Comparison of classical machine learning deep learning to characterise fibrosis inflammation using quantitative MRI

Emily Chan, Matt Kelly, Julia A. Schnabel

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations

Abstract

The quantitative MRI metric, T1, has been used to characterise fibroinflammation in the liver; however, the T1 value alone is unable to differentiate between fibrosis and inflammation. We evaluate the potential utility of classical machine learning techniques (K-Nearest Neighbours, Support Vector Machine and Random Forest) to address this problem using information in the T1 map. We also compare to transfer learning, utilising multiple methods to alleviate the effects of class imbalance. Random Forest with Adaptive Synthetic Sampling was superior to mean T1 in categorising fibroinflammation. Despite the relatively small number of samples (n=289) and large class imbalance, our results demonstrate potential in using the whole T1 map with machine learning for this task.

Original languageEnglish
Title of host publication2021 IEEE 18th International Symposium on Biomedical Imaging, ISBI 2021
PublisherIEEE Computer Society
Pages729-732
Number of pages4
ISBN (Electronic)9781665412469
DOIs
StatePublished - 13 Apr 2021
Externally publishedYes
Event18th IEEE International Symposium on Biomedical Imaging, ISBI 2021 - Nice, France
Duration: 13 Apr 202116 Apr 2021

Publication series

NameProceedings - International Symposium on Biomedical Imaging
Volume2021-April
ISSN (Print)1945-7928
ISSN (Electronic)1945-8452

Conference

Conference18th IEEE International Symposium on Biomedical Imaging, ISBI 2021
Country/TerritoryFrance
CityNice
Period13/04/2116/04/21

Keywords

  • KNN
  • Liver
  • Random Forests
  • SVM
  • T1 mapping
  • Transfer Learning
  • Wideresnet

Fingerprint

Dive into the research topics of 'Comparison of classical machine learning deep learning to characterise fibrosis inflammation using quantitative MRI'. Together they form a unique fingerprint.

Cite this