined dependencies, which pose challenges for accurately predicting future long-term trends. Previous studies have not fully considered the two correlations, and they also still face challenges of excessive time and memory complexity when dealing with long-term predictions. To address these challenges and establish high-precision prediction models, we propose MCNet that consists of a local branch and a global branch. The local branch aims at capturing short-term variations of intra-sequences, as well as capturing inter-sequence correlations. The global branch models long-term dependencies within sequences. Specifically, the local branch consists only of MLP module, which effectively captures short-term variations of intra-sequences by independently modeling the temporal information within and between patches of the most recent time series. Subsequently, inter-sequence dependencies are captured through the channel interaction module, which further explores more key information to improve the performance of MCNet. Meanwhile, global branch models long-term dependencies within the time series through structured global convolution. Experimental results on multiple popular long-term time series forecasting benchmarks demonstrate that MCNet outperforms state-of-the-art methods, yielding a relative improvement of 12% for multivariate time series while also being more efficient.

Fabric4show: real-time vision system for fabric defect detection and post-processing

The exploration of computer vision applications for fabric defect detection has immense potential value. However, current relevant research in this area has primarily focused on detection models that aim for high detection accuracy and algorithmic efficiency, while neglecting the practical industrial production requirements. Therefore, we propose a fabric defect detection and post-processing system that integrates an optimized region with convolutional neural network (CNN) features (i.e., Faster R-CNN) for defect detection, defect localization and detection model evaluation. In addition, the proposed intelligent system incorporates novel approaches, such as a rearranged fabric dataset, anomaly detection, recommended clipping region division, and a replenishment device. This study illustrates an example of artificial intelligence (AI)-driven automated technology in fabric manufacturing. The accuracy and detection speed of different detection models under identical hardware conditions are evaluated and compared with related work. Experimental results demonstrate that the proposed approach achieves comparable performance to other models, while significantly reducing computational resource requirements. The potential efficiency of using two-stage networks on hardware systems for fabric defect detection tasks is highlighted, which is likely to have relevant implications for the textile industry.

Optical feedback-induced spatiotemporal patterns with power law spectra in a liquid crystal light valve.

Pattern formation can be induced by coupling electromagnetic fields to a polarizable and lossy medium. Increasing energy injection patterns can exhibit aperiodic behaviors. We investigate the self-organization of unidimensional aperiodic patterns. Based on a liquid crystal light valve (LCLV) with optical feedback, we observed aperiodic one-dimensional patterns with power laws in the temporal and spatial spectrum density of the light intensity, and their pseudo envelope and phase characteristic of spatiotemporal complexity. Theoretically, a local model describes the system close to nascent bistability and spatial instability. Numerical simulations of this model show chaotic spatiotemporal patterns whose temporal and spatial spectra have exponents similar to those observed experimentally.

Removing ring artifacts in CBCT images via Transformer with unidirectional vertical gradient loss.

Cone-beam computed tomography (CBCT), an important medical modality for disease detection and diagnosis, is currently widely used in clinical practice. However, due to the inconsistent response of CBCT detectors, the lack of proper calibration often leads to the occurrence of ring artifacts in CBCT-reconstructed images. These artifacts may affect physicians' assessment and diagnosis. Therefore, effective elimination of ring artifacts in CBCT images without degrading image quality isimportant. Given the pros and cons of existing methods for removing ring artifacts in CBCT images, this paper is devoted to designing a specific Transformer for this task, leveraging the global and local modeling ability ofTransformer. We design a loss function with dual-domain information fusion for the vanilla Transformer to correct ring artifacts in CBCT images. The method operates in image domain to predict artifact-free outputs and preserve more image details. Meanwhile, we design a tailored loss function incorporating polar domain optimization to remove ring artifacts more effectively. Specifically, an unidirectional gradient loss that constrains vertical gradients in polar domain is imposed, based on the geometric prior that in polar coordinates, ring artifacts predominately affect horizontal gradients while minimally influencing verticalgradients. We conduct extensive ablative and comparative experiments on CBCT/CT image set to validate the performance of the proposed method. First, four ablation experiments demonstrate the feasibility of our approach. Then, we compare our method with several classical methods and the latest state-of-the-arts, and our method achieves the highest quality of corrected images as well as the best evaluation metrics. In these experiments, 5332 CT images were used for training, and 550 CT images, and 500 real CBCT images were used for testing. The source code is available at https://github.com/shasha521/CBCT. Experimental results demonstrate that our method can significantly improve the effectiveness of ring artifact correction. By capitalizing on dual-domain information fusion and a customized loss function, the improved Transformer can not only effectively remove ring artifacts in CBCT images, but also preserve the details of original images quitewell.

