### Abstract

Current thick detectors used in medical imaging allow recording many attributes, such as the 3D location of interaction within the scintillation crystal and the amount of energy deposited. An efficient way of dealing with these data is by storing them in list-mode (LM). To reconstruct the data, maximum-likelihood expectation-maximization (MLEM) is efficiently applied to the list-mode data, resulting in the list-mode maximum-likelihood expectation-maximization (LMMLEM) reconstruction algorithm. In this work, we consider a PET system consisting of two thick detectors facing each other. PMT outputs are collected for each coincidence event and are used to perform 3D maximum-likelihood (ML) position estimation of location of interaction. The mathematical properties of the ML estimation allow accurate modeling of the detector blur and provide a theoretical framework for the subsequent estimation step, namely the LMM-LEM reconstruction. Indeed, a rigorous statistical model for the detector output can be obtained from calibration data and used in the calculation of the conditional probability density functions for the interaction location estimates. Our implementation of the 3D ML position estimation takes advantage of graphics processing unit (GPU) hardware and permits accurate real-time estimates of position of interaction. The LMMLEM algorithm is then applied to the list of position estimates, and the 3D radiotracer distribution is reconstructed on a voxel grid.

Original language | English (US) |
---|---|

Title of host publication | IEEE Nuclear Science Symposium Conference Record |

Pages | 2643-2647 |

Number of pages | 5 |

DOIs | |

State | Published - 2010 |

Event | 2010 IEEE Nuclear Science Symposium, Medical Imaging Conference, NSS/MIC 2010 and 17th International Workshop on Room-Temperature Semiconductor X-ray and Gamma-ray Detectors, RTSD 2010 - Knoxville, TN, United States Duration: Oct 30 2010 → Nov 6 2010 |

### Other

Other | 2010 IEEE Nuclear Science Symposium, Medical Imaging Conference, NSS/MIC 2010 and 17th International Workshop on Room-Temperature Semiconductor X-ray and Gamma-ray Detectors, RTSD 2010 |
---|---|

Country | United States |

City | Knoxville, TN |

Period | 10/30/10 → 11/6/10 |

### Fingerprint

### ASJC Scopus subject areas

- Radiation
- Nuclear and High Energy Physics
- Radiology Nuclear Medicine and imaging

### Cite this

*IEEE Nuclear Science Symposium Conference Record*(pp. 2643-2647). [5874269] https://doi.org/10.1109/NSSMIC.2010.5874269

**List-mode MLEM image reconstruction from 3D ML position estimates.** / Caucci, Luca; Hunter, William C J; Furenlid, Lars R; Barrett, Harrison H.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

*IEEE Nuclear Science Symposium Conference Record.*, 5874269, pp. 2643-2647, 2010 IEEE Nuclear Science Symposium, Medical Imaging Conference, NSS/MIC 2010 and 17th International Workshop on Room-Temperature Semiconductor X-ray and Gamma-ray Detectors, RTSD 2010, Knoxville, TN, United States, 10/30/10. https://doi.org/10.1109/NSSMIC.2010.5874269

}

TY - GEN

T1 - List-mode MLEM image reconstruction from 3D ML position estimates

AU - Caucci, Luca

AU - Hunter, William C J

AU - Furenlid, Lars R

AU - Barrett, Harrison H

PY - 2010

Y1 - 2010

N2 - Current thick detectors used in medical imaging allow recording many attributes, such as the 3D location of interaction within the scintillation crystal and the amount of energy deposited. An efficient way of dealing with these data is by storing them in list-mode (LM). To reconstruct the data, maximum-likelihood expectation-maximization (MLEM) is efficiently applied to the list-mode data, resulting in the list-mode maximum-likelihood expectation-maximization (LMMLEM) reconstruction algorithm. In this work, we consider a PET system consisting of two thick detectors facing each other. PMT outputs are collected for each coincidence event and are used to perform 3D maximum-likelihood (ML) position estimation of location of interaction. The mathematical properties of the ML estimation allow accurate modeling of the detector blur and provide a theoretical framework for the subsequent estimation step, namely the LMM-LEM reconstruction. Indeed, a rigorous statistical model for the detector output can be obtained from calibration data and used in the calculation of the conditional probability density functions for the interaction location estimates. Our implementation of the 3D ML position estimation takes advantage of graphics processing unit (GPU) hardware and permits accurate real-time estimates of position of interaction. The LMMLEM algorithm is then applied to the list of position estimates, and the 3D radiotracer distribution is reconstructed on a voxel grid.

AB - Current thick detectors used in medical imaging allow recording many attributes, such as the 3D location of interaction within the scintillation crystal and the amount of energy deposited. An efficient way of dealing with these data is by storing them in list-mode (LM). To reconstruct the data, maximum-likelihood expectation-maximization (MLEM) is efficiently applied to the list-mode data, resulting in the list-mode maximum-likelihood expectation-maximization (LMMLEM) reconstruction algorithm. In this work, we consider a PET system consisting of two thick detectors facing each other. PMT outputs are collected for each coincidence event and are used to perform 3D maximum-likelihood (ML) position estimation of location of interaction. The mathematical properties of the ML estimation allow accurate modeling of the detector blur and provide a theoretical framework for the subsequent estimation step, namely the LMM-LEM reconstruction. Indeed, a rigorous statistical model for the detector output can be obtained from calibration data and used in the calculation of the conditional probability density functions for the interaction location estimates. Our implementation of the 3D ML position estimation takes advantage of graphics processing unit (GPU) hardware and permits accurate real-time estimates of position of interaction. The LMMLEM algorithm is then applied to the list of position estimates, and the 3D radiotracer distribution is reconstructed on a voxel grid.

UR - http://www.scopus.com/inward/record.url?scp=79960311105&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=79960311105&partnerID=8YFLogxK

U2 - 10.1109/NSSMIC.2010.5874269

DO - 10.1109/NSSMIC.2010.5874269

M3 - Conference contribution

AN - SCOPUS:79960311105

SN - 9781424491063

SP - 2643

EP - 2647

BT - IEEE Nuclear Science Symposium Conference Record

ER -