Article Contents

2024, Volume 7, Issue 4: 459-484. Doi: 10.3934/mfc.2023023

This issue Previous Article Numerical method for the solution of nonlinear boundary value Problems (ODE) Next Article Adversarial subdomain adaptation method based on multi-scale features for bearing fault diagnosis

Role of federated learning in healthcare systems: A survey

Neeta Rana^1, , and
Hitesh Marwaha^2,

1.
FEDA, GNA University, Phagwara, 144401 (Punjab), India
2.
FCS, GNA University, Phagwara, 144401 (Punjab), India

^*Corresponding author: Neeta Rana
^*Corresponding author: Neeta Rana

Received: February 2023

Revised: May 2023

Early access: May 2023

Published: November 2024

Abstract / Introduction Full Text(HTML) Figure(2) / Table(5) Related Papers Cited by

Abstract

Nowadays, machine learning affects practically every industry, but the effectiveness of these systems depends on the accessibility of training data sets. Every device now produces data, and that data can serve as the foundation for upcoming technologies. Traditional machine learning systems need centralised data for their training, but the availability of valid and good amounts of data is not always possible due to various privacy risks. But federated learning can solve this issue [78]. In a federated learning (FL) environment, a model can be trained on decentralised datasets by involving a large number of participants, such as mobile devices or entire enterprises. Researchers are using this technique in various fields and getting great responses. The importance of using federated learning in the healthcare industry is highlighted in this paper since there is a wealth of data available in hospitals or electronic health records that may be used to train medical systems but cannot be shared due to privacy issues. The main contribution of this paper is to highlight the role of federated learning in the medical field. It also presents a list of frameworks available to implement federated learning models. The paper also listed the evaluation metrics used to check the efficiency of a federated learning model. Broadly used evaluation metrics are accuracy, precision, recall, and F1-score. Open issues for research in this area are also discussed at the end of this paper.

Keywords:

Mathematics Subject Classification: Primary: 58F15, 58F17; Secondary: 53C35.

Citation:

Full Text(HTML)

