Cohort Bias Adaptation in Aggregated Datasets for Lesion Segmentation

Brennan Nichyporuk*, Jillian Cardinell, Justin Szeto, Raghav Mehta, Sotirios Tsaftaris, Douglas L. Arnold, Tal Arbel

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Many automatic machine learning models developed for focal pathology (e.g. lesions, tumours) detection and segmentation perform well, but do not generalize as well to new patient cohorts, impeding their widespread adoption into real clinical contexts. One strategy to create a more diverse, generalizable training set is to naively pool datasets from different cohorts. Surprisingly, training on this big data does not necessarily increase, and may even reduce, overall performance and model generalizability, due to the existence of cohort biases that affect label distributions. In this paper, we propose a generalized affine conditioning framework to learn and account for cohort biases across multi-source datasets, which we call Source-Conditioned Instance Normalization (SCIN). Through extensive experimentation on three different, large scale, multi-scanner, multi-centre Multiple Sclerosis (MS) clinical trial MRI datasets, we show that our cohort bias adaptation method (1) improves performance of the network on pooled datasets relative to naively pooling datasets and (2) can quickly adapt to a new cohort by fine-tuning the instance normalization parameters, thus learning the new cohort bias with only 10 labelled samples.

Original languageEnglish
Title of host publicationDomain Adaptation and Representation Transfer, and Affordable Healthcare and AI for Resource Diverse Global Health - 3rd MICCAI Workshop, DART 2021, and 1st MICCAI Workshop, FAIR 2021, Held in Conjunction with MICCAI 2021, Proceedings
EditorsShadi Albarqouni, M. Jorge Cardoso, Qi Dou, Konstantinos Kamnitsas, Bishesh Khanal, Islem Rekik, Nicola Rieke, Debdoot Sheet, Sotirios Tsaftaris, Daguang Xu, Ziyue Xu
PublisherSpringer Science and Business Media Deutschland GmbH
Pages101-111
Number of pages11
ISBN (Print)9783030877217
DOIs
Publication statusPublished - 21 Sep 2021
Event3rd MICCAI Workshop on Domain Adaptation and Representation Transfer, DART 2021, and the 1st MICCAI Workshop on Affordable Healthcare and AI for Resource Diverse Global Health, FAIR 2021, held in conjunction with 24th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2021 - Virtual, Online
Duration: 27 Sep 20211 Oct 2021

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume12968 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference3rd MICCAI Workshop on Domain Adaptation and Representation Transfer, DART 2021, and the 1st MICCAI Workshop on Affordable Healthcare and AI for Resource Diverse Global Health, FAIR 2021, held in conjunction with 24th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2021
CityVirtual, Online
Period27/09/211/10/21

Keywords

  • Cohort bias
  • Deep learning
  • Instance norm conditioning
  • Label style
  • Multiple sclerosis
  • Segmentation
  • Trial merging

Fingerprint

Dive into the research topics of 'Cohort Bias Adaptation in Aggregated Datasets for Lesion Segmentation'. Together they form a unique fingerprint.

Cite this