Premium
Independent component analysis of dynamic contrast‐enhanced magnetic resonance images of breast carcinoma: A feasibility study
Author(s) -
Koh Tong San,
Thng Choon Hua,
Ho Juliana T.S.,
Tan Puay Hoon,
Rumpel Helmut,
Khoo James B.K.
Publication year - 2008
Publication title -
journal of magnetic resonance imaging
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.563
H-Index - 160
eISSN - 1522-2586
pISSN - 1053-1807
DOI - 10.1002/jmri.21391
Subject(s) - independent component analysis , dynamic contrast , voxel , contrast (vision) , breast tumor , magnetic resonance imaging , breast carcinoma , washout , dynamic contrast enhanced mri , breast mri , nuclear medicine , computer science , image resolution , medicine , breast cancer , pattern recognition (psychology) , radiology , artificial intelligence , mammography , cancer
Purpose To study the possibility of using independent component analysis (ICA) to identify breast lesions as separate hemodynamic sources on dynamic contrast‐enhanced (DCE) MR images, as depicted by the passage of contrast medium. Materials and Methods Six patients who were histopathologically confirmed with breast carcinoma underwent DCE MRI with 5 precontrast and 60 postcontrast scans at a time‐resolution of 8 s. A spatial ICA algorithm was applied on the DCE MRI data set to extract spatial component maps corresponding to source locations with different signal time–intensity patterns. To verify the present hypothesis of the ability of ICA to reveal tumor voxels as a separate hemodynamic phase, tumor margins were outlined by an experienced radiologist who was blinded from the ICA results, and the manual outlines were compared with the ICA maps. Results Consistently for each of the six patient study cases, it was found that ICA yields a tumor component map associated with typical tumor enhancement patterns of rapid enhancement with washout or plateau. Tumor outlines manually drawn by the radiologist were in good agreement with the tumor locations depicted in the tumor component maps. Conclusion ICA may provide an objective method for identifying the outlines of enhancing breast tumors on DCE MR images and to automatically extract the tumor signal intensity–time curve for subsequent tracer kinetics analysis. J. Magn. Reson. Imaging 2008. © 2008 Wiley‐Liss, Inc.