Quantitative Experimental Phantom Study Based on Abdominal MR Contrast Media Using Deep Learning

Abstract

This study provides data for the development of oral contrast media for abdominal magnetic resonance imaging (MRI) examinations for potential use in clinical practice. The signal intensities, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) were quantified using various contrast media with longitudinal (T1) and transverse relaxation (T2) pulse sequences. Prediction accuracy error comparisons were conducted according to the mean-squared, mean-absolute, and root-mean-squared errors of the contrast media intensities using the Orange data mining software. The signal strength and SNR were higher in canola oil and pineapple juice (T1-weighted images), while the intensities of blueberry juice and apple juice were high in the T2-weighted images; SNR was high in blueberry and cranberry juice, and CNR was high in Solotop ® and blueberry syrup. The accuracy of the deep-learning prediction errors of MR signal intensities was high. In conclusion, data from ex vivo MRI research can be used for the development of oral contrast media.