Figure 1. Overview of Federated Learning

Download: Full-size image PowerPoint slide

Figure 2. : Steps of the Federated Learning process

Download: Full-size image PowerPoint slide

Table 1. ML and DL techniques with FL to implement healthcare systems

Author/Year	Backbone Method	Description	Results	Limitations
P. Baheti, M. Sikka, K. V. Arya, and R. Rajesh, 2020 [7]	Feed-forward network	Lung node detection from CT scans has been proposed.	Predicts the accuracy of 97.65 % after being tested on "LIDC datasets"	'Not mentioned'
T. Borger et al., Aug-2022 [10]	Feed-forward Neural Network	In order to address the "impatient violence" psychiatric issue, the researcher presented an effective paradigm. In this study, the model uses clinical Natural language processing to identify the likelihood of violence. Electronic health records (EHRs) are used in system training, and four clients provided input for this study.	Results indicate that FL is the best option in training healthcare systems using decentralized data.	In this study, the Doc2Vec method was used for vectorial representations of the texts. However, Pysyft is incompatible with this approach.
J. Lo et al., Dec-2021[47]	Deep neural network	Here, an FL-based approach for "retinal microvasculature segmentation and diabetic retinopathy (RDR) classiﬁcation" using Optical Coherence Tomography (OCT) images has been proposed.	The model produced results that were comparable to those of conventional deep learning (with centralized data) in segmentation as well as in classification, while protecting data privacy.	Possibility of malicious collaborators which can provide corrupt data.
X. Xu, H. Peng, L. Sun, M. Z. A. Bhuiyan, L. Liu, and L. He, Feb-2021[79]	Neural Network with RMS Prop optimizer	A FL model called "FedMood" is proposed to identify depression using mobile health data. This model makes use of a unique kind of virtual keyboard that tracks the duration and time of the button, or the amount of time it takes the user to hit the subsequent key. Accelerometer value is also taken into account in this model to monitor typing speed, which differs from person to person. When a person is depressed, their typing speed changes, this model observes it for predicting the diseases. This model is capable of predicting Bipolar-I and Bipolar-II as well as normal mental health.	In the initial testing of this model, 8 bipolar I patients, 8 bipolar II patients, and 8 normal patients were detected. According to the study, this approach is 10-15 % more accurate than conventional techniques for the identical problem	Poor contribution of few participants in training process
Feng Qian and Andrew Zhang[60]	Use of AI with Federated Learning	Highlight how COVID-19-related issues including diagnosis utilizing a chest CT scan, mortality prediction, adverse drug responses prediction, and hospitalization prediction have been addressed with AI technology. [Paper 1 COVIDFL)	FL models are more efficient	FL has struggles with issues like uneven data distribution, data leakage, optimal algorithm training, privacy and performance Trade-off.
Mustafa Abdul Salam, Sanaa Taha, Mohamed Ramadan[1]	Deep Learning	A FL model for detection of COVID-19 from chest X-ray has been proposed	Accuracy 94.82 %	Have longer training time
Weishan Zhang, Tao Zhou, Qinghua Lu, Xiao Wang, Chunsheng Zhu, Haoyun Sun, Zhipeng Wang, Sin Kit Lo, Fei-Yue Wang[87]	Dynamic Fusion-based	The model is the best way to deal with two serious problems that are present in the default FL settings: underperformance caused by data heterogeneity and massive communication costs. The suggested approach combines two key decision-making points, such as client selection and client participation. The model's dataset consists of 2960 X-ray pictures and 746 CT images.	18 experiments are used to evaluate the system of which 14 showed greater accuracy than default FL and 4 showed lesser accuracy.	"Not Mentioned"
Ines Feki, Sourour Ammar, Yousri Kessentini, Khan Muhammad [18]	Deep CNN (VGG16 and ResNet50)	The suggested model at client sides accepts images of size 224*224 for training. Four clients in total participated in the training procedure. In this study, 76 COVID-19 patients' X-rays and 108 normal X-rays were taken into account.	Results are, federated learning can attain the same performance as centralized model, but without being required to share or centralize sensitive and private information	Datasets size should be more
Qi Dou, Tiffany Y. So, Meirui Jiang, Quande Liu, Varut Vardhanabhuti, Georgios Kaissis, Zeju Li, Weixin Si, Heather H. C. Lee, Kevin Yu, Zuxin Feng, Li Dong, Egon Burian, Friederike Jungmann, Rickmer Braren, Marcus Makowski, Bernhard Kainz, Daniel Rueckert, Ben Glocker, Simon C. H. Yu, and Pheng Ann Heng [16]	CNN-based deep learning model	The proposed model is CNN deep learning model for COVID-19 Detection from CT scans. The training of this model is in FL settings and 75 COVID-19 patients contributed in this training process. These patients belong to three Hong Kong based hospitals. For the training purposed the dataset is divided into internal and external sets. Researchers of this model are ensuring that this AI model is efficient in detection of COVID-19 from CT scans	AUC of approx. 88 % with External set1 and AUC of approx. 91 % with External set 3. These are the highest results attained.	"Not Mentioned"
Micah J. Sheller, Brandon edwards, G. Anthony Reina, Jason Martin, Sarthak pati, Aikaterini Kotrotsou, Mikhail Milchenko, Weilin Xu, Daniel Marcus, Rivka R. colen and Spyridon Bakas [67]	Deep Learning	In this study, four environments—collaborative learning through centralized data sharing, data-private collaborative learning using federated learning, data-private collaborative learning using institutional incremental learning, and data-private collaborative learning using cyclic institutional incremental learning—have been analyzed using a deep learning-based model.	According to analyses, federated learning and institutional incremental learning produce comparable results.	"Not Mentioned"
K. S. Arikumar, Sahaya Beni Prathiba, Mamoun Alazab, Thippa Reddy Gadekallu, Sharnil Pandya, Javed Masood Khan and Rajalakshmi Shenbaga Moorthy[5]	BiLSTM and deep reinforcement learning (DRL) framework	In this study, a Person Movement Identification (PMI) is proposed. Deep reinforcement learning (DRL) and BiLSTM (bidirectional long short-term memory) models are used in this model's implementation in the Federated Learning environment. Unlabeled data from wearable sensors is utilized to extract features using BiLSTM algorithm, and the data is then classified using deep reinforcement model.	Showed efficiency upto 99.67%	The model has few privacy concerns and security risks. These can be handles using blockchain technology.
Yuxia Chang, Chen Fang, Wenzhuo Sun [12]	Blockchain	The FL models now in use are attackable. This study suggests a blockchain based federated learning model that ensures immutability and transparency while sharing knowledge. Here, a verification-based consensus approach is used to govern the FL process in order to guard against single points of failure and poisoning attempts	Analysis on a real-world diabetes dataset revealed that this technique is capable of achieving an equivalent level of accuracy as the original FL in a reasonable amount of time	Improvement can be done in the system by making public partners
Dong Yang, Ziyue Xu, Wenqi Li, Andriy Myronenko, Holger R. Roth, Stephanie Harmon, Sheng Xu, Baris Turkbey, Evrim Turkbey, Xiaosong Wang, Wentao Zhu, Gianpaolo Carraﬁello, Francesca Patella, Maurizio Cariati, Hirofumi Obinata, Hitoshi Mori, Kaku Tamura, Peng An, Bradford. Wood, Daguang Xu [80]	Semi-supervised learning	Semi-supervised learning technique is used in FL settings to detect COVID-19. The dataset of this model consists of CT 1706 CT scans of COVID-19 patients. The dataset is collected from three countries such as China, Italy, Japan. Here, the patch-based training strategy is used for 3D images. The pro- posed framework is capable to grasp valuable information from the clients which only have unlabeled data.	Showing promising results as compared to fully supervised scenarios	Complexity of the model is higher than regular FL
Collaborative Federated Learning For Healthcare:Multi-Modal COVID-19 Diagnosis at the Edge [59]	Clustered-Federated Learning	In order to identify COVID-19 in both X-ray and ultrasound imaging, this study presented a collaborative learning framework employing a clustered federated learning (CFL) process	The evaluation of this approach using two benchmark datasets revealed improvements in the total F1-Score of 11% to 16%.	Large datasets are still not readily available.
A Federated Learning Framework for Healthcare IoT devices [84]	Deep Neural Networks	In federated learning environments, single lead electrocardiograms are monitored to predict arrhythmia. The training set for this study has 74275 ECG signal segments, and the test dataset contains 13107 segments.	The suggested algorithm reduces network traffic by 99.8% and 90%, respectively, when compared to FedAvg and SplitNN	Need to create a system that is more comprehensive and capable of handling numerous learning tasks on IoT healthcare devices with multiple sensors.
