Premium
Sci—Fri AM: Imaging — 01: Feasibility of estimating choline kinase activity with kinetic modeling of 18F‐fluorocholine pet imaging of prostate cancer
Author(s) -
Blais AR,
Lee TY
Publication year - 2012
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4740187
Subject(s) - prostate cancer , positron emission tomography , nuclear medicine , matlab , medical imaging , principal component analysis , prostate , noise (video) , artificial intelligence , computer science , cancer , pattern recognition (psychology) , medicine , image (mathematics) , operating system
Prostate cancer (PCa) detection and delineation remains a challenge for medical imaging. Studies have shown 18 F‐Fluorocholine (FCH) PET imaging to be a promising modality in the detection of recurrent PCa. Detection of denovo PCa is more challenging, as lesions such as benign prostatic hyperplasia (BPH) may adversely affect the sensitivity and specificity of the modality. PCa and BPH have been shown to exhibit similar uptake of FCH, yet it has been shown that phosphocholine levels are much more elevated in PCa compared to BPH. Therefore, it would be useful to measure the activity of phosphorylation via choline kinase (k 3 ) in order to differentiate PCa from BPH. This work examines the feasibility of using a compartmental model to estimate k 3 with dynamic 18 F‐Fluorocholine PET imaging. JSim software [1] was used to simulate the compartmental model for FCH exchange. A simulated tissue curve was generated using predefined parameters and the model's ability to estimate these parameters through fitting of the simulated tissue curve with and without noise was investigated. The fitting procedure was performed using the non‐negative least squares algorithm in MATLAB after the equation governing fitting was linearized. In the noiseless case, the model was able to accurately identify the values of each rate parameter. For the noisy case with an SNR of 10:1, the mean estimated k 3 for 10,000 runs had a coefficient of variation of 14.9%. The kinetic model shows promise for quantifying k 3 , which would allow the differentiation of malignant and benign tumours of the prostate.