Study On Spectral CT Imaging And Its Application In Proton Therapy

Computed tomography(CT)imaging is one of the most important imaging technique in clinical procedure.As the new generation of CT imaging technique,spectral CT has become more and more popular for diagnosis and radiotherapy applications.In the recent years,CT imaging manufacturers have released four types of spectral CT system:single-source dual-kVp CT,single-source fast kVp-switching CT,dual-source CT and dual-layer detector CT.As these machines can provide dual energy image data,they are also called dual-energy CT(DECT).Nowadays,DECT,or spectral CT,has shown its promising and irreplaceable advantages over conventional single-energy CT in clinical diagnosis and radiotherapy procedures.The high system cost and technical barriers of spectral CT,however,obstruct the step of this CT technique into clinical practice.In addition,the advantages of spectral CT haven’t been thoroughly validated in real clinical environment.How could spectral CT imaging become popular and fully used in clinical diagnosis and treatment?One of the key points is to realize spectral CT technique in a cost-effective manner and to reconstruct high-quality spectral images.Another point is to fully utilize spectral imaging to obtain more infomations such as human tissues’ electron density and elemental compositions.The third key point is to comprehensively evaluate the advantages of spectral CT over conventional CT in real clinical procedure.These problems have also been hot topics in both academic and clinical research.In this paper,we propose to reconstruct high-quality spectral images on a conventional clinical CT platform and to evaluate range uncertainty in proton therapy by using current spectral CT system.Our first goal is to acquire multi-energy CT data on a commonly-used single-energy CT system and to reconstruct high-quality spectral images,for the purposes of accurate diagnosis and treatment.We first proposed to utilize kVp-switching scanning protocol to acquire multi-energy projections simultaneously on a conventional CT platform.This would allow us to achieve spectral CT imaging utilizing current clinical situation and technique at low cost.Then we developed a spatial spectral non-local means(ssNLM)reconstruction model and solved it by using a three-steps iterative algorithm.This algorithm is based on the popular non-local means filter method,to utilize the similarity of image patches in spatial domain and to reduce noise effectively.Meanwhile,our proposed algorithm considers structural correlation between different spectral images and the regularize them via this priori constraint.The regularization in both spatial and spectral domains would suppress the under-sampling noise and artifacts while preserving fine structures.We have validated our proposed method by a series of simulation and experimental studies.Of which,we tested our method on a Varian TrueBeam On-Board-Imaging cone-beam CT platform,with three commonly used phantoms.These studies showed that our ssNLM algorithm can reduce noise and streak artifacts effectively and recover fine structures,thus achieve high-quality reconstructed spectral images.Then we proposed a multi-energy element resolved(MEER)CBCT framework.It employs energy-resolved data acquisition on a conventional CBCT platform and then simultaneously reconstructs images of x-ray attenuation coefficient,electron density relative to water(rED)and elemental composition(EC).We realized the MEER-CBCT framework on a Varian TrueBeam CBCT platform using a kVp-switching scanning scheme.A simultaneous image reconstruction and elemental decomposition model was formulated as an optimization problem.The objective function used a least-square term to enforce fidelity between x-ray attenuation coefficients and projection measurements.Spatial regularization was introduced via sparsity under a tight wavelet-frame transform.Consistency between rED,EC and attenuation coefficient was imposed,which inherently served as a regularization term along the energy direction.The optimization problem was solved by a novel alternating-direction minimization scheme.Simulation and experiment studies demonstrated our proposed method’s capability to simultaneously reconstruct x-ray attenuation coefficient,rED,and EC images accurately.Spectral CT machine is not a common device in most radiation oncology centers.One of the main reason is that its advantages over traditional CT haven’t been fully justified in the photon and proton(or heavy ion)treatment procedure.In this study,we did a comprehensive analysis of proton range uncertainties related to stopping-power-ratio(SPR)estimation using spectral CT imaging.The difference between our study and some previous studies is that,we considered a lot of impact factors which may cause SPR estimation errors,not just the inherent model error as other studies did.Our study showed that spectral CT number varied in different scanning conditions(e.g.patient body size or disease sites),these variations would be amplified during the SPR estimation process and then caused proton range uncertainty.However,spectral CT still yield more accuracy and precision in the proton range estimation than that of conventional single-energy CT.These results indicate that spectral CT will achieve more accurate proton therapy than current technique.In addition,we recommend that other centers estimate the range uncertainty of proton therapy in their own systems,although our analysis of the overall uncertainty in SPR estimation using the spectral CT approach may serve as a general guideline.