Meeting Banner
Abstract #2973

A Spectral Quality Control Algorithm for 3D 1H MRSI Data of the Prostate

Alan Wright1, Kobus Thiele1, Kirsten M. Selns2, Ingrid S. Gribbestad2, Elisabeth Weiland3, Tom W. J. Scheenen1, Arend Heerschap1

1Radiology, Radboud University Nijmegen Medical Centre, Nijmegen, Gelderland, Netherlands; 2Department of Circulation and Medical Imaging, NTNU, Trondheim, Norway; 3Siemens Healthcare, Erlangen, Germany

An automated quality control algorithm has been developed for 3D 1H MRSI data sets from prostate cancer patients. This will provide an important post processing tool in automation of the analysis of MRSI data of these patients. The method is based on feature extraction using Independent Component Analysis and uses a Gold Standard of consensus decisions of spectral quality as judged by four experts. The algorithm can separate a Test Set of acceptable from unacceptable data with 95% sensitivity and 95% specificity.

Abstract PDF E-Poster Video

Keywords

acceptability acceptable according account accuracy accurate accurately adding agreed agreement amongst applied array assessment automated automatic best bottleneck bounds brain cancer characteristic chemical choline circulation citrate classifications classified classifier clinical coefficient coefficients coils combined component components concentrations consistent contain control cram curve decisions defined detection determine developed dimensional disease divided either entire error errors excellent expected experienced expert expertise experts features fitting fittings frequency general generate generated give gives gold good graded highest highlighted hyper independent individual inferred input inter judged localization machine major medical metabolites model necessarily noise operator optimization pairs patients peaks poorer practice preventing previously procedure produced prostate quality radiology readers receiver reduced replacing reproduces require required reveal routine satisfactory scale scores sensitivity separated separation sets shape significantly specificity spectra spectral spectroscopic spectrum staging step subsequent subset sufficient support surface tissue training transference trio tune unacceptability unacceptable useful variations vector whether width wright