Figure 2.
Accurate ResNeXt prediction for most morphological classes. Confusion matrix of the predictions obtained by the ResNeXt classifier on the test database annotated by gold standard labels provided by human experts. Plotted values were obtained by fivefold crossvalidation and are normalized row-wise to account for class imbalance. The number of single-cell images included in each category is indicated in the logarithmic plot on the right. Note the enhanced confusion probability between consecutive steps of granulopoiesis and erythropoiesis, as might be expected because of the unsharp delineations between individual morphological classes. Separate confusion maps of individual folds are shown in supplemental Figure 2.

Accurate ResNeXt prediction for most morphological classes. Confusion matrix of the predictions obtained by the ResNeXt classifier on the test database annotated by gold standard labels provided by human experts. Plotted values were obtained by fivefold crossvalidation and are normalized row-wise to account for class imbalance. The number of single-cell images included in each category is indicated in the logarithmic plot on the right. Note the enhanced confusion probability between consecutive steps of granulopoiesis and erythropoiesis, as might be expected because of the unsharp delineations between individual morphological classes. Separate confusion maps of individual folds are shown in supplemental Figure 2.

Close Modal

or Create an Account

Close Modal
Close Modal