טכניון מכון טכנולוגי לישראל
הטכניון מכון טכנולוגי לישראל - בית הספר ללימודי מוסמכים  
Ph.D Thesis
Ph.D StudentKennedy John Anthony
SubjectAn Investigation of Methods for Data and Image Fusion in a
Hybrid PET/CT Imaging System
DepartmentDepartment of Biomedical Engineering
Supervisors Professor Haim Azhari
Professor Emeritus Ora Israel
Full Thesis textFull thesis text - English Version


Abstract

Dual modality PET/CT (positron emission tomography/computed tomography) scanners were introduced commercially in 2001 in order to provide diagnostic information on function from the PET and anatomy from the CT. By overlaying co-registered PET and CT data, clinicians have a powerful noninvasive imaging tool for diagnosis and localization of disease.

PET images are noisy and have low resolution, compared to CT. In addition, CT-based attenuation correction of PET images (CT-AC PET) can erroneously create artifacts in the vicinity of metal implants. The purpose of this thesis was to develop improved methods of PET and CT data fusion in order to enhance the PET image quality, thereby providing increased small lesion detectability and localization, as well as more accurate tracer uptake quantitation.

First, a “super-resolution” method was used to improve the resolution of a routinely used clinical PET/CT scanner (Discovery LS, GE Healthcare Technologies).  Second, PET data were smoothed with a hybrid computed tomography algorithm (HCT) in which anatomical CT information was employed to retain sharper edges. Third, to reduce artifacts near metal implants in CT-AC PET images, an artifact reduction algorithm (ARA) was developed in order to partially correct the attenuation maps. These techniques were validated in phantom and patient PET/CT studies.

In 18F-FDG (18F-fluorodeoxyglucose) PET/CT phantom studies employing either the super-resolution or HCT method, smaller lesions could be resolved as compared with standard methods (≤ 3 mm vs. ≥ 4 mm in diameter). Compared to standard processing, the joint application of both methods increased average contrast ratios in a phantom trial by 54% (range: 45─69 %), superior to 13% (range: 9─15%) for super-resolution alone or 14% (range: 7─18%) for HCT alone. In regions distant from anatomical borders, the HCT algorithm is shown to be equivalent to the Gaussian filtering used in standard reconstructions. In patient images, the combined super-resolution/HCT method allowed for an improved delineation of a pulmonary lesion. In phantom trials, near a metal hip joint, ARA identified two targets absent on routine PET images. In a patient study, artifacts were of lower intensity in ARA-PET as compared to standard images.

In summary, a newly developed reconstruction method incorporating HCT and super-resolution for fusing PET and CT images has been developed, and proven to provide higher resolution metabolic images. In addition, an algorithm for reducing PET artifacts in CT-AC PET has been developed and may potentially improve the detectability of small FDG-avid lesions in the vicinity of metal implants.