MTCSNet: One-stage learning and two-point labeling are sufficient for cell segmentation | IEEE Journals & Magazine | IEEE Xplore