Figure 6
From: Automatic classification of normal and abnormal cell division using deep learning
Examples of false negatives (abnormal mitosis classified as normal) and false positives (normal mitosis classified as abnormal) from the ResNet50 model applied to test data. Each example displays 4 of the 12 frames in each mitosis sample. In (A), a cell clearly divides completely and then reunites, making the mitosis abnormal, but the algorithm might mistake the initial division as normal. In (B), the cell tries to divide but fails, leading the algorithm to mistake a different cell in the image as a second daughter and classify the mitosis as normal. In (C), the cell divides correctly, but a different cell appearing in the image might confuse the network into misclassifying the mitosis as abnormal. In (D), both daughter cells end up separate, but the algorithm might mistake them as one cell, leading to classification failure.
