MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Predictive Value of 0.35T MRI Delta-Radiomics Features in Image-Guided Pancreatic SBRT

G Simpson*, B Farnia , L Portelance , J Ford , N Dogan , F Yang , University of Miami, Miami, FL

Presentations

(Sunday, 7/29/2018) 2:05 PM - 3:00 PM

Room: Davidson Ballroom B

Purpose: MR-guided radiotherapy machines are becoming popular in radiation oncology for their ability to accurately track patient positioning without added dose. Low field strength (0.35T) images provide superior soft tissue contrast compared to conventional onboard imaging (CBCT). The daily images produced provide longitudinal image series. Delta-radiomics texture analysis of low field strength images could provide early treatment response data. The objective of the current study is to assess the discrimination power of delta-radiomics features between pancreatic patients who respond to treatment and those who are non-responsive.

Methods: Thirteen gray-level size zone matrix (GLSZM) features were extracted from the gross tumor volume (GTV) on the first daily treatment image and after the third fraction (image acquired before fourth fraction). The initial patient population consisted of nine patients undergoing SBRT with dose prescriptions ranging from 35 Gy to 50 Gy in five fractions. Delta texture features were calculated by finding the difference of features between the two image series. The delta-radiomic features were used to differentiate treatment responses between responders (partial and near complete response, n=6) or non-responders (no response, n=3). Wilcoxon signed-rank test (JMP Pro, SAS Institute, Cary, NC) was employed for comparison using a significance level of p-value < 0.05.

Results: Despite the small patient population, a univariate analysis identified two gray-level size zone matrix (GLSZM) features as statistically significant. Zone size variance (ZSV) and gray-level variance (GLV) were able to differentiate between responders and non-responders. ZSV and GLV decreased in both responders and non-responders.

Conclusion: Pancreatic tumors are difficult to treat and prediction of response based on delta-radiomics would allow physicians to adapt treatment with two fractions remaining. Delta-radiomic analysis of low field MR images has the ability to differentiate treatment response before the beginning of the fourth fraction and warrants further investigation.

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email