Efficiency of Federated Learning and Blockchain in Preserving Privacy and Enhancing the Performance of Credit Card Fraud Detection (CCFD) Systems

Increasing global credit card usage has elevated it to a preferred payment method for daily transactions, underscoring its significance in global financial cybersecurity. This paper introduces a credit card fraud detection (CCFD) system that integrates federated learning (FL) with blockchain technology. The experiment employs FL to establish a global learning model on the cloud server, which transmits initial parameters to individual local learning models on fog nodes. With three banks (fog nodes) involved, each bank trains its learning model locally, ensuring data privacy, and subsequently sends back updated parameters to the global learning model. Through the integration of FL and blockchain, our system ensures privacy preservation and data protection. We utilize three machine learning and deep neural network learning algorithms, RF, CNN, and LSTM, alongside deep optimization techniques such as ADAM, SGD, and MSGD. The SMOTE oversampling technique is also employed to balance the dataset before model training. Our proposed framework has demonstrated efficiency and effectiveness in enhancing classification performance and prediction accuracy.

Deep learning insights into suspended sediment concentrations across the conterminous United States: Strengths and limitations

Suspended sediment concentration (SSC) is a crucial indicator for aquatic ecosystems and reservoir management but is challenging to predict at large scales. It is unclear whether SSC is predictable using macroscopic environmental attributes and forcings. This study tested the feasibility of deep-network-based models to predict daily SSC at basin outlets given only basin-averaged forcings, readily-available physiographic attributes, and streamflow (from observation or model). We trained long short-term memory (LSTM) deep networks both separately for each of the 377 sites across the conterminous United States (CONUS) (termed “local models”), and on all the sites collectively (Whole-CONUS). The Whole-CONUS and local models presented median coefficient of determination (R2) values of 0.63 and 0.52, respectively. This comparison agrees with previously acknowledged “data synergy” effects for LSTM models, where more data from more sites can help improve predictions overall. Furthermore, the continental-scale analysis provided a wealth of insights about SSC patterns. Both local and Whole-CONUS models tended to be more successful where SSC-streamflow correlations (Rs-q) were high − typically in the humid Eastern US − and with lower SSC. Low Rs-q basins were often found in the arid Southwest with higher SSC. The highly-nonlinear SSC-streamflow relationships seem related to seasonality, basin size, and heterogeneity of land cover and rainfall within the basins, suggesting these basins need to be simulated at higher spatial resolution and may require additional inputs related to SSC induced by seasonality. The Whole-CONUS model also performed well for spatial extrapolation (basins not included in the training dataset, median R2 = 0.55), supporting large-scale modeling efforts. These state-of-the-art results using only minimal inputs suggest data-driven approaches can exploit the natural coevolution of sediment processes and the environment to support sediment modeling.

Quantum steering in a star network

In this work, we will consider the star network scenario where the central party is trusted while all the edge parties (with a number of n) are untrusted. Network steering is defined with an n local hidden state model which can be viewed as a special kind of n local hidden variable model. Two different types of sufficient criteria, nonlinear steering inequality and linear steering inequality will be constructed to verify the quantum steering in a star network. Based on the linear steering inequality, how to detect the network steering with a fixed measurement will be discussed.

Numerical assessment of thermomagnetic convection of ferrofluid in a porous cavity employing a local thermal nonequilibrium model

Utilizing a two-temperature model, this study investigates the thermomagnetic convection of a ferrofluid within a porous cavity. The cavity is heated from the bottom and cooled from the sides to establish thermo-gravitational convection, whereas the thermomagnetic convection is induced by permanent magnets. The equations that govern flow and heat transfer are discretized using the finite volume method and the SIMPLE algorithm. The first part of the study evaluates the impact of the permanent magnet’s position on the flow pattern and ferrofluid and solid temperature fields. The results showed that positioning the magnet at the bottom of the cavity enhances heat transfer rate. The second part investigates the impact of the distance between two magnets at different magnetic, Rayleigh, and Darcy numbers, as well as the conduction ratio. The numerical findings exhibit that the heat transfer rate attains a peak at a certain distance between the magnets, depending on the Rayleigh number. With increasing Rayleigh number and conduction ratio, or decreasing Darcy number, the role of the magnetic field in improving heat transfer becomes less important. It is observed that thermomagnetic convection is replaced by thermo-gravitational one at high conduction ratios. The study emphasizes the need to use a local thermal non-equilibrium model in porous media with high magnetic and Darcy numbers, as well as large conduction ratios, to accurately capture the heat transfer phenomenon. It is observed that by placing two magnets at an appropriate distance, the heat transfer rate could be augmented by more than 136%.

