Click here to


Are you sure ?

Yes, do it No, cancel

T2-Relaxation-Diffusion Correlation Analysis for Prediction of Progression Free Survival (PFS) in Glioblastoma

Y Li*, M Kim, T Lawrence, H Parmar, Y Cao, The University of Michigan, Ann Arbor, MI


(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

To identify potential imaging bio-markers for RT response assessment, we analyzed the T2-relaxation-diffusion correlation in glioblastoma (GBM).

Twenty-four patients with histologically diagnosed GBM were treated with either 60 Gy standard or 75 Gy intensified RT and had diffusion weighted (DW) images pre-RT. The 15 patients had mid-RT scans (week 3) to assess response. DW images were acquired with 11 b-values (0 to 2500 s/mm²) at 2 TEs (93 and 113 ms) on a 3T scanner. Hyper-cellular tumor volume (HCV) of each patient was defined on DW images with b=3000. A T2-diffusin correlation model that considered 2 T2 and 2 diffusion components was applied to pre-RT and mid-RT DW images in the HCV to yield 5 parameters: T2 and diffusion coefficient for fast and slow components (respective T2(f), D(f), T2(s) and D(s)) as well as the fraction of the slow component (V(s)). Changes in the parameters in response to doses at mid-RT were tested using Students’ t test. Kaplan-Meier (KM) analysis and step-wise Cox model were used to test these parameters for prediction of PFS.

T2(f) in the HCV pre-RT was significantly associated with inferior PFS by KM analysis (p=0.05). The boosted doses caused significant increases in T2(f) and D(f) compared to standard doses (p<0.05). Univariate Cox analysis showed that mid-RT Vs and MGMT were significant predictors of PFS (both p<0.05). In a multivariate Cox model, V(s), MGMT and age were significant for PFS (all p<0.05), suggesting that V(s) can add the value to clinical factors for prediction of PFS.

The T2-relaxation-diffusion correlation analysis yield the parameters that cannot be obtained by analyzing T2 and diffusion alone. The parameters have the potential to predict GBM response during RT, which could be valuable for adaptive RT.

Funding Support, Disclosures, and Conflict of Interest: This work was in part supported by NIH/NCI grant UO1 CA183848.


Not Applicable / None Entered.


Not Applicable / None Entered.

Contact Email