Figure 3.

Cellular classification. Classification test results obtained by decision tree classification using as features the size-corrected roundness at the initial position, the final position, and at the point of the maximum deformation. (a) Confusion matrix for classification of individual cells (AUC = 0.58). (b) Confusion matrix for classification of groups of five consecutive cells (from the same class) based on features obtained from their average corrected roundness profile (AUC = 0.72). (c) Cell line prevalence in a mixed sample (n = 5000) obtained using Bayesian inference based on the specificity and sensitivity reported in (a). Plotted are posterior densities (means and 95% posterior probability intervals) for the inferred prevalence of ‘positive’ cells. Densities were computed for observed prevalence values that corresponded to the cell classifier classifying k cells as positive where k ∈ {3100, 3200, …, 4300, 4400}.