Towards continual learning in medical imaging

By Chaitanya Baweja, Ben Glocker, Konstantinos Kamnitsas. 2018

63 citations recorded by the source corpus.

This work investigates continual learning of two segmentation tasks in brain MRI with neural networks. To explore in this context the capabilities of current methods for countering catastrophic forgetting of the first task when a new one is learned, we investigate elastic weight consolidation, a recently proposed method based on Fisher information, originally evaluated on reinforcement learning of Atari games. We use it to sequentially learn segmentation of normal brain structures and then segmentation of white matter lesions. Our findings show this recent method reduces catastrophic forgetting, while large room for improvement exists in these challenging settings for continual learning.

Subfields: Critical Care and Intensive Care Medicine, Health Informatics, Radiology, Nuclear Medicine and Imaging

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Critical Care and Intensive Care Medicine

1. Emma Claire Robinson
2. Bernhard Kainz
3. Samuel Budd

In Health Informatics

1. Emma Claire Robinson
2. Bernhard Kainz
3. Samuel Budd

In Radiology, Nuclear Medicine and Imaging

1. Emma Claire Robinson
2. Bernhard Kainz
3. Christian Wachinger
4. Samuel Budd
5. Ramalingam Chellappa

Citations in the corpus

• Online Coreset Selection for Rehearsal-based Continual Learning (2021, neutral): “In light of this need, continual learning (CL) [40], or lifelong learning, tackles a learning scenario where a model continuously learns over a sequence of tasks [18, 25] within a broad research area, such as classification [15, 9], image generation [47], language learning [24, 4], clinical applicat”
• Learning Without Memorizing (2018, background): “The importance of incremental segmentation has already been underscored in [3].”
• Dynamic memory to alleviate catastrophic forgetting in continuous learning settings (2020, background): “n a test set of all three tasks. while learning segmentations of various brain regions [5]. Baweja et al. applied EWC to sequentially learn normal brain structure and white matter lesion segmentation [1]. Ravishankar et al. propose a pseudorehearsal technique and training of task-specic dense layers”
• A Survey on Active Learning and Human-in-the-Loop Deep Learning for Medical Image Analysis (2019, methodology): “Baweja et al. (2018) investigate continual learning of two MRI segmentation tasks with neural networks for countering catastrophic forgetting of the first task when a new one is learned. They investigate elastic weight consolidation, a method based on Fisher information to sequentially learn segment”
• Importance Driven Continual Learning for Segmentation Across Domains (2020, background): “CL approaches have also been proposed in medical imaging [10,2], but they do not follow the desiderata formulated above.”