Improving Treatment Response Prediction Using A Combination Of Delta-Radiomics And Clinical Biomarker

Recently we showed that a change in radiomics features (delta radiomics features, DRF) from daily CT-guided chemoradiation therapy (CRT) is associated with treatment response for pancreatic cancer. Carbohydrate antigen (CA19-9) is a widely used clinical biomarker for pancreatic cancer. However, the predictive power of such biomarkers (DRF or CA19-9) for CRT response is limited if used alone. The purpose of this work is to investigate if the predictive power can improve combining both biomarkers. Daily non-contrast CTs acquired during routine CT-guided pre-operative CRT for 24 patients with resectable pancreatic head cancer, along with their CA19-9 tests, pathology reports and follow up data were analyzed. For each case, the pancreatic head was segmented manually and inspected on each daily CT to ensure consistency. Changes in 73 radiomic features were extracted from the daily segmented regions. The CA19-9 test results acquired for all available fractions were used. The time between the end of treatment and last follow up was used to build a survival model. Patients were divided into two groups based on their pathological response. Spearman correlation coefficient was used to find the correlation of DRFs and CA19-9. A regression model was built to examine the effect of combining CA19-9 and DRFs on response prediction. A concordance statistic was used to measure model effectiveness. The effect of normalized (or a decrease in) CA19-9 levels during treatment versus failure of CA19-9 levels to normalize (or an increase) on survival was examined. A Cox proportional univariate and multivariate hazard regression analysis were performed to determine the effect of combining CA19-9 and DRFs on survival correlations. A Spearman correlation showed that CA19-9 is correlated to DRFs (Entropy, cluster tendency and coarseness). Increasing CA19-9 levels during treatment were correlated to a bad response, while a decline in CA19-9 levels, was correlated to a good response. Incorporating CA19-9 with DRFs increased the concordance statistic from 0.57 to 0.87 indicating stronger prediction model. The univariate analysis showed that patients with decreasing CA19-9 had an improved median survival (68 month) compared to those with increasing levels (33 month). The 5-year survival was improved for the decreasing CA199 group (55%) compared to the increasing group (30%). The Cox proportional multivariate hazard analysis showed that a treatment related decrease in CA19-9 levels (p=0.031) and DRFs (p=0.001) were independent predictors of survival. The hazard ratio was reduced from 0.73, p=0.032 using CA19-9 only to 0.58, p=0.028 combining DRFs with CA19-9. A new oncologic profile combining delta radiomics and clinical biomarkers has the potential to lead to faster prediction of treatment response, increasing the possibility for response-based treatment adaptation.