Treatment Response Assessment Maps (TRAMs): Increased/Decreased Sensitivity to Tumor As a Function of Acquisition/Analysis Parameters

Abstract

TRAMs calculated from delayed-contrast MRI enable reliable differentiation (sensitivity/specificity>90%) between tumor (blue in the TRAMs) and non-tumor tissues (red). The TRAMs are calculated by subtracting 3D-T1-MRIs acquired 5min (early time point) post-contrast injection from those acquired 60-105min (late point) later. Here we studied the sensitivity of the calculated tumor/treatment-effects volumes to the T1-MRIs acquisition times and TRAMs normalization parameters. 15 patients with high grade glioma and 17 with brain metastases were scanned by the TRAMs protocol (T1-MRIs acquired 5min and 60-105min post contrast injection). Tumor/treatment-effects volumes were calculated from the TRAMs within regions appearing enhanced on T1-MRIs. A sub-cohort of 16 patients was also scanned with a rapid 3D-T1-MRI sequence (20 sec) acquired 2, 5, 12, 17, 20, 24, and 70 min post-contrast. A mathematical model was developed to evaluate the sensitivity of the tumor volumes, calculated from the TRAMs, to the times of the early and late acquisitions. The free parameters of this model were determined by fitting the rapid 3D-T1-MRI data to the model. The model was then used to assess the changes in calculated tumor volumes as a function of the early/late acquisition times post contrast. The TRAMs color window is normalized to the change in the signal intensity between the 2 acquisitions of a chosen vessel. The data of additional 16 patients was used to calculate the TRAMs while choosing different vessels for normalization and studying the normalization effect on the resulting tumor volumes. The model showed that tumor volumes calculated from the TRAMs increase by x4.4 when moving the early time point from 2min to 15.7±2.2min post contrast, where they plateau. At 5min, the time determined optimal for standard TRAMs acquisition, the sensitivity of the tumor volume to changes in the early time point is 10%/min. The sensitivity to changes in the delayed time point was found negligible between 60-105 min post-contrast: 0.01%/min. The maximal change in tumor volumes as a function of the blood vessel chosen for normalization was found to be <5%. The TRAMs were previously shown to provide reliable differentiation between tumor/treatment-effects. The mathematical model confirms that choosing the early time point at 5±0.5min results in an error <5% in calculated tumor volumes. Using shorter delays significantly decreases the sensitivity to tumor while longer delays increase the sensitivity. Therefore, an additional 3D-T1-MRI acquired at 15min may be applied for calculating additional TRAMs with significantly higher sensitivity to tumor, for depicting small tumor regions. Errors in calculated tumor volumes due to changes in the delayed acquisition time and in the vessel normalization were found to be <5%, confirming an overall error of <7% in tumor volumes calculated from standard TRAMs.