Complex quasi-brittle fracture modelling using lattice particle method coupled with a local isotropic damage model

In this paper, a local isotropic damage model is incorporated into the lattice particle method (LPM) to model complex quasi-brittle fracture problems. The issue of particle-size dependency is addressed using fracture energy in the damage model regularized with characteristic length that is particle-size dependent. The characteristic lengths are first estimated from mode-I fracture test by ensuring the prescribed and calculated fracture energies match well for both 2D and 3D cases. Extensive validation against experimental data in terms of crack path and structural response is then carried out to assess the accuracy of the proposed model in 2D and 3D settings. Positive outcomes, in line with the experimental findings, have been obtained for both 2D and 3D problems covering mode I and mixed-mode fractures. Furthermore, by employing a regular lattice network, the present model, when integrated with more sophisticated yield surfaces, can address intricate 3D fracture problems that entail the twisting of crack surfaces. Convergence of structural response and crack path is obtained when the particle spacing is refined.

Gradient Matching Federated Domain Adaptation for Brain Image Classification.

Federated learning has shown its unique advantages in many different tasks, including brain image analysis. It provides a new way to train deep learning models while protecting the privacy of medical image data from multiple sites. However, previous studies suggest that domain shift across different sites may influence the performance of federated models. As a solution, we propose a gradient matching federated domain adaptation (GM-FedDA) method for brain image classification, aiming to reduce domain discrepancy with the assistance of a public image dataset and train robust local federated models for target sites. It mainly includes two stages: 1) pretraining stage; we propose a one-common-source adversarial domain adaptation (OCS-ADA) strategy, i.e., adopting ADA with gradient matching loss to pretrain encoders for reducing domain shift at each target site (private data) with the assistance of a common source domain (public data) and 2) fine-tuning stage; we develop a gradient matching federated (GM-Fed) fine-tuning method for updating local federated models pretrained with the OCS-ADA strategy, i.e., pushing the optimization direction of a local federated model toward its specific local minimum by minimizing gradient matching loss between sites. Using fully connected networks as local models, we validate our method with the diagnostic classification tasks of schizophrenia and major depressive disorder based on multisite resting-state functional MRI (fMRI), respectively. Results show that the proposed GM-FedDA method outperforms other commonly used methods, suggesting the potential of our method in brain imaging analysis and other fields, which need to utilize multisite data while preserving data privacy.

Deep learning-based hydrogen leakage localization prediction considering sensor layout optimization in hydrogen refueling stations

To address the challenge and reliance on subjective experience in monitoring hydrogen leaks at hydrogen refueling stations (HRS), a regression prediction model for leakage localization based on Temporal convolutional network (TCN) and multimodal fusion technology is proposed for the first time. This model can accurately predict the spatial coordinates where the leakage source occurs in HRS. This study constructed a baseline dataset of hydrogen leakage, assisted by CFD simulation. Genetic Algorithm (GA) optimized the sensor layout to lower application costs while increasing monitoring effectiveness. The wind speed, wind direction, and hydrogen concentration data collected are multimodally fused to help the model mine more potential features. We compared the proposed model with current classical algorithms such as LSTM, GRU. The results demonstrate that our model achieves higher accuracy, delivering a localization model with an average error of just 0.54 meters. The proposed method can provide guidance for the layout of monitoring sensors in the large-scale HRS and provide more accurate diagnostic results for hydrogen leakage behaviors to ensure the safe operation of HRS compared to other methods.

Computational Modelling of Heat Transfer through Aluminium Metal Foams for LiFePO4 Battery Cooling

Abstract: Temperature is crucial for battery pack durability and power. Folded fin and serpentine channel cooling methods are mostly used to cool the pack. However, fluid absorption during cooling can reduce capacity and cause downstream temperatures to be higher than upstream. Consistent cooling is vital to prevent temperature variation and increase battery pack lifespan. This work is concerned with the computational study of heat dissipation from open-cell aluminium metal foam for cooling LiFePO4 battery packs. The battery module consists of six pieces of pouch cell and three pieces of the aluminium foam heat sink. In the present study, aluminium foams are positioned between the LiFePO4 battery modules that are arranged in a vertical manner. Thermal interaction between the battery module and aluminum foam was studied. The effect of pore density on heat dissipation performance at different mass flow rates was explored. It has been discovered that aluminium foam with suitable porosity and pore density can efficiently cool the LiFePO4 battery pack. This paper provides a theoretical framework for designing a thermal management system for lithium- ion batteries using aluminium foam. Background: Metal foam cooling is an established technique for thermal management of Lithiumion batteries in electric vehicles. Objective:: The present study aims to analyze heat transfer through aluminium metal foams for vertically aligned LiFePO4 battery pack cooling. Methods: The Darcy extended Forchheimer (DEF) model examines fluid flow through metallic foams, using the local thermal non-equilibrium model to determine heat transfer. Results: The impact of the density of pores in the aluminium foam on the average wall temperature and temperature difference along the battery surface is determined. The variation of heat transfer of lithium-ion battery modules for different mass flow rates is also studied. Conclusion: The results indicate that utilizing aluminium foam as a heat transfer medium for battery modules significantly enhances their thermal management performance.

