Certification of Deep Learning Models for Medical Image Segmentation

被引:0
|
作者
Laousy, Othmane [1 ,2 ,3 ]
Araujo, Alexandre [4 ]
Chassagnon, Guillaume [2 ]
Paragios, Nikos [5 ]
Revel, Marie-Pierre [2 ]
Vakalopoulou, Maria [1 ,3 ]
机构
[1] Univ Paris Saclay, Cent Supelec, MICS, Gif Sur Yvette, France
[2] Paris Cite Univ, Hop Cochin, AP HP, Paris, France
[3] Inria Saclay, Gif Sur Yvette, France
[4] NYU, New York, NY USA
[5] Therapanacea, Paris, France
关键词
Certified Robustness; Randomized Smoothing; Denoising Diffusion Models; Segmentation; CHEST RADIOGRAPHS;
D O I
10.1007/978-3-031-43901-8_58
中图分类号
TP31 [计算机软件];
学科分类号
081202 ; 0835 ;
摘要
In medical imaging, segmentation models have known a significant improvement in the past decade and are now used daily in clinical practice. However, similar to classification models, segmentation models are affected by adversarial attacks. In a safety-critical field like healthcare, certifying model predictions is of the utmost importance. Randomized smoothing has been introduced lately and provides a framework to certify models and obtain theoretical guarantees. In this paper, we present for the first time a certified segmentation baseline for medical imaging based on randomized smoothing and diffusion models. Our results show that leveraging the power of denoising diffusion probabilistic models helps us overcome the limits of randomized smoothing. We conduct extensive experiments on five public datasets of chest X-rays, skin lesions, and colonoscopies, and empirically show that we are able to maintain high certified Dice scores even for highly perturbed images. Our work represents the first attempt to certify medical image segmentation models, and we aspire for it to set a foundation for future benchmarks in this crucial and largely uncharted area.
引用
收藏
页码:611 / 621
页数:11
相关论文
共 50 条
  • [1] Medical Ultrasound Image Segmentation With Deep Learning Models
    Wang, Chuantao
    Zhang, Jinhua
    Liu, Siyu
    [J]. IEEE ACCESS, 2023, 11 : 10158 - 10168
  • [2] Rethinking Boundary Detection in Deep Learning Models for Medical Image Segmentation
    Lin, Yi
    Zhang, Dong
    Fang, Xiao
    Chen, Yufan
    Cheng, Kwang-Ting
    Chen, Hao
    [J]. INFORMATION PROCESSING IN MEDICAL IMAGING, IPMI 2023, 2023, 13939 : 730 - 742
  • [3] DEEP LEARNING ARCHITECTURES FOR MEDICAL IMAGE SEGMENTATION
    Subramaniam, Sudha
    Jayanthi, K. B.
    Rajasekaran, C.
    Kuchelar, Ramani
    [J]. 2020 IEEE 33RD INTERNATIONAL SYMPOSIUM ON COMPUTER-BASED MEDICAL SYSTEMS(CBMS 2020), 2020, : 579 - 584
  • [4] Two-layer Ensemble of Deep Learning Models for Medical Image Segmentation
    Dang, Truong
    Nguyen, Tien Thanh
    McCall, John
    Elyan, Eyad
    Moreno-Garcia, Carlos Francisco
    [J]. COGNITIVE COMPUTATION, 2024, 16 (03) : 1141 - 1160
  • [5] Ensemble of deep learning models with surrogate-based optimization for medical image segmentation
    Truong Dang
    Anh Vu Luong
    Liew, Alan Wee Chung
    McCall, John
    Tien Thanh Nguyen
    [J]. 2022 IEEE CONGRESS ON EVOLUTIONARY COMPUTATION (CEC), 2022,
  • [6] Medical image segmentation using deep learning: A survey
    Wang, Risheng
    Lei, Tao
    Cui, Ruixia
    Zhang, Bingtao
    Meng, Hongying
    Nandi, Asoke K.
    [J]. IET IMAGE PROCESSING, 2022, 16 (05) : 1243 - 1267
  • [7] Deep Learning -Enhanced Image Segmentation for Medical Diagnostics
    Malathy, V.
    Maiti, Niladri
    Kumar, Nithin
    Lavanya, D.
    Aswath, S.
    Banu, Shaik Balkhis
    [J]. 2024 INTERNATIONAL CONFERENCE ON ADVANCES IN COMPUTING, COMMUNICATION AND APPLIED INFORMATICS, ACCAI 2024, 2024,
  • [8] Medical image semantic segmentation based on deep learning
    Jiang, Feng
    Grigorev, Aleksei
    Rho, Seungmin
    Tian, Zhihong
    Fu, YunSheng
    Jifara, Worku
    Adil, Khan
    Liu, Shaohui
    [J]. NEURAL COMPUTING & APPLICATIONS, 2018, 29 (05): : 1257 - 1265
  • [9] Variability and reproducibility in deep learning for medical image segmentation
    Renard, Felix
    Guedria, Soulaimane
    De Palma, Noel
    Vuillerme, Nicolas
    [J]. SCIENTIFIC REPORTS, 2020, 10 (01)
  • [10] Variability and reproducibility in deep learning for medical image segmentation
    Félix Renard
    Soulaimane Guedria
    Noel De Palma
    Nicolas Vuillerme
    [J]. Scientific Reports, 10