Artificially Generated Training Datasets for Supervised Machine Learning Techniques in Magnetic Resonance Imaging: An Example in Myocardial Segmentation

Machine learning techniques have become increasingly successful over the last few years in medical image analysis and radiology. However, the low availability, relativeness and size of the training data sets required by the associated learning algorithms prevents their further development or delays...

Full description

Saved in:
Bibliographic Details
Published in2019 Computing in Cardiology (CinC) pp. Page 1 - Page 2
Main Authors Xanthis, Christos G, Haris, Kostas, Filos, Dimitrios, Aletras, Anthony H
Format Conference Proceeding
LanguageEnglish
Published Creative Commons 01.09.2019
Subjects
Image analysis
Image segmentation
Machine learning
Magnetic resonance imaging
Supervised learning
Training
Training data
Online AccessGet full text

Cover

More Information
Summary:Machine learning techniques have become increasingly successful over the last few years in medical image analysis and radiology. However, the low availability, relativeness and size of the training data sets required by the associated learning algorithms prevents their further development or delays their application in clinical practice.This study presented for the first time the development of artificially generated training datasets for supervised learning techniques through the incorporation of a realistic simulation framework in the field of Magnetic Resonance Imaging (MRI). An example in left-ventricle segmentation was utilized and the performance of a fully convolutional network on true cardiac MR data was evaluated.
ISSN:2325-887X
DOI:10.22489/CinC.2019.220