Meeting Banner
Abstract #4201

An Automatic Parenchyma Extraction Method for MRI R2* Relaxoemtry of Iron Loaded Liver

Meiyan Feng1, Huashuai Gao1, Xinyuan Zhang1, Yanqiu Feng1, Wufan Chen1, Xuegang Xin1, Taigang He2, 3

1School of Biomedical Engineering, Southern Medical University, Guangzhou, China; 2Cardiovascular Biomedical Research Unit, Royal Brompton Hospital, London, United Kingdom; 3National Heart and Lung Institute, Imperial College, London, United Kingdom

The whole liver ROI method for R2* measurement has been shown better reproducibility than the routinely used mROI method, but still suffers from noise, partial volume effect and subjective segmentation of liver parenchyma and vessels. We proposed an automatic parenchyma extraction (APE) method of R2* measurement, the measurement accuracy of which was evaluated in both simulation and patient studies. The mean R2* evaluation error percentage and the coefficient of variation (CoV) of inter-observer reproducibility for the APE method severally were 0.34% and 1.39%. The proposed APE method may be important for increasing the diagnostic confidence of R2* measurement.

Abstract PDF Presentation Video

Keywords

account accuracy adopted aimed artifacts assess assessing audience automatic better biomedical bland cardiovascular china clinical clinicians coefficient college concentration confidence considered consisted corrected correlation defined delineated dependence dependent deviation devoid diagnostic diminish distributed divided dramatically drawn effective engineering engineers error errors evaluate evaluated evaluation example exponential expressed extract extraction filter fitting five function fuzzy gradient heart hepatic heterogeneity hospital impact imperial important increasing independent indicated institute intensity inter iron kingdom liver loaded local lung males manual manually mathematical matrix measured medical morphological national noise observer operator output partial patients percentage performance pixel pixels placement plot plots practice prognosis propose proposed protocols quantification ranging reduced reliable repetition reproducibility robust royal sample sampling scanned scanner scatter school segmentation simulation simulations slice sonata southern statistical statistically steps still subjective suffer target tissues took traditionally transfused transfusion true unit variability variation varying vessels visible volume whole wise years