Trust, but Verify: Robust Image Segmentation using Deep Learning

Fahim Ahmed Zaman; Xiaodong Wu; Weiyu Xu; Milan Sonka; Raghuraman Mudumbai

doi:10.1109/IEEECONF59524.2023.10476801

Back

Conference proceeding

Trust, but Verify: Robust Image Segmentation using Deep Learning

Fahim Ahmed Zaman, Xiaodong Wu, Weiyu Xu, Milan Sonka and Raghuraman Mudumbai

2023 57th Asilomar Conference on Signals, Systems, and Computers, pp.1070-1074

10/29/2023

DOI: 10.1109/IEEECONF59524.2023.10476801

View Online

Abstract

We describe a method for verifying the output of a deep neural network for medical image segmentation that is robust to several classes of random as well as worst-case perturbations i.e. adversarial attacks. This method is based on a general approach recently developed by the authors called "Trust, but Verify" wherein an auxiliary verification network produces predictions about certain masked features in the input image using the segmentation as an input. A well-designed auxiliary network will produce high-quality predictions when the input segmentations are accurate, but will produce low-quality predictions when the segmentations are incorrect. Checking the predictions of such a network with the original image allows us to detect bad segmentations. However, to ensure the verification method is truly robust, we need a method for checking the quality of the predictions that does not itself rely on a black-box neural network. Indeed, we show that previous methods for segmentation evaluation that do use deep neural regression networks are vulnerable to false negatives i.e. can inaccurately label bad segmentations as good. We describe the design of a verification network that avoids such vulnerability and present results to demonstrate its robustness compared to previous methods.

adversarial attack

Artificial neural networks

Closed box

Computers

Deep learning

Image segmentation

Measurement

Perturbation methods

robust deep learning

Details

Title: Subtitle: Trust, but Verify: Robust Image Segmentation using Deep Learning
Creators: Fahim Ahmed Zaman - University of Iowa
Xiaodong Wu - University of Iowa
Weiyu Xu - University of Iowa
Milan Sonka - University of Iowa
Raghuraman Mudumbai - University of Iowa
Resource Type: Conference proceeding
Publication Details: 2023 57th Asilomar Conference on Signals, Systems, and Computers, pp.1070-1074
DOI: 10.1109/IEEECONF59524.2023.10476801
eISSN: 2576-2303
Publisher: IEEE
Language: English
Date published: 10/29/2023
Academic Unit: Roy J. Carver Department of Biomedical Engineering; Electrical and Computer Engineering; Iowa Technology Institute; Radiation Oncology; The Iowa Institute for Biomedical Imaging; Fraternal Order of Eagles Diabetes Research Center; Injury Prevention Research Center; Ophthalmology and Visual Sciences
Record Identifier: 9984621036602771

Metrics

45 Record Views