Supervised learning of functional maps for infarct classification

Anirban Mukhopadhyay, Ilkay Oksuz, Sotirios A. Tsaftaris

Research output: Contribution to journalArticlepeer-review

2 Citations (Scopus)

Abstract

© Springer International Publishing Switzerland 2016.Our submission to the STACOM Challenge at MICCAI 2015 is based on the supervised learning of functional map representation between End Systole (ES) and End Diastole (ED) phases of Left Ventricle (LV), for classifying infarcted LV from the healthy ones. The Laplace- Beltrami eigen-spectrum of the LV surfaces at ES and ED, represented by their triangular meshes, are used to compute the functional maps. Multi-scale distortions induced by the mapping, are further calculated by singular value decomposition of the functional map. During training, the information of whether an LV surface is healthy or diseased is known, and this information is used to train an SVM classifier for the singular values at multiple scales corresponding to the distorted areas augmented with surface area difference of epicardium and endocardium meshes. At testing similar augmented features are calculated and fed to the SVM model for classification. Promising results are obtained on both cross validation of training data as well as on testing data, which encourages us in believing that this algorithm will perform favourably in comparison to state of the art methods.
Original languageEnglish
Pages (from-to)162-170
Number of pages9
JournalLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Early online date9 Jan 2016
DOIs
Publication statusPublished - 2016

Keywords

  • Cardiac remodelling
  • Infarct
  • Laplace-beltrami
  • SVD
  • SVM

Fingerprint

Dive into the research topics of 'Supervised learning of functional maps for infarct classification'. Together they form a unique fingerprint.

Cite this