WAVELET-TRANSFORM OPTIMIZATION METHODS IN THE FILTERING OF FUNCTIONALMR-IMAGING DATA SETS (VOL 201(P), PG 259, 1996)

PURPOSE: To compare optimization methods for wavelet filtering of func tional MR imaging data sets. MATERIALS AND METHODS: Four different wav elet transform (WT) filtering optimization methods were applied to 4 B OLD functional MR imaging SPGR data sets prior to the generation of cr oss-correlation functional maps: (1) Amplitude thresholding of the WT was optimized by using an iterative algorithm by maximizing the signal -output paradigm Pearson correlation. (2) The lowest resolution WT com ponent of the background was discarded, and the remaining maximum was used as an amplitude threshold. (3) The percentage compression from an iterative Pearson correlation optimization algorithm was used. (4) Wa velet phase plane tree-sorted decomposition and reconstruction was opt imized by using an autocorrelation technique. Activation volume was as sessed for the entire imaging section and the sensorimotor cortex. RES ULTS: The Ist method (amplitude thresholding) resulted in loss of acti vation volume from the unfiltered state in 2 cases. The percentage com pression and background thresholding methods (2 and 3) resulted in com parable increases in sensorimotor cortex activation. Phase plane decom position/reconstruction resulted in diffusely distributed activation i ncrease. CONCLUSION: Background amplitude thresholding and iterative o ptimization of the WT by using percentage compression are effective me thods for filtration of functional MR imaging data sets. Potential pit falls in WT filtering include excessive signal damping from amplitude thresholds. TAKE HOME POINTS: Percentage compression WT optimization i s useful in denoising functional MR imaging data sets. Amplitude thres holds alone are a potential pitfall in wavelet WT functional MR imagin g filtration.