Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Videos

A Dual-Reweighting Defense Strategy Against Data Poisoning Attacks in Medical Image Classification Models.

Xiaolong Yu¹, Yuping Zhou², Pengzhan Zheng¹

¹College of Computer, Minnan Normal University, Zhangzhou, Fujian, China.

Journal of Imaging Informatics in Medicine

|March 12, 2026

Summary

This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

Plumbagin Induces Ferroptosis in Nonfunctioning Pituitary Adenomas via Nrf2/FTH1-Dependent Ferritinophagy.

Drug design, development and therapy·2026

Same author

Non-classical NMDAR subunit GluN3A in a specialized hippocampal region regulates stress-coping strategies in male mice.

Communications biology·2026

Same author

Corrigendum to "Intermittent fasting attenuates ovarian chronic inflammation in obese mice by inhibiting the CXCL12-CXCR4 axis" [Int. Immunopharmacol. 179 (2026) 116603].

International immunopharmacology·2026

Same author

Zinc regulates the balance of specific gut microbiota through MTF1/Nrf2 and mitigates ferroptosis through activation of PINK1/Parkin/mitophagy via the gut-kidney axis in sepsis-associated acute kidney injury in pregnant mice.

Apoptosis : an international journal on programmed cell death·2026

Same author

CHST5 gene mutations contribute to high myopia by disrupting collagen fiber organization.

Journal of genetics and genomics = Yi chuan xue bao·2026

Same author

Synergistic effects of probiotics and low-FODMAP diet on clinical and inflammatory outcomes in ulcerative colitis: a retrospective cohort study.

Frontiers in medicine·2026

Same journal

Kolmogorov-Arnold Guided Local-Global Attention for Medical Image Classification.

Journal of imaging informatics in medicine·2026

Same journal

Artificial Intelligence-Assisted Inner Ear Computed Tomography Analysis: Radiomics-Based Comparison of Affected and Unaffected Ears in Idiopathic Sudden Sensorineural Hearing Loss.

Journal of imaging informatics in medicine·2026

Same journal

High Adoption, Higher Expectations: A Cross-Sectional Survey of Radiologist Engagement with Artificial Intelligence in the United Arab Emirates.

Journal of imaging informatics in medicine·2026

Same journal

Complex-valued Multi-scale Hybrid Attention Network for Fast MRI via Sparsified Data Learning.

Journal of imaging informatics in medicine·2026

Same journal

Automatic Phase and Sequence Identification in Gd-EOB-DTPA-Enhanced Liver MRI Using Deep Convolutional and Sequential Learning.

Journal of imaging informatics in medicine·2026

Same journal

Ultrasound-Based AI in Predicting Hormone Receptor Status in Breast Cancer: Is "Digital Biopsy" Possible.

Journal of imaging informatics in medicine·2026

See all related articles

This study introduces Dweighted, a novel defense scheme to protect deep learning models in medical imaging from data poisoning attacks. Dweighted enhances model security and accuracy against malicious threats.

Area of Science:

Artificial Intelligence
Medical Imaging
Cybersecurity

Background:

Deep learning models are advancing disease detection and medical image analysis.
Security concerns, particularly data poisoning attacks, threaten model performance and patient safety.
Malicious actors can compromise data or model parameters, leading to misdiagnoses.

Purpose of the Study:

To propose a novel defense scheme, Dweighted, to enhance the security of deep learning models against data poisoning attacks.
To improve the robustness and accuracy of medical image analysis models in the face of adversarial threats.
To develop a method for accurately identifying and eliminating malicious clients in federated learning settings.

Main Methods:

Dweighted integrates dual weighting with clustering analysis for dynamic client weight adjustment.

Keywords:

K-means clustering Data poisoning attack Dual weighting Federated learning Medical image classification Principal component analysis

Related Experiment Videos

It considers dataset size, model parameter differences, and similarity analysis.

Principal Component Analysis (PCA) and K-means clustering are employed to detect and remove malicious clients.

Main Results:

Dweighted significantly enhances global model security and robustness against data poisoning.
The scheme maintains high classification accuracy.
Achieved 94.89% overall accuracy and reduced attack success rate to 2.43% in the IID setting.

Conclusions:

Dweighted offers a robust defense against data poisoning attacks in medical deep learning.
The proposed method effectively balances security enhancement with classification performance.
Dweighted represents a significant advancement in securing AI for healthcare applications.