A robust segmentation method with triple‐factor non‐negative matrix factorization for myocardial blood flow quantification from dynamic ⁸²Rb positron emission tomography

Abstract

Purpose: In this work, we proposed a triple-factor non-negative matrix factorization (TNMF) method to semiautomatically segment the regions of interest (ROIs) of the left ventricular (LV) cavity and myocardium to improve the reproducibility of myocardial blood flow (MBF) quantification from dynamic 82 Rb positron emission tomography (PET). Methods: The proposed TNMF method was evaluated using NCAT phantom simulation with three noise levels. The segmented ROIs, time-activity curves (TACs), and K 1 derived from the TNMF method were compared with the ground truth simulated. The TNMF method was further evaluated in two patients each undergone both rest and stress 82 Rb PET studies. The TNMF and manual segmentations were implemented by two different observers, and the interoperator variations of MBF and myocardial flow reserve (MFR) were compared between the two methods. Results: Our simulation results showed that the TNMF method for dynamic PET image segmentation was robust as evidenced by the high Dice similarity coefficient, regardless of high or low count level. The relative bias in K 1 estimation was less than 1%. Our patient results also showed that reasonable ROIs for the LV cavity and myocardium could be obtained precisely for patients with and without myocardial perfusion defects. The TACs derived from the TNMF method were highly correlated with those obtained with the manual method (R 2 ≥ 0.964). The interoperator variations of MBF and MFR were markedly reduced using the TNMF method. Conclusions: In conclusion, the TNMF method is highly feasible for semiautomatic segmentation of the LV cavity and myocardium, with the potential to improve the precision of MBF quantification by improving segmentation.

Cardiac PET imaging: Lost in quantification. It’s time to find the way

“…[7][8][9][10][11] Semiautomatic and automatic segmentation methods, based on various statistical scenarios, have been proposed to improve the reproducibility of PET quantification. [12][13][14] Factor analysis is widely used for extracting tissue TAC in dynamic PET images. 13 This approach is based on the assumption that dynamic PET noise and the model approximation errors follow Gaussian distributions.…”

“…More recently, a triple-factor non-negative matrix factorization (TNMF) method for semiautomatic LV cavity and myocardial segmentation has been proposed and it has been found to be highly feasible for MBF and MPR quantification. 14 In the current issue of Journal of Nuclear Cardiology, Liu et al 15 proposed a new software tool called ''Yale-MQ,'' based on an optimized TNMF segmentation workflow and on one-tissue compartment model. The authors evaluated quantitative precision and intra-/ inter-observer variabilities of Yale-MQ in MBF and MFR quantification in a cohort of normal healthy volunteers (n = 18) and a group of CAD patients (n = 62) who underwent cardiac imaging by 82 Rb.…”