Automated Quality Control of Prostate Cancer MRSI Using Independent Component Analysis

Alan James Wright¹, Thiele Kobus¹, Thomas Hambrock¹, Tom W. Scheenen¹, Arend Heerschap¹

Magnetic Resonance Spectroscopic Imaging (MRSI) of prostate cancer patients can provide information on the detection and localization of prostate cancers. Automatic processing of MRSI data requires an automated quality control step. We present a method for quality control of 3T MRSI data from prostate cancer patients that separates raw spectral data as voxels of acceptable and unacceptable quality. This is done with a feature extraction method based on independent component analysis. The separation achieved is comparable to the gold standard of expert decision.