FedSGDCOVID: Federated SGD COVID-19 Detection under Local Differential Privacy Using Chest X-ray Images and Symptom Information [24]	2D convolutional neural networks (CNNs) with "spatial pyramid pooling (SPP) layer"	This article suggests a COVID-19 detection system called FedSGDCOVID. This model is trained using data from two datasets. 15,153 X-ray images and 5434 symptoms from 21 columns, including fever and dry cough, are included in the collection.	The suggested federated model achieves the best accuracy of 95.32% on chest X-ray pictures and 96.65% on symptom data after thorough comparison with seven models.	Dataset can be increased
S. H. Khan and M. G. R. Alam [35]	Deep Learning	The research is related to detect Pneumonia from chest X-ray images. The detection is based on the impression of foggy region in the chest X-ray.	Accuracy approx 90%	It's supposed that the in normal Chest X-ray the ribs are highlighted when a canny filter is applied, but this is not the case in Pneumonia lungs affected. But in certain cases, the normal chest X-rays become foggier with this filter as compared to Pneumonia affected lungs. If this issue can be resolved then the accuracy of this model can be increased.
S. M. Jalal [29]	Random Forest with Hyper parameter Tuning	S. M. Jalal [29] Random Forest with Hyper parameter Tuning This model is using Random Forest based statergy to detect heart disease. Here, in this model multiple decision trees are created which are becoming the base of final decision that is named as "Final Class". The optimization of the final result is done through hyperparameter tuning. The approx accuracy with this system is 87	The approx accuracy with this system is 87%.	"Not mentioned"
X. Yuan, J. Zhang, J. Luo, J. Chen, Z. Shi and M. Qin [85]	Federated Learning based on Client-selection	This study consists of two sub-algorithms; the first deals with client selection, and the second deals with customer participation in system training.	X. Yuan, J. Zhang, J. Luo, J. Chen, Z. Shi, and M. Qin [85] Federated Learning based on Client-selection This study consists of two sub-algorithms; the first deals with client selection, and the second deals with customer participation in system training. Five Deep Learning models, including LR, DT, RF, and SVM, are compared to the results. However, the suggested model FLFS excels with 86	The model needs to be optimized to increase accuracy.
A. Bhattacharya, R. Rana, V. Udutalapally and D. Das [9]	CNN Model	A. Bhattacharya, R. Rana, V. Udutalapally and D. Das [9] CNN Model A lightweight "CoviFL CNN model" model is proposed in this study which was utilised to train AIoMT edge devices utilising local datasets. Additionally, these AIoMT devices are capable of detecting COVID-19 through coughing audio. 93.01	93.01% accuracy	Can be deployed across a wide area to improve performance.
T. Ngo et al. [53]	MobileNetV2, ResNet101V2, DenseNet, VGG16	To identify Cerebellar Ataxia (CA) disease, a FL model based on deep learning algorithm has been presented. The dataset in this study was gathered from four clinics in Australia. The "Romber's balancing Test, " "IMU Kinematic Sensor, " and "Biokin Mobile App" are the foundations for the data collecting and feature extraction processes. MobileNetV2 excels when the performance of four deep learning models is compared in FL environment.	After comparison, the highest accuracy is recorded with MobileNetV2 which is 86.69 %.	Performance can be improved by creating a hybrid model that combines features from two or more distinct models.
M. N. Hossen, V. Panneerselvam, D. Koundal, K. Ahmed, F. M. Bui and S. M. Ibrahim [25]	CNN algorithms	This model is implemented to detect skin diseases like rosacea, eczema, acne, and psoriasis. The collected dataset consists of images of these four diseases that were taken from the "DermNet Image Library". A novel CNN model is proposed to detect these diseases. The effectiveness of the suggested model is compared with the popular CNN models like AlexNet and VGG16.	The accuracy was approximately 81% when the dataset was spread among 1000 clients, but it increased to almost 94.15% when it was dispersed among 25,000 contributors.	"Not mentioned"
M. Nasajpour, M. Karakaya, S. Pouriyeh and R. M. Parizi [51]	Transfer Learning (Alex Net)	In this study, transfer learning-based model in Federated Learning environment is proposed to detect Diabetic Retinopathy. Here the FL aggregation algorithms such as FedAvg and FedProx performance is compared with Transfer Learning based model.	The model outperforms with accuracy 92%(approx). Here FedProx outperform with heterogeneous data.	To ensure aggregated parameter privacy, differential privacy settings can be done.
L. Peng, N. Wang, N. Dvornek, X. Zhu, S. Member and X. L. Member [56]	Graph Generative Adversarial Network	Here, a new FL framework is proposed with a name FedNI. This model used GCN model to predict the population-based diseases.	The evaluation is done on "ABIDE and ADNI datasets" and gave 66.7% and 75.8% accuracy.	Data heterogeneity issue
S. Baghersalimi, T. Teijeiro, D. Atienza and A. Aminifar [6]	1-Dimensional Convolutional Neural Network	A 1D-CNN based FL model is proposed to detect Epileptic Seizure Detection from ECG signals.	Sensitivity approx 81% and approx speciﬁcity of 82%.	The performance of the proposed model can be increased by using ECG signal with EEG.
H. Shamseddine, S. Otoum and A. Mourad [66]	Logistic Regression, Neural Networks, Decision Trees, K-Nearest Neighbors	Here, the model is for "autism spectrum disorder (ASD)" detection. FL scheme is used to screen the patients at local screening center and the collected data is used to locally train the model and updated models are aggregated at a center point. The dataset consists of behavioral and facial images. This model uses four ML models such as Logistic Regression for feature extraction, Neural Networks for classification, Decision Trees, K-Nearest Neighbors for classification of unlabelled data.	Accuracy 63%	Not mentioned
A. Jiménez-Sánchez, M. Tardy, M. A. G. Ballester, D. Mateus and G. Piella [32]	Curriculum Learning	A memory- aware CL model is used in FL settings for Breast-cancer classification. 3 datasets of Full Field Digital Mammography (FFDM), coming from three different vendors: Hologic, GE and Siemens	AUC of 0.95 is achieved	"Not Mentioned"
C. Rønn Hansen et al.[22]	Cox regression Model	The goal of the study was to create a survival model for larynx cancer patients who had received radiotherapy. An open-source federated learning platform with Cox regression algorithm was used to train the model on data from three cooperating centers such as Odences, Christie and Liverpool in order to implement an effective model by handling patients' data leaking issue.	Dataset from three hospitals was taken in consideration and the C-indices for Odences is 0.75, for Christie – 0.65 and for Liverpool, its 0.69.	Implementing these types of models is easier when hospitals start contributing with their whole data without the need of any legal agreements.
R. Durga and E. Poovammal [17]	Deep Learning Blockchain	A deep Learning based model which is trained in FL environment has proposed to detect COVID-19 from CT-scans. For the secured data retrieval of FL, a blockchain architecture is used.	Accuracy of 98.2%	Latency of the blockchain can be reduced.
J. Liu et al.[45]	Dual Attention Gates (DAGs)-U-Net	In the proposed model, "feature extraction and vertebral body segmentation" is done using DAF-U-Net. The selection of U-Net is because of its high efficiency during segmentation of medical images. The name given to this model is "Federated Learning-based Vertebral Body Segment Framework (FLVBSF)"	Approx 98% accuracy with U-Net based DAGs	"Not mentioned"
D. G. Nair, J. J. Nair, K. Jaideep Reddy and C. V. Aswartha Narayana [50]	Support Vector Machine	The proposed system is for the detection of Facial Paralysis. Here, the dataset separately consists of Facial Paralytic patients' face images and normal images. The classification model used in this study is Support vector Machine (SVM). While training, only the authentic participant (ensure the authenticity before making him/her a participant) contributed in the training process. During every iteration, updated parameters are sent to the server further the clustering at server-side is done using Fuzzy C-mean clustering algorithm. The motive of ensuring patients privacy is fully achieved here.	Accuracy approx 91%	The proposed system is for the detection of Facial Paralysis. Here, the dataset separately consists of Facial Paralytic patients' face images and normal images. The classification model used in this study is Support vector Machine (SVM). While training, only the authentic participant (ensure the authenticity before making him/her a participant) contributed in the training process. During every iteration, updated parameters are sent to the server further the clustering at server-side is done using Fuzzy C-mean clustering algorithm. The motive of ensuring patients privacy is fully achieved here.

