Fuzzy logic-based moving average filters for reducing noise from Tc-99m-sestamibi parathyroid images
Anil Kumar Pandey1, Navneet Kumar, Shweta Dhiman
1Department of Nuclear Medicine, All India Institute of Medical Sciences, New Delhi, India.
Introduction:
The objective of the study was to use fuzzy logic-based moving average filters for reducing noise from Tc-99m-sestamibi parathyroid images and to compare its performance with classical moving average filters.
Methods:
Sixty-eight Tc-99m-sestamibi parathyroid images (33 image zoom 1.0, 35 images zoom 2.0) were filtered using symmetric triangular fuzzy filters with the moving average (TMAV), asymmetric triangular fuzzy filters with the moving average (ATMAV) and classical moving average filter (MAV) with moving average within a square window of dimension N × N pixels (N=3,5,7,9,11). The relative filtering performance was compared both objectively (using Brisque score) and subjectively [by two nuclear medicine physicians on a 4-point scale (1 = nondiagnostic; 2 = diagnostic; 3 = good; and 4 = excellent image quality)]. The nonparametric two-sample Kolmogorov-Smirnov test was applied to find the statistically significant difference between the quality of input and their corresponding filtered images.
Results:
The Brisque score assigned to MAV filtered zoom 2.0 images (MAV_3, median = -0.61) were significantly smaller than that of their input images (median = 53.84, at P = 1) and fuzzy filtered images (TMAV_3, median = 0.44, at P = 0.89 and ATMAV_3, median =8.26, at P = 0.97). The sum of average subjective image quality score for input, MAV_3, TMAV_3, TMAV_5, ATMAV_3, and ATMAV_5 were 148, 221, 221.5, 198,171,253 and 237.5, respectively.
Conclusion:
On the basis of subjective assessment, the performance of ATMAV_3 fuzzy filter was found to be better compared to the classical moving average filter in reducing noise from Tc-99m-sestambi parathyroid images.
