Figure 5
From: Label-free macrophage phenotype classification using machine learning methods
Label-free macrophage classification accuracy. (a) Classification accuracy using KNN, SVM and FCNN classifiers tested for their ability to classify the six macrophage phenotypes, where the FCNN is noted to provide the highest classification accuracy. (b) The classification accuracy of any 2, 3, 4, 5 phenotypes combinations, comparing to the 6 phenotypes. (c) The confusion matrix representing the accuracy of macrophage phenotyping using intrinsic autofluorescence. In blue are the correctly identified cells, while phenotypes not correctly identified are in pale red. Difference in color intensity is indicative for numerical values with darker shades representing higher values. The table on the right demonstrates overall summary for correctly and not correctly identified cells (i.e., true positives and false positive). (d) A representation of the utilized Fully Connected Neural Network (FCNN) architecture with a single wide hidden layer. The number of neurons in the figure are for illustration only, the actual utilized number is 45 for the input, 100 for the hidden layer, and 6 for the output as it will be further explained in the “Detailed methods and data analysis” section.
