Support-vector machine

Question:

If we would give you a new data point for which you should predict whether the tumor is benign or malignant, using either The radial-basis kernel, setting up a search grid for = {0.1, 1, 2, 5, 10} OR The polynomial kernel, setting up a search grid for d = {1, 2, 5, 10}, both with 5-fold cross validation, which would you chose and why based on the ROC curves? Consider also ethical implications of False Positive versus False Negative of reporting a tumor as malignant.