Table 2.

Cell classification experiment, based on cropped-out patches of the manually expert labeled (MEL), object detection-augmented and expert labeled (ODAEL) and the clustering + object detection-augmented and expert labeled (CODAEL) dataset variants.

Metric	MEL variant	ODAEL variant	CODAEL variant
Precision	0.675 ± 0.055	0.642 ± 0.040	0.679 ± 0.044
Recall	0.898 ± 0.023	0.883 ± 0.031	0.891 ± 0.028
ROC AUC	0.926 ± 0.014	0.930 ± 0.002	0.944 ± 0.007

ROC AUC indicates the area under the receiver operating characteristic curve. Values represent mean ± standard deviation of five independent training and inference runs.