| Show Table

DownLoad: CSV

Table 2. Federated Learning Aggregation Algorithms

Algorithm Name	Description
Basic FL aggregation Algorithm
Federated Stochastic Gradient Descent (FedSGD) [70]	As, stochastic gradient descent (SGD) had produced excellent results in deep learning. So, the researchers opted to use SGD as the foundation for the FL training method. SGD can be crudely applied to the FL optimization problem, where in each round of communication; a "batch gradient computation" is performed. This method is computationally effective, but it needs a lot of training rounds to build good models.
Federated Averaging (FedAvg) [61]	FedAvg algorithm consists of n communications having: The server which chooses m clients and distribute a global model (Mi) to all clients. Each client partitions its local data into batches and performs E epochs of Stochastic Gradient Decent (SGD) for updating their local models. Finally, clients upload their trained local models to the server, which then computes a weighted sum of all received local models to generate the new global model Mi+1. In the next iteration, Mi+1 model is distributed to clients for training. Each iteration follows the same process [55].
Adaptive Federated Learning (AdaFed) algorithm[21]	The core concentration of this FL algorithm is on two functionalities such as a) The weighting to the clients contributing in the training process is based on their performance. b) The adaptive loss function is introduced in the FL setting [20]. The main motive is to improve the training process.
LocalFed [11]	Researchers may easily construct and run their ideas in a federated learning using the localfed federated learning. In contrast to others, localfed views federated learning as a collection of many elements that should be changed immediately and with few adjustments[93].
FedMA [74]	In FL systems, model parameters rise with every iteration, driving up communication costs. FedMA is a suitable model to manage this issue, which only extracts and transfers informative parameters after training. For this, the current FL averaging algorithm is combined with the information-gain based sampling technique to assist address the limits that come up during deployments with limited bandwidth [69].
Algorithms to handle heterogeneity issue in FL: Various FL algorithms are able to handle heterogeneity issue where server and clients are of different in nature FedProx, MHAT, FLOX are few algorithms which are especially designed to handle the heterogeneity in FL.
Federated Proximal (FedProx) Method [43]	As, in FedAvg is stochastic gradient descent (SGD) based which can cause issue when numerous amounts of local updates happen. The solution to this problem is FedProx which use "proximal point update" for local optimization [86] [42].
MHAT (Model-Heterogeneous Aggregation Training) [27]	This model uses the knowledge distillation in FL settings. The algorithm is optimal as compared to the existing aggregation algorithms. In this aggregation scheme, an auxiliary model is trained based on the updated sent by the clients. This algorithm consists of four steps: firstly, each client chooses its network architecture. Each client sends the updates to the server in the following phase. Third, a server-trained auxiliary model is used to combine the "probability distributions" uploaded by each client. The auxiliary model is then used to make a prediction on the publicly available data, and the results are subsequently broadcast to each client. Then, once the average accuracy of each local model converges, each client updates its training samples by include the probability distribution corresponding to the public data given by the server, then go back to step one and repeating the entire process.
Federated Learning on funcX (FLOX) [37]	This is a heterogeneous aggregation algorithm. In cross-device FL, when the server model and client models are not identical then there is the need of some heterogeneous aggregation algorithms. In this training process, firstly the global model is divided into sub-models at server and then each client got a sub-model for training. After training, clients' send the updates to server for aggregation. Same process continues for number of iterations. During every iteration, the client trains the same sub-model (every tith iteration train the updated sub-model aggregated after ti-1th iteration) on its local data.
Attentive Aggregation: The aggregation algorithms aggregate the local updates at server-side without bothering about the contribution of each client. The problem can be resolved using some attentive algorithms in FL settings.
Attentive Federated Optimization (FedAtt) [30]	The researchers claimed that the communication cost decreases by using this aggregation algorithm. This aggregation algorithm is implemented using attentive algorithms with FL to optimize the aggregation process at server-side. This algorithm ensures the optimization at server-side through the aggregation of the "layer-wise contribution" of neural model of selected clients.
Federated Attentive Message Passing (FedAMP) [26]	FedAMP is an attentive aggregation algorithm in FL. In the non-IID settings, heterogeneous clients are involved in the training process. In this pool of clients, few clients share similar properties and can work in collaboration with preaching their privacies. The main motive of this algorithm is to increase the collaboration among similar type of clients with massage passing, so to get an efficient model after training it up to preset count of iterations.
HeurFedAMP [26]	HeurFedAMP [26] This aggregation algorithm is the Improvement of FedAMP. It is required when the clients in FL system are using Deep Neural Networks.
Adaptive Aggregation algorithms: Extensive aggregation algorithms have been proposed for "non-convex distributed problems". But still FL faces lots of challenges, such as communication issue as the sizes of datasets and models increase in every iteration. These algorithms also need large number of training iterations. These problems can be resolved by including adaptivity in FL algorithms. Few Adaptive optimization FL algorithms are listed below:
FedADAGRAD, FedADAM, FedYOGI [61]	As per the available results of FedAvg algorithm, the coverage rate is high with homogenous clients, but it's low when clients are heterogeneous in nature. Author claimed that these three adaptive FL algorithms are first of its type. These adaptive algorithms are used on the server side and on client's side SGD has been used. Authors ensure that these methods have comparable communication cost and work efficiency as FEDAVG in cross-device settings. These algorithms are evaluated on image as well as text data and have shown good results. The efficiency of FedYogi is similar to FedAdam.
Faster adaptive FL algorithm (FAFED) [76]	The adaptive aggregation algorithms discussed above, adjust the adaptive learning rate on the server side only. But same adaptivity is not applied on client side and to achieve it extra global learning rate is required. This drawback of these adaptive aggregation FL algorithms is resolved in FAFED. It has "shared adaptive learning rates" to resolve the issues related to heterogeneous data.
Federated Mediation (FedMed) [75]	This adaptive algorithm gave chance to train the global model to only those clients (workers) those have less loss value. Firstly, the loss value on dataset is calculated of each local client. Further the calculated loss value of each client is stored in an array, the mediator of FedMed check the array and detects the clients having less loss value. The detected clients got the chance to train the global model. This process continues in all iterations.

