A fuzzy rank-based deep ensemble methodology for multi-class skin cancer classification DOI Creative Commons

Arindam Halder,

A Dalal,

S. Gharami

et al.

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Feb. 20, 2025

Skin cancer is widespread and can be potentially fatal. According to the World Health Organisation (WHO), it has been identified as a leading cause of mortality. It essential detect skin early so that effective treatment provided at an initial stage. In this study, widely-used HAM10000 dataset, containing high-resolution images various lesions, employed train evaluate. Our methodology for dataset involves balancing imbalanced by augmenting followed splitting into train, test validation set, preprocessing images, training individual models Xception, InceptionResNetV2 MobileNetV2, then combining their outputs using fuzzy logic generate final prediction. We examined performance ensemble standard metrics like classification accuracy, confusion matrix, etc. achieved impressive accuracy 95.14% result demonstrates effectiveness our approach in accurately identifying lesions. To further assess efficiency model, additional tests have performed on DermaMNIST from MedMNISTv2 collection. The model performs well transcends benchmark 76.8%, achieving 78.25%. Thus efficient classification, showcasing its potential clinical applications.

Language: Английский

MedMNIST v2 - A large-scale lightweight benchmark for 2D and 3D biomedical image classification DOI Creative Commons
Jiancheng Yang, Rui Shi, Donglai Wei

et al.

Scientific Data, Journal Year: 2023, Volume and Issue: 10(1)

Published: Jan. 19, 2023

Abstract We introduce MedMNIST v2 , a large-scale MNIST-like dataset collection of standardized biomedical images, including 12 datasets for 2D and 6 3D. All images are pre-processed into small size 28 × (2D) or (3D) with the corresponding classification labels so that no background knowledge is required users. Covering primary data modalities in designed to perform on lightweight 3D various scales (from 100 100,000) diverse tasks (binary/multi-class, ordinal regression, multi-label). The resulting dataset, consisting 708,069 9,998 total, could support numerous research/educational purposes image analysis, computer vision, machine learning. benchmark several baseline methods v2, 2D/3D neural networks open-source/commercial AutoML tools. code publicly available at https://medmnist.com/ .

Language: Английский

Citations

347
Recent progress in transformer-based medical image analysis DOI
Zhaoshan Liu, Qiujie Lv, Ziduo Yang

et al.

Computers in Biology and Medicine, Journal Year: 2023, Volume and Issue: 164, P. 107268 - 107268

Published: July 20, 2023

Language: Английский

Citations

59
PMC-CLIP: Contrastive Language-Image Pre-training Using Biomedical Documents DOI

Weixiong Lin,

Ziheng Zhao, Xiaoman Zhang

et al.

Lecture notes in computer science, Journal Year: 2023, Volume and Issue: unknown, P. 525 - 536

Published: Jan. 1, 2023

Language: Английский

Citations

44
A generalist vision–language foundation model for diverse biomedical tasks DOI
Kai Zhang, Rong Zhou, Eashan Adhikarla

et al.

Nature Medicine, Journal Year: 2024, Volume and Issue: 30(11), P. 3129 - 3141

Published: Aug. 7, 2024

Language: Английский

Citations

30
Quantum Vision Transformers DOI Creative Commons
El Amine Cherrat,

Iordanis Kerenidis,

Natansh Mathur

et al.

Quantum, Journal Year: 2024, Volume and Issue: 8, P. 1265 - 1265

Published: Feb. 22, 2024

In this work, quantum transformers are designed and analysed in detail by extending the state-of-the-art classical transformer neural network architectures known to be very performant natural language processing image analysis. Building upon previous which uses parametrised circuits for data loading orthogonal layers, we introduce three types of training inference, including a based on compound matrices, guarantees theoretical advantage attention mechanism compared their counterpart both terms asymptotic run time number model parameters. These can built using shallow produce qualitatively different classification models. The proposed layers vary spectrum between closely following exhibiting more characteristics. As building blocks transformer, propose novel method matrix as states well two new trainable adaptable levels connectivity quality computers. We performed extensive simulations standard medical datasets that showed competitively, at times better performance benchmarks, best-in-class vision transformers. trained these small-scale require fewer parameters benchmarks. Finally, implemented our superconducting computers obtained encouraging results up six qubit experiments.

Language: Английский

Citations

22
A comprehensive survey on deep active learning in medical image analysis DOI
Haoran Wang, Qiuye Jin, Shiman Li

et al.

Medical Image Analysis, Journal Year: 2024, Volume and Issue: 95, P. 103201 - 103201

Published: May 13, 2024

Language: Английский

Citations

20
Neural Architecture Search for biomedical image classification: A comparative study across data modalities DOI
Zeki Kuş, Musa Aydın, Berna Kıraz

et al.

Artificial Intelligence in Medicine, Journal Year: 2025, Volume and Issue: 160, P. 103064 - 103064

Published: Jan. 5, 2025

Language: Английский

Citations

2
Personalized Federated Learning with Adaptive Batchnorm for Healthcare DOI
Lu Wang, Jindong Wang, Yiqiang Chen

et al.

IEEE Transactions on Big Data, Journal Year: 2022, Volume and Issue: 10(6), P. 915 - 925

Published: May 23, 2022

There is a growing interest in applying machine learning techniques to healthcare. Recently, federated (FL) gaining popularity since it allows researchers train powerful models without compromising data privacy and security. However, the performance of existing FL approaches often deteriorates when encountering non-iid situations where there exist distribution gaps among clients, few previous efforts focus on personalization In this article, we propose FedAP tackle domain shifts obtain personalized for local clients. learns similarity between clients via statistics batch normalization layers while preserving specificity each client with different normalization. Comprehensive experiments five healthcare benchmarks demonstrate that achieves better accuracy compared state-of-the-art methods (e.g., 10%+ improvement PAMAP2) faster convergence speed.

Language: Английский

Citations

43
MetaFed: Federated Learning Among Federations With Cyclic Knowledge Distillation for Personalized Healthcare DOI
Yiqiang Chen, Lu Wang,

Xin Qin

et al.

IEEE Transactions on Neural Networks and Learning Systems, Journal Year: 2023, Volume and Issue: 35(11), P. 16671 - 16682

Published: July 28, 2023

Federated learning (FL) has attracted increasing attention to building models without accessing raw user data, especially in healthcare. In real applications, different federations can seldom work together due possible reasons such as data heterogeneity and distrust/inexistence of the central server. this article, we propose a novel framework called MetaFed facilitate trustworthy FL between federations. obtains personalized model for each federation server via proposed cyclic knowledge distillation. Specifically, treats meta distribution aggregates manner. The training is split into two parts: common accumulation personalization. Comprehensive experiments on seven benchmarks demonstrate that achieves better accuracy compared with state-of-the-art methods e.g., 10 $\%+$ improvement baseline physical activity monitoring dataset (PAMAP2) fewer communication costs. More importantly, shows remarkable performance real-healthcare-related applications.

Language: Английский

Citations

41
A quantum convolutional network and ResNet (50)-based classification architecture for the MNIST medical dataset DOI

Esraa Hassan,

M. Shamim Hossain, Abeer Saber

et al.

Biomedical Signal Processing and Control, Journal Year: 2023, Volume and Issue: 87, P. 105560 - 105560

Published: Oct. 7, 2023

Language: Английский

Citations

36