Study Of Application Of Channelized Hotelling Observer In Pet System Image Quality Evaluation

Positron Emission Tomography(PET)is an advanced clinical examination imaging technology in nuclear medicine field.It can accurately reflect physiological and biochemical changes inside the body.After the 21 st century,PET is in rapid development.At present,it has been the very important means of examination in early diagnosis of cancer and brain function imaging.PET image quality assessment is a very important aspect of PET imaging,but also a key step in optimizing PET imaging system parameters.Human Observer(HO)is currently the main means to assess image quality,but it does not suitable for the study of system parameter optimization.Because doctors need to spend a lot of time and effort,and in order to get objective results,multiple doctors are required.Numerical Observer(NO)can overcome these shortcomings very well and is increasingly used in image quality assessment.Among many numerical observers,considering the universality of the applications,Channelized Hotelling Observer(CHO)is chosen as the method of PET image quality assessment in this dissertation.The first part of the dissertation introduced the commonly used image quality assessment methods include the numerical observer CHO used primaryly in this dissertation and it's research situation of now? imaging principle of the PET system and image reconstruction process?as well as the factors that affect the quality of reconstructed image;then introduces the GATE(Geant4 Application for Tomographic Emission)simulation method which is the way data of this dissertation get from.The second part of the dissertation uses CHO to evaluate the effect of PET system parameters on image quality.The factors affecting the PET image quality include hardware parameters and software parameters.The hardware parameters are mainly the crystal thickness and axial field of view(AFOV).The software parameters are mainly the filters and the number of iterations.Effects of the hardware parameters crystal thickness,axial field of view(AFOV)on image quality: crystal thickness and axial field of view(AFOV)are two major factors in hardware.The increase of thickness and AFOV can increase system sensitivity and improve image quality,but the cost will also increase,scan speed will drop.After considering the sensitivity,scanning speed and cost,the axial field of view is set to 201.6mm and 403.2mm respectively,and the crystal thickness is set to 15 mm and 17.5mm respectively.Using CHO to evaluate their effect on image quality,the results showed that system with an axial field of view of 403.2 mm produces the more clear PET images than the system with an axial field of view of 201.6 mm when the crystal thickness is the same.A system with a crystal thickness of 17.5 mm produces more clear PET images than a system with a crystal thickness of 15 mm when the axial field of view is the same.Effects of the software parameters filter,the number of iterations on image quality: The number of filters and iterations are two software parameters studied most frequently.Their values affect the quality of the image very often.The quality of the image usually gets better then gets worse with the increase of the value of its parameters.So it is very important to determine their optimal parameters in order to get the best image quality.CHO can quantify the image quality to select the best parameters.Firstly,the best scanning time is selected by considering the image quality,simulation time and so on.Second,the filter used in this dissertation is a Gaussian filter and a non-local mean(NLM)filter.And the Gaussian parameter ? is set to 0.5,0.8,1.1,1.4,2.0,3.0 respectively.NLM parameter f is set to 0.5,0.7,0.9,1.2,2.0,3.0,4.0 respectively.They were optimized by CHO and the filtering effect of the two filters under the optimal parameters was evaluated.The results showed that ? = 1.4 and f = 0.5 are the best parameters of Gaussian filter and NLM filter.And the filtering effect of NLM filter is better than Gaussian filter.Finally,number of iterations is set to 1,2,3,4,6,8 and 10 respectively.When the number of iterations becomes larger,the image becomes clear,but because the noise in the image is amplified,the image is blurred due to noise effects when the number of iterations is lager than a certain value.In order to get the optimal one,CHO is used to evaluate the images of number of iterations metioned before.The result showed that the image quality is the best when the number of iterations is 4.The simulation results showed that CHO can be used to evaluate the effect of PET system parameters on image quality,which is important for PET system development and auxiliary diagnosis of PET images.The innovation of this paper is that the researchers can quantify the parameters by calculating the AUC value by CHO,and this is the first study to evaluate the influence of PET system parameters on image quality by CHO method.