Click here to


Are you sure ?

Yes, do it No, cancel

A Novel Mutual Information-Based Algorithm for IMRT/VMAT Quality Assurance

B Ebrahimi*, R Rahimi Darabad , K Padgett , N Dogan , University of Miami, Miami, FL


(Sunday, 7/29/2018) 4:30 PM - 5:00 PM

Room: Exhibit Hall | Forum 4

Purpose: To verify the accuracy of delivered dose in IMRT/VMAT treatments, measured via portal dosimeter, by comparing to predicted dose calculated by the TPS while utilizing a novel algorithm that may potentially address some of the limitations of conventional Gamma analysis, as well as enhance the sensitivity to subtle errors.

Methods: Portal dosimetry verification plans for 3 patients (11 total Arcs) were prepared in Eclipse and delivered using a Varian Trilogy Linac. To increase the temporal resolution, during the conversion of the treatment plans to verification plans, each arc was split into 90â?° sub-arcs (38 sub-arcs). Predicted and measured dose distributions were compared using both the novel normalized point-wise mutual information (nPMI)-based algorithm and the conventional Gamma analysis. Moreover, sensitivity of the algorithm to the dose gradient and flat dose distribution were assessed by calculating the nPMI score in two sets of simulated dose distributions in the presence of low and high dose gradient regions and spatial shifts.

Results: The Gamma analysis gave 91.4% [81.4-94.2] and 98.0% [95.9-99.2] for 2%/2mm and 3%/3mm criteria, respectively, while the nPMI scores for 2% and 3% dose mismatch tolerance were 0.80 [0.76-0.87] and 0.87 [0.81-0.96]. nPMI showed a closer correlation to dose differences as compared to Gamma. Additionally, in simulated gradients of 2 and 20% per-pixel, nPMI scores were independent of the gradient size and only varied, [0.46-0.49] at the edges, and [0.96-0.98] for the large areas with flat dose, these metrics depend primarily on the surface size of the shifted area and the tolerance level used.

Conclusion: Our results indicate that nPMI is potentially more sensitive to subtle but clinically meaningful discrepancies between the predicted and measured doses than Gamma analysis which struggles to detect discrepancies in areas of high dose gradients and flat dose distributions.


Not Applicable / None Entered.


TH- External beam- photons: Quality Assurance - IMRT

Contact Email