Continual Active Learning Using Pseudo-Domains for Limited Labelling Resources and Changing Acquisition Characteristics
Open Access
- 16 March 2022
- journal article
- research article
- Published by Machine Learning for Biomedical Imaging in Machine Learning for Biomedical Imaging
- Vol. 1 (IPMI 2021) , 1-28
Abstract
Machine learning in medical imaging during clinical routine is impaired by changes in scanner protocols, hardware, or policies resulting in a heterogeneous set of acquisition settings. When training a deep learning model on an initial static training set, model performance and reliability suffer from changes of acquisition characteristics as data and targets may become inconsistent. Continual learning can help to adapt models to the changing environment by training on a continuous data stream. However, continual manual expert labelling of medical imaging requires substantial effort. Thus, ways to use labelling resources efficiently on a well chosen sub-set of new examples is necessary to render this strategy feasible. Here, we propose a method for continual active learning operating on a stream of medical images in a multi-scanner setting. The approach automatically recognizes shifts in image acquisition characteristics – new domains –, selects optimal examples for labelling and adapts training accordingly. Labelling is subject to a limited budget, resembling typical real world scenarios. In order to avoid catastrophic forgetting while learning on new domains the proposed method utilizes a rehearsal memory. To demonstrate generalizability, we evaluate the effectiveness of our method on three tasks: cardiac segmentation, lung nodule detection and brain age estimation. Results show that the proposed approach outperforms other active learning methods on a continuous data stream with domain shifts.Keywords
All Related Versions
This publication has 23 references indexed in Scilit:
- A survey on active learning and human-in-the-loop deep learning for medical image analysisMedical Image Analysis, 2021
- Active, continual fine tuning of convolutional neural networks for reducing annotation effortsMedical Image Analysis, 2021
- Models GenesisMedical Image Analysis, 2021
- Importance Driven Continual Learning for Segmentation Across Domains2020
- Unlearning Scanner Bias for MRI Harmonisation in Medical Image Segmentation2020
- O‐MedAL: Online active deep learning for medical image analysisWIREs Data Mining and Knowledge Discovery, 2020
- Continual lifelong learning with neural networks: A reviewNeural Networks, 2019
- Validation, comparison, and combination of algorithms for automatic detection of pulmonary nodules in computed tomography images: The LUNA16 challengeMedical Image Analysis, 2017
- U-Net: Convolutional Networks for Biomedical Image Segmentation2015
- The Pascal Visual Object Classes (VOC) ChallengeInternational Journal of Computer Vision, 2009