Fig. 2

Cascade R-CNN training to distinguish non-senescent and senescent cells. A Concept of the Cascade R-CNN system. Microscopic images of iMSCs during culture were acquired, and the Cascade R-CNN was trained to detect non-senescent and senescent cells. B The AP, mAP, and AR showed the performance of the Cascade R-CNN trained by passage data. C Precision-recall curve of the trained Cascade R-CNN. D The mAP of the small, medium, and large objects. E The AR of the small, medium, and large objects. F Several evaluation indexes for non-senescent and senescent cell detection. G The performance of the Cascade R-CNN in serial passages. H The performance of deep learning methods. I The Pearson correlation coefficient between the senescence proportion output of the Cascade R-CNN and senescence-related indicators. Data were representative of three independent experiments, n = 3