MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel
  1. Home
  2. Meeting Program
  3. All
  4. Quantification of Parametric Response Maps for Response-Guided Adaptive Radiotherapy

Quantification of Parametric Response Maps for Response-Guided Adaptive Radiotherapy

H Zhong1*, H Sharifi2 , F Kong3 , X Li4 , C Liu5 , I Chetty6 , (1) Henry Ford Health System, Detroit, MI,(2) Henry Ford Health System, Detroit, MI, (3) Indiana University- School of Medicine, Indianapolis, IN, (4) Medical College of Wisconsin, Milwaukee, WI, (5) Henry Ford Health System, Detroit, MI, (6) Henry Ford Hospital System, Detroit, MI

Presentations

(Monday, 7/30/2018) 4:30 PM - 6:00 PM

Room: Davidson Ballroom B

Purpose: Parametric response maps (PRMs) derived from functional images such as FDG-PET or DW-MRI may serve as theragnostic images for dose painting. Quantification and standardization of these images, however, often involve large uncertainties. This study sought to evaluate the variation of PRMs constructed by different computational methods.

Methods: Two sets of FDG-PET images (PET1 and PET2) were acquired before RT and after fraction 18, respectively, for 4 lung cancer patients. For each patient, two CT-based deformable image registration algorithms (image-intensity-based B-Spline registration and finite element modeling), two registration orientations, and two SUV mapping techniques (SUV-interpolation and SUV-projection) were employed to generate 8 deformed PETs. In each pair of deformed and target PETs, PET1 (or mapped PET1) was taken as the baseline from which voxel-wise differences from the mapped PET2 (or PET2) constitute a PRM. To compare the computed PRMs, SUVs in the baseline PET were divided equally into 100 bins, with Si denoting the total SUVs in bin i. SUV differences between the baseline and its correspondent PET2 were summed in bin i to generate a response value Ri. An accumulated-response histogram at point p was defined by ARH(p)=(∑_(S_i≥p)▒R_i )/(∑_(S_i≥p)▒S_i ).

Results: For the 4 patients, CT-measured tumor volume was reduced by 25.4% and total lesion glycolysis (TGL) reduced by 46.6% on average. The mean and standard-deviation of ARH(p) calculated by the 8 methods were averaged over the 4 patients, with
results: 54.0±7.5% at p=2.5, and 64.8±7.4% at p=SUVmax. Among all computed ARH(2.5), the difference between the two registration algorithms was 4.96%, 6.01% between the two registration orientations, and 11.1% between the two SUV mapping techniques.

Conclusion: PRMs constructed with different techniques showed large variations which may result in dose painting uncertainties, affecting effectiveness of response-guided radiotherapy. Further investigation of these techniques with a large pool of data is warranted.

Keywords

Not Applicable / None Entered.

Taxonomy

TH- response assessment : PET imaging-based

Contact Email

Additional Links