| Show Table

DownLoad: CSV

Table 3. Federated Learning Frameworks

Name of Platform	Link
FATE [92]	https://fate.fedai.org/
Substra [91]	https://www.substra.ai/
TensorFlow Federated [90]	https://www.tensorflow.org/federated
IBM Federated Learning [48]	https://ibmfl.mybluemix.net/
Flower [8]	https://flower.dev/
FedML[89]	https://fedml.ai/

| Show Table

DownLoad: CSV

Table 4. Links of GitHub repositories of FL models

Name of Disease	Description	Link
COVID-19 detection	FL model for COVID-19 detection form lung CT scans.	https://github.com/med-air/FL-COVID
LocalFed	localfed is a federated learning framework that allows researchers to easily build their theories and run them in a federated learning context.	https://github.com/arafeh94/localfed
FedGraphNN	FedGraphNN FedGraphNN is a system that helps in research on federated GNNs. The results are showing that this system is secure and computationally efficient.	https://github.com/FedML-AI/FedML/tree/master/python/app/fedgraphnn

| Show Table

DownLoad: CSV

Table 5. Links of GitHub repositories of FL models

Measures	Equation	References
Accuracy	$ \frac{T\rho+T\eta}{T\rho+T\eta+F\rho+F\eta} *100 $	[10], [38], [13], [36], [58], [46], [47], [28], [77], [35], [7], [79], [64], [29], [1], [85], [9], [53], [25], [51], [66], [17], [45], [50]
Sensitivity (Recall)	$ \frac{T\rho}{T\rho+F\eta} $	[10], [38], [40], [47], [28], [64], [35], [29], [85], [9], [25] [51], [17], [45], [50]
Precision	$ \frac{T\rho}{T\rho+F\rho} $	[10], [40], [47], [28], [64], [35], [29], [85], [9], [25] [51], [17], [50]
Specificity	$ \frac{T\eta}{T\eta+F\rho} $	[51], [17], [45]
F1-Score	$ 2* \frac{Precision*Recall}{Precision+Recall} $	[10], [40], [4], [47], [28], [64], [35], [29], [85], [9], [17]
ROC- AUC Score	$ \int_{a}^{b} f(x) dx $	[10], [35], [9]
PR-AUC Score	$ \int_{a}^{b} f(x) dx $	[10], [36], [22]
Root Mean Square Error (RMSE)	$ \sqrt{\frac{1}{n}} \sum_{i=1}^{n} \left(\frac{d_i-f_i}{\sigma_i}\right)^2 $	[44]
Cohen's Kappa	$ k \equiv \frac{{p_0 - p_e}}{{1 - p_e}} = 1 - \frac{{1 - p_0}}{{1 - p_e}} $	[35]
Harell 's C-indices	$ \frac{{\sum_{i \neq j}^{} 1 \{ \eta_i< \eta_j \} (T_i> T_j) d_j}}{{\sum_{i \neq j}^{} 1 (T_i - T_j) d_j}} $	[22]
Dice similarity coefficient	$ \frac{2T\rho}{2T\rho+F\rho+F\eta} $	[45]
Jccard similarity coefficient	$ \frac{T\rho}{T\rho+F\rho+F\eta} $	[45]

