MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

CT Textures Predict Post-Treatment Radiation-Related Fibrosis in Head-And-Neck Cancer Patients

D Toomeh1*, D Kwon3, G Azzam4, S Samuels5,I Mihaylov2, (1) University of Miami, Miami, FL, (2) Univ Miami, Miami, FL, (3) ,Miami, FL, (4) ,,,(5) ,Miami, FL

Presentations

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

Room: AAPM ePoster Library

Purpose: Radiation therapy can induce skin and muscle changes which result in fibrosis formation which reduces the quality of life after treatment. Our hypothesis is that CT textures may serve as a predictor for radiation-induced fibrosis in the neck.
Methods: Sixty head-and-neck patients who underwent radiation treatment alone or chemoradiation were reviewed. Delineation of Skin, subcutaneous tissue (SCT) and sternomidomastoid (SCM) muscle was performed. Skin and SCT volumes were cropped at the 10 Gy isodose line. Approximately 90 textures were extracted using our in-house developed software. Radiation-induced fibrosis grade was dichotomized according to validated scleroderma (SSPRO) survey into mild [score 0-25], and severe [score 26-100]. Odds ratio (OR) along with the corresponding 95% confidence interval, and p-value for each imaging feature were estimated using a univariate logistic regression model. In addition, the predictive value of the features were determined by Random forest histogram-based area under the curve (AUC).
Results: In the statistical analyses only 3 skin, 1 SCT and 3 SCM features indicate significant associations with the dichotomized RIF grade (two-tailed p-values < 0.05), showing substantial effect with AUC of 0.757 skin, 0.806 SCT, and 0.786 SCM.

Conclusions:
The presented results suggest that the pre-treatment CT textures can provide a valuable prediction of the development of RIF in head and neck cancer patients undergoing RT. Future work include, expansion of the database and validation through multivariate analyses.

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email