A deep learning–based model for postoperative resection assessment in glioblastoma: a comparative validation study