| Show Table

DownLoad: CSV

Related Papers

Cited by

References

[1]	M. Abdul Salam, S. Taha and M. Ramadan, COVID-19 detection using federated machine learning, Public Library of Science San Francisco, 16 (2021), e0252573.
[2]	Agrawal and Pulkit, et al., Data platform for machine learning, Proceedings of the 2019 International Conference on Management of Data, (2019), 1803-1816.
[3]	S. Alam, L. Liu, M. Yan and M. Zhang, FedRolex: Model-heterogeneous federated learning with rolling sub-model extraction, preprint, (2022), arXiv: 2212.01548.
[4]	J. R. Amalraj and R. Lourdusamy, Security and privacy issues in federated healthcare - An overview, Open Comput. Science, 12 (2022), 57-65.
[5]	K. S. Arikumar, S. B. Prathiba, M. Alazab, T. R. Gadekallu, S. Pandya, J. M. Khan and R. S. Moorthy, Federated learning-based person movement identification through wearable devices in smart healthcare systems, Sensors, 22 (2022), 1377.
[6]	S. Baghersalimi, T. Teijeiro, D. Atienza and A. Aminifar, Personalized real-time federated learning for epileptic seizure detection, IEEE Journal of Biomedical and Health Informatics, 26 (2021), 898-909.
[7]	P. Baheti, M. Sikka, K. V. Arya and R. Rajesh, Federated learning on distributed medical records for detection of lung nodules, Proceedings of the 15th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, (2020), 445-451.
[8]	D. J. Beutel, T. Topal, A. Mathur, X. Qiu, J. Fernandez-Marques, Y. Gao, L. Sani, K. H. Li, T. Parcollet and P. P. B. de Gusmão, et al., Flower: A friendly federated learning framework, (2022).
[9]	A. Bhattacharya, R. Rana, V. Udutalapally and D. Das, CoviFL: Edge-assisted federated learning for remote COVID-19 detection in an AIoMT framework, IEEE Symposium on Computers and Communications (ISCC), (2022), 1-6.
[10]	T. Borger, et al., Federated learning for violence incident prediction in a simulated cross-institutional psychiatric setting, Expert Syst. Appl., 199 (2022), 2371-2392.
[11]	M. Chahoud, et al., A Dynamic and Efficient Multi-Criteria Federated Learning Client Deployment Scheme, 2022. Available from: http://arXiv.org/abs/2211.02906.
[12]	Y. Chang, C. Fang and W. Sun, et al., A blockchain-based federated learning method for smart healthcare, Computational Intelligence and Neuroscience, 2021, (2021).
[13]	Y. Chen, B. Wang, H. Jiang, P. Duan, Y. Ping and Z. Hong, PEPFL A framework for a practical and efficient privacy preserving federated learning, Digit Communication Networks, (2022).
[14]	T. K. Dang, X. Lan, J. Weng and M. Feng, Federated learning for electronic health records, ACM Trans. Internet Technol., 13 (2022), 1-17.
[15]	E. Darzidehkalani, M. Ghasemi-rad and P. m. a. van Ooijen, Federated learning in medical imaging. Part II: Methods, challenges, and considerations, Journal of the American College of Radiology, 19 (2022), 975-982.
[16]	Q. Dou, T. Y. So, M. Jiang, Q. Liu, V. Vardhanabhuti, G. Kaissis, Z. Li, W. Si, H. H. Lee and K. Yu, et al., Federated deep learning for detecting covid lung abnormalities, NPJ Digital Medicine, 4 (2021), 60.
[17]	R. Durga and E. Poovammal, FLED-block: Federated learning ensembled deep learning blockchain model for COVID-19 prediction, Frontiers in Public Health, 10 (2022).
[18]	I. Feki, S. Ammar, Y. Kessentini and K. Muhammad, Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information, Elsevier Connect, (2020).
[19]	L. Gao, L. Li, Y. Chen, C. Z. Xu and M. Xu, FGFL A blockchain-based fair incentive governor for federated learning, IEEE Sens., 163 (2022), 283-299.
[20]	A. Giuseppi, L. Della Torre, D. Menegatti, F. D. Priscoli, A. Pietrabissa and C. Poli, An adaptive model averaging procedure for federated learning, J. Adv. Inf. Technology, 13 (2022), 539-548.
[21]	A. Giuseppi, S. Manfredi, D. Menegatti, C. Poli and A. Pietrabissa, Decentralised federated learning for hospital networks with application to COVID 19 detection, IEEE Accessy, 10 (2022), 92681-92691.
[22]	C. R. Hansen, G. Price, M. Field, N. Sarup, R. Zukauskaite, J. Johansen, J. G. Eriksen, F. Aly, A. McPartlin and L. Holloway, et al., Larynx cancer survival model developed through open-source federated learning, Radiotherapy and Oncology, 176 (2022), 179-186.
[23]	A. L Heureux, K. Grolinger, H. F. Elyamany and M. A. M. Capretz, Machine learning with big data: Challenges and approaches, IEEE Access, 5 (2017), 7776-7797.
[24]	T.-T. Ho, K.-D. Tran and Y. Huang, FedSGDCOVID: Federated SGD COVID-19 detection under local differential privacy using chest X-ray images and symptom information, Sensors, 22 (2022), 3728.
[25]	M. N. Hossen, V. Panneerselvam, D. Koundal, K. Ahmed, F. M. Bui and S. M. Ibrahim, Federated machine learning for detection of skin diseases and enhancement of internet of medical things (IoMT) security, IEEE Journal of Biomedical and Health Informatics, (2022).
[26]	Y. Huang, et al., Personalized cross-silo federated learning on non-iid data, Proceedings of the AAAI Conference on Artificial Intelligence, 35 (2021), 7865-7873.
[27]	L. Hu, H. Yan, L. Li, Z. Pan, X. Liu and Z. Zhang, MHAT An efficient model-heterogenous aggregation training scheme for federated learning, Inf. Sci. (Ny), 560 (2021), 493-503.
[28]	O. Ibitoye, M. O. Shafiq and A. Matrawy, DiPSeN: Differentially private self-normalizing neural networks for adversarial robustness in federated learning, preprint, (2021), arXiv: 2101.03218.
