Optimized Detection of Left Ventricular Hyperpolarized [1‐ 13 C]Pyruvate Signal in Human Cardiac Metabolic Imaging

ABSTRACT Purpose This study aimed to improve the analysis of the left ventricular (LV) blood pool hyperpolarized [1‐ 13 C]pyruvate signal for a more reliable quantification of downstream metabolic outcome. Methods Hyperpolarized [1‐ 13 C]pyruvate magnetic resonance spectroscopic images were acquired dynamically at 3 Tesla in 106 participants (NCT03685175). An exponential model was fitted to the distribution of summed LV blood pool [1‐ 13 C]pyruvate signal intensities from all dynamic frames to extract a characteristic numerical fitting value. A threshold formula was introduced to detect a suitable region of interest in the blood pool. A correlation‐based analysis between the LV blood pool [1‐ 13 C]pyruvate signal and the metabolic outcome, i.e., combined myocardial [1‐ 13 C]lactate and [ 13 C]bicarbonate signals, was performed to determine an optimal threshold and was compared to conventional peak‐intensity and percentile‐based approaches. Results The exponential numerical fitting value (0.0105 ± 0.0036) of [1‐ 13 C]pyruvate signal intensity distribution was normalized to the area under the curve of the average signal‐to‐noise ratio. A threshold range of 13%–25% was identified where the resulting values representative for the LV blood pool [1‐ 13 C]pyruvate signal reached an optimal correlation with the metabolic outcome. The proposed model showed a more stable correlation between LV blood pool [1‐ 13 C]pyruvate signal and the metabolic outcome compared to conventional peak‐intensity and percentile‐based approaches. Conclusions A mathematical model using exponential fitting and thresholding of the intensity distribution of [1‐ 13 C]pyruvate signal in the LV blood pool provides a strong correlation with the metabolic outcome, allowing for a semiautomated method to robustly assess the left ventricular pyruvate signal in dynamic nuclear polarization studies.