Improving adversarial robustness of medical imaging systems via adding global attention noise

Dai, Yinglong; Qian, Yaguan; Liu, Feng; Wang, Bin; Gu, Zhaoquan; Wang, Wei; Wan, Jian; Zhang, Yanchun

doi:10.1016/j.compbiomed.2023.107251

Cited by 10 publications

(2 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The combination of large-scale supervised transfer learning with self-supervised learning was shown to improve the out-of-distribution generalization performance of medical imaging DNNs (32). The addition of Global Attention Noise during training (33), as well as adversarial training, where adversarial inputs are included in the training process (31), have been shown to improve the accuracy of medical imaging DNNs against adversarial attacks. Multi-task learning was used to address the specific challenges of prediction instability and explainability in the classification of smartphone photos of chest radiographs (21).…”

Section: Prior Work On Robustness Of Medical Imaging Ai Modelsmentioning

confidence: 99%

“…Our work makes the following contributions that go above and beyond the previous efforts. While noise-added training is a well-known technique to improve the robustness of neural networks ( 34 ) and has recently been applied to medical imaging specifically for adversarial robustness ( 31 , 33 ), our work applies it to achieve robustness to natural sources of noise. Input mixing and DCT-based denoising have not been previously applied to the medical imaging domain to the best of our knowledge.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

RoMIA: a framework for creating Robust Medical Imaging AI models for chest radiographs

Anand,

Krithivasan,

Roy

2024

Front. Radiol.

View full text Add to dashboard Cite

Artificial Intelligence (AI) methods, particularly Deep Neural Networks (DNNs), have shown great promise in a range of medical imaging tasks. However, the susceptibility of DNNs to producing erroneous outputs under the presence of input noise and variations is of great concern and one of the largest challenges to their adoption in medical settings. Towards addressing this challenge, we explore the robustness of DNNs trained for chest radiograph classification under a range of perturbations reflective of clinical settings. We propose RoMIA, a framework for the creation of Robust Medical Imaging AI models. RoMIA adds three key steps to the model training and deployment flow: (i) Noise-added training, wherein a part of the training data is synthetically transformed to represent common noise sources, (ii) Fine-tuning with input mixing, in which the model is refined with inputs formed by mixing data from the original training set with a small number of images from a different source, and (iii) DCT-based denoising, which removes a fraction of high-frequency components of each image before applying the model to classify it. We applied RoMIA to create six different robust models for classifying chest radiographs using the CheXpert dataset. We evaluated the models on the CheXphoto dataset, which consists of naturally and synthetically perturbed images intended to evaluate robustness. Models produced by RoMIA show 3%–5% improvement in robust accuracy, which corresponds to an average reduction of 22.6% in misclassifications. These results suggest that RoMIA can be a useful step towards enabling the adoption of AI models in medical imaging applications.

show abstract