[29]	S. M. Jalal, M. R. Hasan, M. A. Haque and M. G. R. Alam, A horizontal federated random forest for heart disease detection from decentralized local data, IEEE 10th Region 10 Humanitarian Technology Conference R10 HTC, (2022), 191-196.
[30]	S. Ji, S. Pan, G. Long, X. Li, J. Jiang and Z. Huang, Learning private neural language modeling with attentive aggregation, Proc. Int. Jt. Conf. Neural Networks, (2019).
[31]	S. Ji, T. Saravirta, S. Pan, G. Long and A. Walid, Emerging trends in federated learning: From model fusion to federated x learning, preprint, (2021), arXiv: 2102.12920.
[32]	A. Jiménez-Sánchez, M. Tardy, M. A. G. Ballester and G. Piella, Memory-aware curriculum federated learning for breast cancer classification, Computer Methods and Programs in Biomedicine, 229 (2023), 107318.
[33]	M. Joshi, A. Pal and M. Sankarasubbu, Federated learning for healthcare domain - Pipeline, applications and challenges, ACM Transactions on Computing for Healthcare, 3 (2022), 1-36.
[34]	P. Kairouz, H. B. McMahan and B. AvenT, et al., Advances and Open Problems in Federated Learning, 2019. Available from: http://arXiv.org/abs/1912.04977.
[35]	S. H. Khan and M. G. R. Alam, A federated learning approach to pneumonia detection, International Conference on Engineering and Emerging Technologies (ICEET), (2021), 1-6.
[36]	C. Koetsier, J. Fiosina, J. N. Gremmel, J. P. Müller, D. M. Woisetschläger and M. Sester, Detection of anomalous vehicle trajectories using federated learning, ISPRS Open J. Photogramm. Remote Sens., 4 (2022), 1000013.
[37]	N. Kotsehub, et al., FLoX federated learning with FaaS at the edge, Proceedings of IEEE 18th Int. Conference e Science, (2022), 11-20.
[38]	R. Kumar, et al., Blockchain federated learning and deep learning models for COVID 19 detection using CT imaging, IEEE Sens., 21 (2022), 16301-16314.
[39]	M. Kumar and A. Singh, Probabilistic data structures in smart city: Survey, applications, challenges, and research directions, J. Ambient Intell. Smart Environ., 14 (2022), 229-284.
[40]	T. T. Kuo and A. Pham, Detecting model misconducts in decentralized healthcare federated learning, Int. J. Med. Inform., 158 (2022), 104658.
[41]	W. Lee, Reward-based participant selection for improving federated reinforcement learning, ICT Express, (2022).
[42]	T. Li, A. K. Sahu, M. Zaheer, M. Sanjabi, A. Talwalkar and V. Smith, Federated optimization in heterogeneous networks, preprint, (2018), arXiv: 1812.06127.
[43]	T. Li, A. K. Sahu, M. Zaheer, M. Sanjabi, A. Talwalkar and V. Smith, Federated optimization in heterogeneous networks, preprint, (2018), arXiv: 1812.06127.
[44]	Y. Liu, J. Nie, X. Li, S. H. Ahmed, W. Y. B. Lim, and C. Miao, Federated Learning in the Sky: Aerial-Ground Air Quality Sensing Framework with UAV Swarms, IEEE Internet Things, 8 (2021), 9827–9837.
[45]	J. Liu, X. Liang, R. Yang, Y. Luo, H. Lu, L. Li, S. Zhang and S. Yang, Federated learning-based vertebral body segmentation, Engineering Applications of Artificial Intelligence, 116 (2022), 105451.
[46]	Li u, J. Cheng, X. Wang, X. Lu and J. Yin, Hybrid differential privacy based federated learning for internet of things, J. Syst. Archit, 124 (2022), 102418.
[47]	J. Lo, et al., Federated learning for microvasculature segmentation and diabetic retinopathy classification of OCT data, Ophthalmol. Sci., 1 (2021), 100069.
[48]	H. Ludwig, et al., Ibm federated learning: An enterprise framework white paper v0. 1, preprint, (2020), arXiv: 2007.10987.
[49]	H. B. Mcmahan, F. X. Yu, A. T. Suresh, D. Bacon and P. Richt, C. O. E. Fficiency, F$1$ S. I. C. E, (2017), 1-10.
[50]	D. G. Nair, J. J. Nair, K. J. Reddy and C. A. Narayana, A privacy preserving diagnostic collaboration framework for facial paralysis using federated learning, Engineering Applications of Artificial Intelligence, 116 (2022), 105476.
[51]	M. Nasajpour, M. Karakaya, S. Pouriyeh and R. M. Parizi, Federated transfer learning for diabetic retinopathy detection using CNN architectures, SoutheastCon 2022, (2022), 655-660.
[52]	S. Naz, K. T. Phan and Y. P. P. Chen, A comprehensive review of federated learning for COVID-19 detection, Int. J. Intell. Syst., 37 (2022), 2371-2392.
[53]	T. Ngo, D. C. Nguyen, P. N. Pathirana, L. A. Corben, M. B. Delatycki, M. Horne, D. J. Szmulewicz and M. Roberts, Federated deep learning for the diagnosis of cerebellar ataxia: Privacy preservation and auto-crafted feature extractor, IEEE Transactions on Neural Systems and Rehabilitation Engineering, 30 (2022), 803-811.
[54]	L. T. Nguyen, J. Kim and B. Shim, Gradual federated learning with simulated annealing, IEEE Transactions on Signal Processing, 69 (2021), 6299-6313.
[55]	A. Nilsson, S. Smith, G. Ulm, E. Gustavsson and M. Jirstrand, A performance evaluation of federated learning algorithms, Proc. 2nd Work. Distrib. Infrastructures Deep Learn. Part Middlew, (2018), 1-18.
[56]	L. Peng, N. Wang, N. Dvornek, X. Zhu and X. Li, Fedni: Federated graph learning with network inpainting for population-based disease prediction, IEEE Transactions on Medical Imaging, (2022)
[57]	B. Pfitzner, N. Steckhan and B. Arnrich, Federated learning in a medical context a systematic literature review, ACM Trans. Internet Technol., 21 (2021), 1-31.
[58]	D. Połap and M. Woźniak, A hybridization of distributed policy and heuristic augmentation for improving federated learning approach, Neural Networks, 146 (2022), 130-140.
