Local respiratory motion correction for PET/CT imaging: Application to lung cancer.

F. Lamare, H. Fayad, P. Fernandez, D. Visvikis
Med. Phys.. 2015-09-17; 42(10): 5903-5912
DOI: 10.1118/1.4930251

PubMed
Read on PubMed



1. Med Phys. 2015 Oct;42(10):5903-12. doi: 10.1118/1.4930251.

Local respiratory motion correction for PET/CT imaging: Application to lung
cancer.

Lamare F(1), Fayad H(2), Fernandez P(1), Visvikis D(2).

Author information:
(1)INCIA, UMR 5287, University of Bordeaux, Talence F-33400, France and Nuclear
Medicine Department, University Hospital, Bordeaux 33000, France.
(2)INSERM, UMR1101, LaTIM, Université de Bretagne Occidentale, Brest 29609,
France.

PURPOSE: Despite multiple methodologies already proposed to correct respiratory
motion in the whole PET imaging field of view (FOV), such approaches have not
found wide acceptance in clinical routine. An alternative can be the local
respiratory motion correction (LRMC) of data corresponding to a given volume of
interest (VOI: organ or tumor). Advantages of LRMC include the use of a simple
motion model, faster execution times, and organ specific motion correction. The
purpose of this study was to evaluate the performance of LMRC using various
motion models for oncology (lung lesion) applications.
METHODS: Both simulated (NURBS based 4D cardiac-torso phantom) and clinical
studies (six patients) were used in the evaluation of the proposed LRMC approach.
PET data were acquired in list-mode and synchronized with respiration. The
implemented approach consists first in defining a VOI on the reconstructed motion
average image. Gated PET images of the VOI are subsequently reconstructed using
only lines of response passing through the selected VOI and are used in
combination with a center of gravity or an affine/elastic registration algorithm
to derive the transformation maps corresponding to the respiration effects. Those
are finally integrated in the reconstruction process to produce a motion free
image over the lesion regions.
RESULTS: Although the center of gravity or affine algorithm achieved similar
performance for individual lesion motion correction, the elastic model, applied
either locally or to the whole FOV, led to an overall superior performance. The
spatial tumor location was altered by 89% and 81% for the elastic model applied
locally or to the whole FOV, respectively (compared to 44% and 39% for the center
of gravity and affine models, respectively). This resulted in similar associated
overall tumor volume changes of 84% and 80%, respectively (compared to 75% and
71% for the center of gravity and affine models, respectively). The application
of the nonrigid deformation model in LRMC led to over an order of magnitude gain
in computational efficiency of the correction relative to the application of the
deformable model to the whole FOV.
CONCLUSIONS: The results of this study support the use of LMRC as a flexible and
efficient correction approach for respiratory motion effects for single lesions
in the thoracic area.

DOI: 10.1118/1.4930251
PMID: 26429264 [Indexed for MEDLINE]

Know more about