B-0801 Treatment response assessment in Hodgkin lymphoma: in search for morphological correlates of metabolic activity

 T. Knogler, G. Karanikas, M. Weber, K. El-Rabadi, M.E. Mayerhoefer | Monday, March 11, 10:30 – 12:00 / Room F1

Purpose: To predict early treatment response with three-dimensional texture features (TF) in patients with Hodgkin lymphoma (HL) after radio-chemotherapy extracted from contrast-enhanced CT.
Methods and Materials: 21 patients with histologically proven HL were included in this study. Contrast-enhanced (18)F-FDG PET/CT was obtained on a dedicated PET/CT scanner. Volumes-of-interest and long- and short-axis diameter were manually defined on 48 HL manifestations prior and post-radio-chemotherapy on the CT image stack. Three-dimensional texture features derived from the grey-level histogram, co-occurrence matrix, run-length matrix and absolute gradient were calculated for the VOIs. A stepwise logistic regression with forward selection was performed to find classic radiologic features (i.e. lesion diameter, lesion volume) and TF, which correctly classify treatment response (i.e. full response and partial response). Classification in PET/CT was used as reference.
Results: Difference in short axis diameter best fit as classic feature with a sensitivity of 100 %, a specificity of 54,5% and an accuracy of 89,6%. Combination of “S_0_0_1_Entrp” and difference of “vertl_fraction” best fit as TF with a sensitivity of 97,3%, a specificity of 72,7% and an accuracy of 91,7%.
Conclusion: Texture features extracted from contrast-enhanced CT in patients with HL are superior in differentiation between responders and non-responders without the need for PET examinations, compared to classical radiological features. However, PET/CT as state-of-the-art imaging technique has a sensitivity and specificity >90%, so that further research with larger patient number is needed to investigate this new method.

Be Sociable, Share!

•