WAVELET-BASED NEURAL-NETWORK WITH FUZZY-LOGIC ADAPTIVITY FOR NUCLEAR IMAGE-RESTORATION

A novel wavelet-based neural network with fuzzy-logic adaptivity (WWNF A) is proposed for image restoration using a nuclear medicine gamma ca mera based on the measured system point spread function. The objective is to restore image degradation due to photon scattering and collimat or photon penetration with the gamma camera and allow improved quantit ative external measurements of radionuclides in vivo. The specific cli nical model proposed is tile imaging of bremsstrahlung radiation using P-32 and Y-90 because of the enhanced image degradation effects of ph oton scattering, photon penetration and poor signal-to-noise ratio (SN R) in measurements of this type with the gamma camera. The theoretical basis for four-channel multiresolution wavelet decomposition of the n uclear image into different subimages is developed with the objective of isolating the signal from noise. A fuzzy rule is generated to train a membership function rising least mean squares (LMS) to obtain as op timal balance between image restoration and the stability of the neura l network (NN), while maintaining a linear response for the camera to radioactivity dose. A multichannel modified Hopfield neural network (H NN) architecture is then proposed for multichannel image restoration u sing the dominant signal subimages. This algorithm model avoids the co mmon inverse problem associated with other image restoration filters s uch as the Wiener filter. The relative performance of the WNNFA for im age restoration is compared to a previously reported order statistic n eural network hybrid (OSNNH) filter by these investigators and a tradi tional Weiner filter and a modified HNN using simulated degraded image s with different noise levels. Quantitative metrics arch as the normal ized mean square error (NMSE) and SNR are used to compare filter perfo rmance. The WNNFA yields comparable results for image restoration with a slightly better performance for the images with higher noise levels as often encountered in bremsstrahlung detection with the gamma camer a. Experimental attenuation measurements were also performed in a wate r tank using two radionuclides, P-32 and Y-90, typically used for anti body therapy. Similar values for an effective attenuation coefficient was observed for the restored images using the OSNNH filters and WNNFA which demonstrate that the restoration filters preserves the total co unts in the image as required for quantitative in vivo measurements. T he WNNFA was computationally more efficient by a factor 4-6 compared t o the OSNNH filter. The filter architecture, in turn, is also optimum for parallel processing or VLSI implementation as required for planar mid particularly for tomographic mode of detection wing the gamma came ra. The proposed WNNFA method should also prove to be useful for quant itative imaging of single photon emitters for other nuclear medicine t omographic imaging applications using positron emitters mid direct X-r ay photon detection.