MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel
  1. Home
  2. Meeting Program
  3. All
  4. Delta-Radiomics Features Predict Response and Outcome in Head and Neck Cancer Patients Treated with Chemoradiation

Delta-Radiomics Features Predict Response and Outcome in Head and Neck Cancer Patients Treated with Chemoradiation

E Florez*, T Thomas , S Lirette , A Fatemi , University of Mississippi Med. Center, Jackson, MS

Presentations

(Monday, 7/15/2019) 1:15 PM - 1:45 PM

Room: Exhibit Hall | Forum 2

Purpose: Surveillance imaging for patients with chemoradiation inherent pitfalls due to a difficulty to differentiate residual disease from radiation changes. This study assessed the tonsillar cancer patients treated with chemoradiation purpose to differentiate residual disease from radiation changes using radiomics features extracted from pre-treatment and post-treatment CT images.

Methods: Retrospective analysis of 80 patients with squamous cell carcinoma of the tonsil treated with chemoradiation. (15/80, 19%) patients reported residual disease on CT after 2 months. Among them, 12 patients underwent a PET/CT one month later revealed residual disease in 7 patients. Next, 6/7 patients underwent surgery. Of them, 4/6 had pathologically proven residual disease and 2/6 had no evidence of disease. All GTVs transferred from CT to MIM Software. Oncologist contoured on 2 months follow up CT and 3 months follow up PET. GTVs from baseline and post-therapies CT exported to MatLab. CT radiomics features (Rf=250) were measured on all tumor contours using different approaches, including first-, second-, and higher-order statistics, both with and without normalization.

Results: 250 texture parameters examined for predictive utility in an individual univariate logistic model with positive PET/CT as the outcome. These 250 areas under the curve (AUCs) were then ranked and any variable whose AUC was greater than or equal to 0.70 was added to a multivariate cross-validated ridge regression model. This process was stratified by normalization scheme, having been the first group conformed by the GTV without normalization, and the other two groups normalized by standard deviation [μ-3σ, μ + 3σ] and by the brightness level [1% -99%], respectively. Excellent AUCs values were obtained, mainly by the scheme without normalization (AUC = 0.889) which indicates that the segmentation process of the GTVs was quite accurate.

Conclusion: Radiomic analysis in both CT and PET were able to predict the residual tumor from radiation changes.

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email

Additional Links