Z Razi¹*, A Bumin¹, H Gao¹, Z Zhang², M Arreola¹, (1) University of Florida, Gainesville, FL (2) Washington University School of Medicine, Saint Louis, MO

• 
  Z Razi

PO-GePV-I-44 (Sunday, 7/10/2022)   [Eastern Time (GMT-4)]

ePoster Forums

Purpose: A physical tissue relaxation time-mimicking phantom is essential for quantitative evaluation of MRI protocols. The aim of our current study is to address the need for improved standardization in MRI relaxometry and develop individual calculators to predict the amount of T1|T2 modifiers’ content. Quantitative analysis between T1|T2 modifiers’ content and relaxation times are presented.

Methods: A total of 30 homogeneous mixtures with varying concentrations of agarose and Gd-BOPTA were prepared for phantom design. A dedicated procedure was applied to prevent potential susceptibility artifact and air bubbles. An IR-based spin-echo sequence (TI range: 65-1610 ms) was used for T1-mapping. A multi-echo spin-echo (TE range: 8.5-272 ms) was used for T2-mapping. The quantitative maps were calculated with a homemade Matlab™ script and validated using qMRLab software. Generalized linear regression (GLR) were conducted for relaxation rates. Best model was selected based on small-sample Akaike Information Criterion. A machine-learning predictor was also developed using linear regression(LR) and random-forest(RF) classifiers for linear and non-linear models respectively.

Results: The ranges of agarose and Gd-BOPTA are 0-3% and 0.07-0.45μmol/kg and T1|T2 are 256-2296 ms and 46-1292 ms respectively. The achieved root-mean-squared-error(RMSE) using GLR for R1 model was 67.6s⁻¹ and for R2 was 12.2s⁻¹. RMSEs for agarose and Gd-BOPTA using LR and RF are 0.21. 0.08, 0.16, and 0.05 respectively. Gd-BOPTA is the most important term of the six-parameter model for R1, for both quadratic and cubic terms; while all other parameters accounted for less than 2% of the explained variance. A five-parameter model is the best for R2 with agarose being the most important. Other parameters accounted for less than 1.5% of the explained variance although the p-value was less than 0.05, hence statistically significant.

Conclusion: It is feasible by using derived models to accurately predict Gd-BOPTA and agarose content based on T1|T2 relaxation time measurement.

Keywords

MRI, Phantoms, Relaxation Times

Taxonomy

Not Applicable / None Entered.

Contact Email