A Feature-Pooling and Signature-Pooling Method for Feature Selection for Quantitative Image Analysis: Application to a Radiomics Model for Survival in Glioma

Zhenwei Shi*, Chong Zhang, Inge Compter, Maikel Verduin, Ann Hoeben, Danielle Eekers, Andre Dekker, Leonard Wee

*Corresponding author for this work

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

23 Downloads (Pure)

Abstract

We proposed a pooling-based radiomics feature selection method and showed how it would be applied to the clinical question of predicting one-year survival in 130 patients treated for glioma by radiotherapy. The method combines filter, wrapper and embedded selection in a comprehensive process to identify useful features and build them into a potentially predictive signature. The results showed that non-invasive CT radiomics were able to moderately predict overall survival and predict WHO tumour grade. This study reveals an associative inter-relationship between WHO tumour grade, CT-based radiomics and survival, that could be clinically relevant.
Original languageEnglish
Title of host publicationRadiomics and Radiogenomics in Neuro-oncology. RNO-AI 2019
EditorsH. Mohy-ud-Din, S. Rathore
PublisherSpringer, Cham
Pages70-80
ISBN (Electronic)978-3-030-40124-5
ISBN (Print)978-3-030-40123-8
DOIs
Publication statusPublished - 25 Feb 2020

Publication series

SeriesLecture Notes in Computer Science
Volume11991
ISSN0302-9743

Fingerprint

Dive into the research topics of 'A Feature-Pooling and Signature-Pooling Method for Feature Selection for Quantitative Image Analysis: Application to a Radiomics Model for Survival in Glioma'. Together they form a unique fingerprint.

Cite this