Fig. 9: DM performance with brightness test.
From: Deep-Manager: a versatile tool for optimal feature selection in live-cell imaging analysis

Classification performance of DM selection in combination with SVM classification model in terms of ACC (left y-axis panel a) and F1-score (left y-axis panel b) vs AUC (right y-axis) in recognizing BT-474 cells morphological changes across 4 h of culture at day 0. The independent test set of images has been modified according to the brightness test for the 2D scenario. Blue lines denote performance results calculated over the training set, red lines denote performance results calculated over the modified test set, green lines represent the performance results achieved by not selecting features, cyan line identifies the performance of the fine-tuned DENSENET201 network, black lines represent the performance of applying DM feature selection procedure to the features extracted by the fine-tuned DENSENET201 network. n = 10 repetitions were used to extract boxplot.