The Study On Dedicated Breast PET System And Image Reconstruction Algorithms

The breast cancer ranks first of malignancy tumor in women, the early diagnosis andtreatment can greatly improve the cure rate. Because of the high sensation and precision，Positron Emission Tomography (PET) has became an irreplaceable role in the clinicalapplications. This paper presented a kind of partial ring dedicated breast PET scanner system. Itreduced the crystals number, the cost of hardware and data storage. It could provide geometricflexibility in detector placement as well as combine with other modalities, like X-ray, CT, MRIand radiation therapy of breast tumor. The dedicated breast scanner system has high spatialresolution and sensitivity. But the partial ring PET system will inevitably lead to drop in signalsampling, the lack of the detection angle and the deterioration of the image quality, and wasdifficult to achieve good results of the detection. In this paper, we established a sparseregularization image reconstruction model using time of flight information and compressedsensing theory.This paper first introduced the research status, the principle theory of PET imaging, andsummarized the PET image reconstruction algorithms. On the basis of previous studies, thepaper designed five different structure breast PET systems, full ring scanner,2/3(120in-planeangular coverage) curved scanner,1/2(90in-plane angular coverage) curved scanner,2/3rotation curved scanner and1/2rotation curved scanner. The diameter of the full ring scanner as15cm and axial length was15.6cm. The simulation were performed to analyses the absolutesensitivity, noise equivalent count rate (NECR), scatter fraction (SF) and special resolution ofthese scanners. The simulation results showed that with the decrease of coverage angle thesensitivity and noise equivalent count rate of the scanner gradually reduced, the spatial resolutiondeclined in the edge, particularly in the gaps, while the performance of the rotation system wasbetter than no rotation system in partial ring.Secondly the paper studied the image reconstruction algorithms under the conditions ofundersampling. Based on the compressed sensing theory, the paper used the Total Variation (TV)to sparse the image and used it as regularization, then solved the objective function using thealternate methods. This method performed better in noise suppression, keeping the edge not beendestroyed and improving the reconstruction image quality.Finally, the paper further studied the reconstruction algorithms with TOF and combined sparse regularization. By using the wavelet and TV sparse regularization and introducing theTOF information to further improved the reconstruction images. The paper then studied theinfluences of time resolution and time bin on image reconstruction quality. The experimentalresults showed that compared with the traditional methods the combined bases sparseregularization could effectively suppress noise and improve the image quality, and the highertime resolution and short time bin could reduce better reconstruction quality. In breast phantomreconstruction, data acquisition in full ring system and rotation system was well-distributed andthe reconstruction quality was better. The TOF information could further improve the quality ofimage reconstruction. The rotation partial ring system could reach the performance of full ringsystem after introducing TOF.In this paper, the designed PET system reduced the hardware cost, enhanced the PETintegration with other systems and improved the quality of image reconstruction. This dedicatedsystem and image reconstruction algorithms have a good prospect.