Generation of brain pseudo-CTs using an undersampled, single-acquisition UTE-mDixon pulse sequence and unsupervised clustering

摘要: Purpose: MR-based pseudo-CT has an important role in radiation therapy planning and PET attenuation correction. The purpose of this study is to establish a clinically feasible approach, including image acquisition, correction, CT formation, for generation the brain using single-acquisition, undersampled ultrashort echo time (UTE)-mDixon pulse sequence. Methods: Nine patients were recruited study. For each patient, 190-s, undersampled, single acquisition UTE-mDixon sequence was acquired (TE = 0.1, 1.5, 2.8 ms). A novel method retrospective trajectory correction free induction decay (FID) signal performed based on point-spread functions three external MR markers. Two-point Dixon images reconstructed first second data 1.5 R2∗ (1/T2∗) then estimated used provide bone information. Three features, i.e., Dixon-fat, Dixon-water, R2∗, unsupervised clustering. Five tissue clusters, air, brain, fat, fluid, bone, fuzzy c-means (FCM) algorithm. two-step, automatic tissue-assignment approach proposed designed according prior information given feature space. Pseudo-CTs generated by voxelwise linear combination membership FCM. low-dose patient as gold standard comparison. Results: contrast sharpness FID improved after applied. mean delay 0.774 μs (max: 1.350 μs; min: 0.180 μs). FCM-estimated centroids different types showed distinguishable pattern tissues, significant differences found between centroid locations types. Pseudo-CT can additional skull detail low bias absolute error numbers voxels (−22 ± 29 HU 130 16 HU) when compared CT. Conclusions: features processing methods may representative clustering could thus be clinical generation.