Figure 3.
Generalization of SpinX to segment images of new marker proteins and cell types acquired using distinct microscopes. (a) Representative images and SpinX segmentation output of mCherry-Tubulin and SiR-Tubulin datasets used to train the spindle model (specialized). (b and c) Representative images and SpinX segmentation output of YFP-Astrin (b), mRFP-EB3 and mKate2-EB3 (c) datasets used to assess the extent to which the spindle model can be generalized. The dashed box shows the original image border (YFP-Astrin and mKate2-EB3 datasets were padded for analysis). Cartoons in a, b, and c show differing localization patterns (red) of spindle marker proteins. Images acquired using a widefield or higher resolution confocal microscope are highlighted, where the associated objective used is indicated. (d) Bar graph shows SpinX’s segmentation accuracy of spindles labeled using mRFP-EB3, YFP-Astrin, and Tubulin (mCherry-Tubulin and SiR-Tubulin combined) acquired with a widefield microscope, and mKate2-EB3 using a confocal microscope as indicated. Accuracy was manually scored by experts using the error classification system indicated in Fig. 2, c and e. (e) Representative images and SpinX segmentation output of bovine one-cell embryo, bovine two-cell embryo, HEK293, and mESC datasets used to assess the extent to which the spindle model can be generalized to spindles from other cell types. Images provided by Kletter et al. (2022) were acquired using a confocal microscope, where the associated objective used is highlighted. (f) Bar graph shows SpinX’s segmentation accuracy of spindles from other cell types, segregated into either the “Spindle pole inclusion” or “Spindle morphology” category. Categories were created based on the error classification system outlined in Fig. 2, whereby “Spindle pole inclusion” includes both images classified as “Correct” and “U-minor,” while “Spindle morphology” includes only images classified as “Correct.” Spindles with a visible midzone were chosen. Accuracy was manually scored by experts. mRFP-EB3 N = 5 cells, 1,540 images; YFP-Astrin N = 10 cells, 330 images; SiR/mCherry-Tubulin N = 10 cells, 630 images; mKate2-EB3 N = 5 cells, 1,920 images; bovine one-cell embryo N = 10 cells, 30 images; bovine two-cell embryo N = 10 cells, 30 images; HEK293 N = 10 cells, 130 images; mESC N = 10 cells, 50 images. Scale bars: 10 µm.

Generalization of SpinX to segment images of new marker proteins and cell types acquired using distinct microscopes. (a) Representative images and SpinX segmentation output of mCherry-Tubulin and SiR-Tubulin datasets used to train the spindle model (specialized). (b and c) Representative images and SpinX segmentation output of YFP-Astrin (b), mRFP-EB3 and mKate2-EB3 (c) datasets used to assess the extent to which the spindle model can be generalized. The dashed box shows the original image border (YFP-Astrin and mKate2-EB3 datasets were padded for analysis). Cartoons in a, b, and c show differing localization patterns (red) of spindle marker proteins. Images acquired using a widefield or higher resolution confocal microscope are highlighted, where the associated objective used is indicated. (d) Bar graph shows SpinX’s segmentation accuracy of spindles labeled using mRFP-EB3, YFP-Astrin, and Tubulin (mCherry-Tubulin and SiR-Tubulin combined) acquired with a widefield microscope, and mKate2-EB3 using a confocal microscope as indicated. Accuracy was manually scored by experts using the error classification system indicated in Fig. 2, c and e. (e) Representative images and SpinX segmentation output of bovine one-cell embryo, bovine two-cell embryo, HEK293, and mESC datasets used to assess the extent to which the spindle model can be generalized to spindles from other cell types. Images provided by Kletter et al. (2022) were acquired using a confocal microscope, where the associated objective used is highlighted. (f) Bar graph shows SpinX’s segmentation accuracy of spindles from other cell types, segregated into either the “Spindle pole inclusion” or “Spindle morphology” category. Categories were created based on the error classification system outlined in Fig. 2, whereby “Spindle pole inclusion” includes both images classified as “Correct” and “U-minor,” while “Spindle morphology” includes only images classified as “Correct.” Spindles with a visible midzone were chosen. Accuracy was manually scored by experts. mRFP-EB3 N = 5 cells, 1,540 images; YFP-Astrin N = 10 cells, 330 images; SiR/mCherry-Tubulin N = 10 cells, 630 images; mKate2-EB3 N = 5 cells, 1,920 images; bovine one-cell embryo N = 10 cells, 30 images; bovine two-cell embryo N = 10 cells, 30 images; HEK293 N = 10 cells, 130 images; mESC N = 10 cells, 50 images. Scale bars: 10 µm.

This Feature Is Available To Subscribers Only

or Create an Account

Close Modal
Close Modal