Automated Principal Component Analysis (PCA) Filtering for Denoising DCE-MRI Data

Balvay Daniel¹, Nadjia Kachenoura², Isabelle Thomassin¹, Rokhaya Thiam¹, Laure Fournier^1,3, Yves Rozenholc⁴, Charles Andr Cuenod^1,3

Dynamic Contrast Enhanced MRI is impaired, for small or heterogeneous lesion, by low signal to noise ratio (SNR), providing poor drawing of region of interest, of noisy microcirculatory parametric maps. To improve SNR without limiting the spatial or the time resolution by low pass filter, a Principal Component Analysis time filtering (PCA-TF) was performed. The number of PCA-TF factors was automatically identified by using the Fraction of Residual Information. The tests on DCE-MRI series showed an obvious denoising efficiency of the adapted PCA-TF with a minimal loss of information.