Skip to main content

Table 4 Effect of evaluation time parameters on the average precision of Cellpose models trained with seven training images, initialized with cyto weights, patch size 448\(\times\)448, and no random rotation in the data augmentation step

From: Optimizing deep learning-based segmentation of densely packed cells using cell surface markers

 

default

flow threshold

prob threshold

  

0.3

0.5

-1

1

1_CD8

0.730

0.728

0.726

0.706

0.729

2_CD3

0.746

0.733

0.748

0.730

0.741

3_CD4

0.795

0.780

0.782

0.772

0.778

4_CD3

0.649

0.630

0.646

0.619

0.647

5_CD3

0.699

0.664

0.698

0.666

0.687

6_CD3

0.710

0.684

0.703

0.661

0.716

7_CD3

0.648

0.623

0.640

0.621

0.657

mAP

0.711

0.692

0.706

0.682

0.708

Back to article page

BMC Medical Informatics and Decision Making

ISSN: 1472-6947

Contact us