TY - GEN
T1 - Novel parametric PET image quantification using texture and shape analysis
AU - Rahmim, A.
AU - Coughlin, J.
AU - Gonzalez, M.
AU - Endres, C. J.
AU - Zhou, Y.
AU - Wong, D. F.
AU - Wahl, R. L.
AU - Sossi, V.
AU - Pomper, M. G.
PY - 2012
Y1 - 2012
N2 - Kinetic parameter estimation at the individual voxel level has the powerful ability to represent both spatial distributions and quantitative physiological parameters of interest. In practice, parametric images are commonly quantified by computing mean values for specified ROIs. Nonetheless, the mean operator vastly oversimplifies the available spatial uptake information. It may be hypothesized that a given tracer will exhibit increasingly differential or heterogeneous uptake due to disease: subsequently, we have implemented and explored a comprehensive texture and shape analysis framework wherein extensive information is generated from parametric PET images, through: (1) 3D moment invariants analysis, (2) intensity histogram analysis, (3) gray-level spatial-dependence (GLSD) analysis, and (4) neighborhood gray tone difference (NGTD) analysis. In the present work, we applied this approach to imaging of 11C-DPA-713, a novel PET ligand with high binding to the translocator protein (TSPO), a marker of neuroinflammation. In particular, for tracers such as DPA with relatively wide-spread uptake, where a reliable reference tissue does not exist, quantification of heterogeneity may provide additional valuable information. Our center has been the first to perform DPA PET studies in humans, and is gathering a large collection of PET studies; e.g. subjects with systemic lupus erythematosus (SLE), traumatic brain injury (TBI; former NFL players), along with young and elderly controls. Our preliminary analysis has revealed that, compared to conventional mean ROI analysis, a number of metrics in the proposed framework yield enhanced discrimination (as measured using AUC) between patients vs. controls in a range of ROIs, in TBI as well as SLE vs. controls, consistent with increased heterogeneity of uptake with disease, though in the case of SLE this can be, at least partly, attributed to changes in ROI shapes. Overall, the proposed framework has the potential to bring about a new quantification paradigm in parametric imaging.
AB - Kinetic parameter estimation at the individual voxel level has the powerful ability to represent both spatial distributions and quantitative physiological parameters of interest. In practice, parametric images are commonly quantified by computing mean values for specified ROIs. Nonetheless, the mean operator vastly oversimplifies the available spatial uptake information. It may be hypothesized that a given tracer will exhibit increasingly differential or heterogeneous uptake due to disease: subsequently, we have implemented and explored a comprehensive texture and shape analysis framework wherein extensive information is generated from parametric PET images, through: (1) 3D moment invariants analysis, (2) intensity histogram analysis, (3) gray-level spatial-dependence (GLSD) analysis, and (4) neighborhood gray tone difference (NGTD) analysis. In the present work, we applied this approach to imaging of 11C-DPA-713, a novel PET ligand with high binding to the translocator protein (TSPO), a marker of neuroinflammation. In particular, for tracers such as DPA with relatively wide-spread uptake, where a reliable reference tissue does not exist, quantification of heterogeneity may provide additional valuable information. Our center has been the first to perform DPA PET studies in humans, and is gathering a large collection of PET studies; e.g. subjects with systemic lupus erythematosus (SLE), traumatic brain injury (TBI; former NFL players), along with young and elderly controls. Our preliminary analysis has revealed that, compared to conventional mean ROI analysis, a number of metrics in the proposed framework yield enhanced discrimination (as measured using AUC) between patients vs. controls in a range of ROIs, in TBI as well as SLE vs. controls, consistent with increased heterogeneity of uptake with disease, though in the case of SLE this can be, at least partly, attributed to changes in ROI shapes. Overall, the proposed framework has the potential to bring about a new quantification paradigm in parametric imaging.
UR - http://www.scopus.com/inward/record.url?scp=84881570870&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84881570870&partnerID=8YFLogxK
U2 - 10.1109/NSSMIC.2012.6551507
DO - 10.1109/NSSMIC.2012.6551507
M3 - Conference contribution
AN - SCOPUS:84881570870
SN - 9781467320306
T3 - IEEE Nuclear Science Symposium Conference Record
SP - 2227
EP - 2230
BT - 2012 IEEE Nuclear Science Symposium and Medical Imaging Conference Record, NSS/MIC 2012
T2 - 2012 IEEE Nuclear Science Symposium and Medical Imaging Conference Record, NSS/MIC 2012
Y2 - 29 October 2012 through 3 November 2012
ER -