Tumor mask

[chongweiwu]

dear author, thanks for your excellent work. I have several questions about the tumor masks : first, wether the tumor used in the measurement contains the edema areas? second, have you done any post-treatment on the segmented tumor masks? third, what is the dilation structure, in your work, during the morphological dilation of tumors?

[cwmok]

Dear @chongweiwu,

1. During the evaluation, the tumor mask we used only contained the tumor core (whole tumor without the edema).
2. No, we don't have any post-processing to the segmented tumor masks.
3. The structure we used is an element with square connectivity equal to one. (scipy.ndimage.binary_dilation)

[chongweiwu]

I am really appreciate for your detailed reply. I’m worried that my divided two landmarks sets may differ from yours, so can you provide me with the source code, dilated tumor images or the divided two landmarks sets? my email adress is chongweiwu@hust.edu.cn thanks

[cwmok]

Suppose tumor_seg is a segmentation map of the tumor core, we first do the binary dilation as follows:

from scipy.ndimage import binary_dilation tumor_seg = binary_dilation(tumor_seg, iterations=30)

Then, to determine the class (tumor/non-tumor region) of each landmark, a simple if-else statement is used:

for i in range(Y_label.shape[0]):
    if tumor_seg[int(Y_label[i][0]), int(Y_label[i][1]), int(Y_label[i][2])] == 1:
        print("Tumor", Y_label[i])
        Y_label_tumor.append(Y_label[i])
        X_label_tumor.append(X_label[i])
    else:
        print("Non-Tumor", Y_label[i])
        Y_label_nontumor.append(Y_label[i])
        X_label_nontumor.append(X_label[i])

[chongweiwu]

Thank you again for your kindly reply!