[59]	A. Qayyum, K. Ahmad, M. A. Ahsan, A. Al-Fuqaha and J. Qadir, Collaborative federated learning for healthcare: Multi-modal covid-19 diagnosis at the edge, IEEE Open Journal of the Computer Society, 3 (2022), 172-184.
[60]	F. Qian and A. Zhang, The value of federated learning during and post-COVID-19, International Journal for Quality in Health Care, 33 (2021).
[61]	S. Reddi, et al., Adaptive Federated Optimization, 2020. Available from: http://arXiv.org/abs/2003.00295.
[62]	H. Ren, D. Anicic and T. A. Runkler, TinyReptile: TinyML with federated meta-learning, preprint, (2023), arXiv: 2304.05201.
[63]	N. Rieke, et al., The future of digital health with federated learning, NPJ Digital Medicine, 3 (2020), 119.
[64]	O. Rudovic, et al., Personalized federated deep learning for pain estimation from face images, preprint, (2021), arXiv: 2101.04800.
[65]	M. Shaheen, M. S. Farooq, T. Umer and and B. S. Kim, Applications of federated learning. Taxonomy, challenges, and research trends, Electron., 11 (2022), 670.
[66]	H. Shamseddine, S. Otoum and A. Mourad, A federated learning scheme for neuro-developmental disorders: Multi-aspect ASD detection, preprint, (2022), arXiv: 2211.00643.
[67]	M. J. Sheller, B. Edwards, G. A. Reina, J. Martin, S. Pati, A. Kotrotsou, M. Milchenko, W. Xu, D. Marcus and R. R. Colen, et al., Federated learning in medicine facilitating multi institutional collaborations without sharing patient data, Springer, 10 (2020), 1-12.
[68]	P. P. Shinde, A review of machine learning and deep learning applications, Fourth Int. Conf. Comput. Commun. Control Autom., (2018), 1-6.
[69]	S. Shukla and N. Srivastava, Federated matched averaging with information-gain based parameter sampling, preprint, (2021).
[70]	S. Singh, Federated Optimization Algorithms FedSGD and FedAvg, 2023. Available from: https: //shreyansh26.github.io/post/2021-12-18_federated_optimization_fedavg/.
[71]	K. Singhal, H. Sidahmed, Z. Garrett, S. Wu, K. Rush and S. Prakash, Federated reconstruction: Partially local federated learning, Adv. Neural Inf. Process, 14 (2021), 11220-11231.
[72]	C. Thapa, P. C. M. Arachchige, S. Camtepe and L. Sun, SplitFed when federated learning meets split learning, Proc. 36th AAAI Conf. Artif. Intell. AAAI 2022, 36 (2021), 8485-8493.
[73]	N. Thapa, Z. Liu, D. B. Kc, B. Gokaraju and K. Roy, Comparison of machine learning and deep learning models for network intrusion detection systems, Futur. Internet, 12 (2020), 1-16.
[74]	H. Wang, M. Yurochkin, Y. Sun, D. Papailiopoulos and Y. Khazaeni, Federated Learning with Matched Averaging, 2020. Available from: http://arXiv.org/abs/2002.06440.
[75]	X. Wu, Z. Liang and J. Wang, Fedmed A federated learning framework for language modeling, Sensors (Switzerland), 20 (2020), 1-17.
[76]	X. Wu, F. Huang, Z. Hu and H. Huang, Faster adaptive federated learning, preprint, (2022), arXiv: 2212.00974.
[77]	B. Xiong, X. Yang, F. Qi and C. Xu, A unified framework for multi-modal federated learning, Neurocomputing, 480 (2022), 110-118.
[78]	J. Xu, B. S. Glicksberg, C. Su, P. Walker, J. Bian and F. Wang, Federated learning for healthcare informatics, J. Healthc. Informatics Res., 5 (2021).
[79]	X. Xu, H. Peng, L. Sun, M. Z. A. Bhuiyan, L. Liu and L. He, Fedmood: Federated learning on mobile health data for mood detection, preprint, (2021), arXiv: 2102.09342.
[80]	D. Yang, Z. Xu, W. Li, A. Myronenko and H. R. Roth, Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID- 19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information, Elsevier, (2020).
[81]	Z. Yang, M. Chen, K. K. Wong, H. V. Poor and S. Cui, Federated learning for 6G: Applications, challenges, and opportunities, Engineering Elsevier Ltd., 8 (2022), 33-41.
[82]	Q. Yng, et al., Vertical federated learning, Fed. Learn., (2020), 69-82.
[83]	T. Yu, E. Bagdasaryan and V. Shmatikov, Salvaging Federated Learning by Local Adaptation,
[84]	B. Yuan, S. Ge and W. Xing, A federated learning framework for healthcare iot devices, preprint, (2020), arXiv: 2005.05083.
[85]	X. Yuan, J. Zhang, J. Luo, J. Chen, Z. Shi and M. Qin, An efficient digital twin assisted clustered federated learning algorithm for disease prediction, IEEE 95th Vehicular Technology Conference: (VTC2022-Spring), (2022), 1-6.
[86]	X.-T. Yuan and P. Li, On convergence of FedProx: Local dissimilarity invariant bounds, non-smoothness and beyond, preprint, (2018), arXiv: 2206.05187.
[87]	W. Zhang, et al., Dynamic-fusion-based federated learning for COVID-19 detection, IEEE Internet Things, 8 (2021), 15884-15891. 2020. Available from: http://arXiv.org/abs/2002.04758.
[88]	Y. Zhu, Z. Liu, P. Wang and C. Du, A dynamic incentive and reputation mechanism for energy-efficient federated learning in 6g, Digital Communications and Networks, (2022).
[89]	FedML-AI/FedML: FedML, 2023. Available from: https: //github.com/FedML-AI/FedML.
[90]	GitHub - Tensorflow/federated A Framework for Implementing Federated Learning, 2023. Available from: https: //github.com/tensorflow/federated.
[91]	Substra documentation–Substra 0.24.0 documentation, 2023. Available from: https: //docs.substra.org/en/stable/.
[92]	GitHub FederatedAI FATE An Industrial Grade Federated Learning Framework, 2023. Available from: https: //github.com/FederatedAI/FATE.
[93]	GitHub Arafeh94 Localfed, 2023. Available from: https: //github.com/arafeh94/localfed.