Toward efficient resource utilization at edge nodes in federated learning

Federated learning (FL) enables edge nodes to collaboratively contribute to constructing a global model without sharing their data. This is accomplished by devices computing local, private model updates that are then aggregated by a server. However, computational resource constraints and network communication can become a severe bottleneck for larger model sizes typical for deep learning (DL) applications. Edge nodes tend to have limited hardware resources (RAM, CPU), and the network bandwidth and reliability at the edge is a concern for scaling federated fleet applications. In this paper, we propose and evaluate a FL strategy inspired by transfer learning in order to reduce resource utilization on devices, as well as the load on the server and network in each global training round. For each local model update, we randomly select layers to train, freezing the remaining part of the model. In doing so, we can reduce both server load and communication costs per round by excluding all untrained layer weights from being transferred to the server. The goal of this study is to empirically explore the potential trade-off between resource utilization on devices and global model convergence under the proposed strategy. We implement the approach using the FL framework FEDn. A number of experiments were carried out over different datasets (CIFAR-10, CASA, and IMDB), performing different tasks using different DL model architectures. Our results show that training the model partially can accelerate the training process, efficiently utilizes resources on-device, and reduce the data transmission by around 75% and 53% when we train 25%, and 50% of the model layers, respectively, without harming the resulting global model accuracy. Furthermore, our results demonstrate a negative correlation between the number of participating clients in the training process and the number of layers that need to be trained on each client’s side. As the number of clients increases, there is a decrease in the required number of layers. This observation highlights the potential of the approach, particularly in cross-device use cases.

The effects of non-local observations on the adjoint estimation of local model parameters: An example of Manning’s n coefficient in a tidal model over the Bohai, Yellow, and East China Seas

Accurate parameter estimation is a key issue for enhancing the performance of hydrological models. With the rapid development of satellite technology, integrating satellite data and numerical modelling for parameter estimation has gained more prominence. The limited spatial coverage of altimeters necessitates areas without direct local observations to rely only on information from nearby non-local observations. However, the relation is far from clear between non-local observations and the estimation of local model parameters. In this study, a two-dimensional adjoint tidal model over the Bohai, Yellow, and East China Seas (BYECS) is developed, assimilating observations from TOPEX/Poseidon, Jason-1, and Jason-2 altimeters in conjunction with tide-gauge stations. In twin experiments, the feasibility and reliability of utilizing non-local observations are emphasized by varying synthetic observations. The experiments also reveal how the number and distance of non-local observations affect the ill-posedness of local parameter estimation. The estimation errors can be reduced by increasing observations within a ‘critical distance’, beyond which the accuracy does not improve further. Subsequently, the robustness of ‘critical distance’ is discussed under different prescribed parameter distributions, sizes of predefined optimized area, and random observational errors, indicating that the ‘critical distance’ is approximately 1.0°-2.0°. These results provide a positive reference for practical experiment. In BYECS, the order of Manning’s n coefficient ranges from 10-4 to 10-1, with an average value of 0.0246 in shallow waters. The spatial distribution reveals 9 high-value areas, primarily modulated by a concurrence of sandy sediment and weak tidal currents. This study offers valuable insights into refining the adjoint model for physical parameter estimation, particularly in anticipation of the expanded spatial coverage of SWOT. Compared to only assimilating T/P series altimeters, incorporating SWOT reduces discrepancies by over 38%. Additionally, the spatial Manning’s n coefficient provides guidance for enhancing tidal simulation in other models.

Federated Active Learning Framework for Efficient Annotation Strategy in Skin-Lesion Classification

Federated Learning (FL) enables multiple institutes to train models collaboratively without sharing private data. Current FL research focuses on communication efficiency, privacy protection, and personalization and assumes that the data of FL have already been ideally collected. In medical scenarios, however, data annotation demands both expertise and intensive labor, which is a critical problem in FL. Active learning (AL), has shown promising performance in reducing the number of data annotations in medical image analysis. We propose a federated AL (FedAL) framework in which AL is executed periodically and interactively under FL. We exploit a local model in each hospital and a global model acquired from FL to construct an ensemble. We use ensemble-entropy-based AL as an efficient data-annotation strategy in FL. Therefore, our FedAL framework can decrease the amount of annotated data and preserve patient privacy while maintaining the performance of FL. To our knowledge, this FedAL framework applied to medical images has not been previously reported. We validated our framework on real-world dermoscopic datasets. Using only 50% of samples, our framework was able to achieve state-of-the-art performance on a skin-lesion classification task. Our framework performed better than several state-of-the-art AL methods under FL and achieved comparable performance to full-data FL.