FIGURE 7.

Prediction of metastatic phase from comprehensive flow cytometric data for immune cells. (A) Workflow for prediction of metastatic phase with the use of machine learning or deep learning. Normalized flow cytometry data (lung or peripheral blood) was split into training data or testing data at a ratio of 4:1, and training data were applied to machine learning or deep learning. The testing data were then used to predict metastatic phase and to calculate the prediction accuracy. The normalized flow cytometry data were randomly split five times, with learning and testing (prediction) by machine learning or deep learning also being performed five times in a corresponding manner and the mean accuracy calculated. (B) Accuracy of metastatic phase prediction. Data are means +95% confidence interval (n = 5), with the mean accuracy also being shown above each bar. (C) Mean accuracy for prediction of each metastatic phase by logistic regression and an example of the prediction results.