A priori motion models for four-dimensional reconstruction in gated cardiac SPECT

David S. Lalush, Lin Cui, Benjamin Tsui

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

We investigate the benefit of incorporating a priori assumptions about cardiac motion in a fully four-dimensional (4D) reconstruction algorithm for gated cardiac SPECT. Previous work has shown that non-motion-specific 4D Gibbs priors enforcing smoothing in time and space can control noise while preserving resolution. In this paper, we evaluate methods for incorporating known heart motion in the Gibbs prior model. The new model is derived by assigning motion vectors to each 4D voxel, defining the movement of that volume of activity into the neighboring time frames. Weights for the Gibbs cliques are computed based on these `most likely' motion vectors. To evaluate, we employ the mathematical cardiac-torso (MCAT) phantom with a new dynamic heart model that simulates the beating and twisting motion of the heart. Sixteen realistically-simulated gated datasets were generated, with noise simulated to emulate a real Tl-201 gated SPECT study. Reconstructions were performed using several different reconstruction algorithms, all modeling nonuniform attenuation and three-dimensional detector response. These include ML-EM with 4D filtering, 4D MAP-EM without prior motion assumption, and 4D MAP-EM with prior motion assumptions. The prior motion assumptions included both the correct motion model and incorrect models. Results show that reconstructions using the 4D prior model can smooth noise and preserve time-domain resolution more effectively than 4D linear filters. We conclude that modeling of motion in 4D reconstruction algorithms can be a powerful tool for smoothing noise and preserving temporal resolution in gated cardiac studies.

Original languageEnglish (US)
Title of host publicationIEEE Nuclear Science Symposium & Medical Imaging Conference
PublisherIEEE
Pages1923-1927
Number of pages5
Volume3
StatePublished - 1996
Externally publishedYes
EventProceedings of the 1996 IEEE Nuclear Science Symposium. Part 1 (of 3) - Anaheim, CA, USA
Duration: Nov 2 1996Nov 9 1996

Other

OtherProceedings of the 1996 IEEE Nuclear Science Symposium. Part 1 (of 3)
CityAnaheim, CA, USA
Period11/2/9611/9/96

Fingerprint

Acoustic variables control
Detectors

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Lalush, D. S., Cui, L., & Tsui, B. (1996). A priori motion models for four-dimensional reconstruction in gated cardiac SPECT. In IEEE Nuclear Science Symposium & Medical Imaging Conference (Vol. 3, pp. 1923-1927). IEEE.

A priori motion models for four-dimensional reconstruction in gated cardiac SPECT. / Lalush, David S.; Cui, Lin; Tsui, Benjamin.

IEEE Nuclear Science Symposium & Medical Imaging Conference. Vol. 3 IEEE, 1996. p. 1923-1927.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Lalush, DS, Cui, L & Tsui, B 1996, A priori motion models for four-dimensional reconstruction in gated cardiac SPECT. in IEEE Nuclear Science Symposium & Medical Imaging Conference. vol. 3, IEEE, pp. 1923-1927, Proceedings of the 1996 IEEE Nuclear Science Symposium. Part 1 (of 3), Anaheim, CA, USA, 11/2/96.
Lalush DS, Cui L, Tsui B. A priori motion models for four-dimensional reconstruction in gated cardiac SPECT. In IEEE Nuclear Science Symposium & Medical Imaging Conference. Vol. 3. IEEE. 1996. p. 1923-1927
Lalush, David S. ; Cui, Lin ; Tsui, Benjamin. / A priori motion models for four-dimensional reconstruction in gated cardiac SPECT. IEEE Nuclear Science Symposium & Medical Imaging Conference. Vol. 3 IEEE, 1996. pp. 1923-1927
@inproceedings{6c95ea1fc48f47afbaabf9da2e862187,
title = "A priori motion models for four-dimensional reconstruction in gated cardiac SPECT",
abstract = "We investigate the benefit of incorporating a priori assumptions about cardiac motion in a fully four-dimensional (4D) reconstruction algorithm for gated cardiac SPECT. Previous work has shown that non-motion-specific 4D Gibbs priors enforcing smoothing in time and space can control noise while preserving resolution. In this paper, we evaluate methods for incorporating known heart motion in the Gibbs prior model. The new model is derived by assigning motion vectors to each 4D voxel, defining the movement of that volume of activity into the neighboring time frames. Weights for the Gibbs cliques are computed based on these `most likely' motion vectors. To evaluate, we employ the mathematical cardiac-torso (MCAT) phantom with a new dynamic heart model that simulates the beating and twisting motion of the heart. Sixteen realistically-simulated gated datasets were generated, with noise simulated to emulate a real Tl-201 gated SPECT study. Reconstructions were performed using several different reconstruction algorithms, all modeling nonuniform attenuation and three-dimensional detector response. These include ML-EM with 4D filtering, 4D MAP-EM without prior motion assumption, and 4D MAP-EM with prior motion assumptions. The prior motion assumptions included both the correct motion model and incorrect models. Results show that reconstructions using the 4D prior model can smooth noise and preserve time-domain resolution more effectively than 4D linear filters. We conclude that modeling of motion in 4D reconstruction algorithms can be a powerful tool for smoothing noise and preserving temporal resolution in gated cardiac studies.",
author = "Lalush, {David S.} and Lin Cui and Benjamin Tsui",
year = "1996",
language = "English (US)",
volume = "3",
pages = "1923--1927",
booktitle = "IEEE Nuclear Science Symposium & Medical Imaging Conference",
publisher = "IEEE",

