Open AccessAutomatic and manual image fusion of111In-pentetreotide SPECT and diagnostic CT in neuroendocrine tumor imaging - An evaluation
Author(s) -
Elisabeth Hedlund,
JanErik Karlsson,
Sven-Åke Starck
Publication year - 2010
Publication title -
journal of medical physics/journal of medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.292
H-Index - 24
eISSN - 1998-3913
pISSN - 0971-6203
DOI - 10.4103/0971-6203.71766
Subject(s) - image fusion , nuclear medicine , medicine , radiology , image registration , neuroendocrine tumors , single photon emission computed tomography , scintigraphy , tomography , artificial intelligence , computer science , image (mathematics) , pathology
In the clinical diagnosis of neuroendocrine tumors (NET), the results of examinations, such as high-resolution computed tomography (CT) and single photon computerized tomography (SPECT), have conventionally been interpreted separately. The aim of the present study was to evaluate Hermes Multimodality™ 5.0 H Image Fusion software-based automatic and manual image fusion of SPECT and CT for the localization of NET lesions. Out of 34 NET patients who were examined by means of somatostatin receptor scintigraphy (SRS) with 111In- pentetreotide along with SPECT, 22 patients had a CT examination of the abdomen, which was used in the fusion analysis. SPECT and CT data were fused using software with a registration algorithm based on normalized mutual information. The criteria for acceptable fusion were established at a maximum cranial or caudal dislocation of 25 mm between the images and at a reasonable consensus (in order of less than 1 cm) between outline of the reference organs. The automatic fusion was acceptable in 13 of the 22 examinations, whereas 9 fusions were not. However all the 22 examinations were acceptable at the manual fusion. The result of automatic fusion was better when the slice thickness of 5 mm was applied at CT examination, when the number of slices was below 100 in CT data and when both examinations included uptakes of pathological lesions. Retrospective manual image fusion of SPECT and CT is a relatively inexpensive but reliable method to be used in NET imaging. Automatic image fusion with specified software of SPECT and CT acts better when the number of CT slices is reduced to the SPECT volume and when corresponding pathological lesions appear at both SPECT and CT examinations.