It is typically advised to have a suitably enough sample size to ensure the ROC curve's reliability and correctness. A general rule of thumb is to have at least 50 positive and 50 negative instances, however this may not always be enough in all circumstances.

Thank you for response. it was a question from a reviewer, he asked me to refer to a professional statistician to judge my results. did you have any reference ?

Hello Ikram Chebieb

You may find it useful to see the following resource.

• Hanley JA, McNeil BJ (1982) The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology, 143:29-36.

It is one of the oldest resources available that most softwares use.

Abolfazl Ghoodjani thank you so much, I appreciate your help