}

TY - GEN

T1 - A priori motion models for four-dimensional reconstruction in gated cardiac SPECT

AU - Lalush, David S.

AU - Cui, Lin

AU - Tsui, Benjamin

PY - 1996

Y1 - 1996

N2 - We investigate the benefit of incorporating a priori assumptions about cardiac motion in a fully four-dimensional (4D) reconstruction algorithm for gated cardiac SPECT. Previous work has shown that non-motion-specific 4D Gibbs priors enforcing smoothing in time and space can control noise while preserving resolution. In this paper, we evaluate methods for incorporating known heart motion in the Gibbs prior model. The new model is derived by assigning motion vectors to each 4D voxel, defining the movement of that volume of activity into the neighboring time frames. Weights for the Gibbs cliques are computed based on these `most likely' motion vectors. To evaluate, we employ the mathematical cardiac-torso (MCAT) phantom with a new dynamic heart model that simulates the beating and twisting motion of the heart. Sixteen realistically-simulated gated datasets were generated, with noise simulated to emulate a real Tl-201 gated SPECT study. Reconstructions were performed using several different reconstruction algorithms, all modeling nonuniform attenuation and three-dimensional detector response. These include ML-EM with 4D filtering, 4D MAP-EM without prior motion assumption, and 4D MAP-EM with prior motion assumptions. The prior motion assumptions included both the correct motion model and incorrect models. Results show that reconstructions using the 4D prior model can smooth noise and preserve time-domain resolution more effectively than 4D linear filters. We conclude that modeling of motion in 4D reconstruction algorithms can be a powerful tool for smoothing noise and preserving temporal resolution in gated cardiac studies.

AB - We investigate the benefit of incorporating a priori assumptions about cardiac motion in a fully four-dimensional (4D) reconstruction algorithm for gated cardiac SPECT. Previous work has shown that non-motion-specific 4D Gibbs priors enforcing smoothing in time and space can control noise while preserving resolution. In this paper, we evaluate methods for incorporating known heart motion in the Gibbs prior model. The new model is derived by assigning motion vectors to each 4D voxel, defining the movement of that volume of activity into the neighboring time frames. Weights for the Gibbs cliques are computed based on these `most likely' motion vectors. To evaluate, we employ the mathematical cardiac-torso (MCAT) phantom with a new dynamic heart model that simulates the beating and twisting motion of the heart. Sixteen realistically-simulated gated datasets were generated, with noise simulated to emulate a real Tl-201 gated SPECT study. Reconstructions were performed using several different reconstruction algorithms, all modeling nonuniform attenuation and three-dimensional detector response. These include ML-EM with 4D filtering, 4D MAP-EM without prior motion assumption, and 4D MAP-EM with prior motion assumptions. The prior motion assumptions included both the correct motion model and incorrect models. Results show that reconstructions using the 4D prior model can smooth noise and preserve time-domain resolution more effectively than 4D linear filters. We conclude that modeling of motion in 4D reconstruction algorithms can be a powerful tool for smoothing noise and preserving temporal resolution in gated cardiac studies.

UR - http://www.scopus.com/inward/record.url?scp=0030352402&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0030352402&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:0030352402

VL - 3

SP - 1923

EP - 1927

BT - IEEE Nuclear Science Symposium & Medical Imaging Conference

PB - IEEE

ER -