Table 1:
Average classification accuracy for the LDA classifier using as input MLP (Multiparameter Persistence Landscape), ECC, or ECP. Data for each tumor are split into 80/20 train–test splits and classification accuracy is reported as the mean over 100 repetitions of splitting, training, and testing.
| CD68+ vs. FoxP3+ | CD8+ vs. FoxP3+ | CD8+ vs. CD68+ | CD8+ vs. CD68+ vs. FoxP3+ | |
|---|---|---|---|---|
| MPL - ECC - ECP | MPL - ECC - ECP | MPL - ECC - ECP | MPL - ECC - ECP | |
| T_A | 0.584 - 0.938 - 0.941 | 0.672 - 0.994 - 0.988 | 0.669 - 0.894 - 0.856 | 0.486 - 0.896 - 0.886 |
| T_B | 0.794 - 0.917 - 0.922 | 0.88 - 0.992 - 0.992 | 0.54 - 0.943 - 0.962 | 0.568 - 0.921 - 0.940 |
| T_C | 0.723 - 0.947 - 0.904 | 0.7 - 0.884 - 0.859 | 0.605 - 0.811 - 0.699 | 0.505 - 0.842 - 0.755 |
| T_D | 0.811 - 0.960 - 0.933 | 0.899 - 0.986 - 0.985 | 0.644 - 0.802 - 0.807 | 0.613 - 0.862 - 0.874 |
| T_E | 0.732 - 0.941 - 0.940 | 0.644 - 0.867 - 0.869 | 0.593 - 0.806 - 0.688 | 0.511 - 0.842 - 0.719 |
| T_F | 0.738 - 0.655 - 0.933 | 0.644 - 0.619 - 0.830 | 0.73 - 0.709 - 0.850 | 0.511 - 0.578 - 0.824 |
| T_G | 0.771 - 0.788 - 0.858 | 0.782 - 0.791 - 0.904 | 0.675 - 0.614 - 0.609 | 0.599 - 0.673 - 0.659 |
| T_H | 0.710 - 0.651 - 0.885 | 0.682 - 0.747 - 0.955 | 0.628 - 0.695 - 0.891 | 0.555 - 0.659 - 0.845 |
| T_I | 0.733 - 0.788 - 0.737 | 0.758 - 0.716 - 0.679 | 0.540 - 0.693 - 0.713 | 0.548 - 0.716 - 0.493 |
| T_J | 0.727 - 0.642 - 0.767 | 0.535 - 0.678 - 0.857 | 0.602 - 0.808 - 0.868 | 0.449 - 0.507 - 0.699 |
| T_K | 0.510 - 0.872 - 0.770 | 0.570 - 0.784 - 0.816 | 0.502 - 0.823 - 0.877 | 0.404 - 0.594 - 0.635 |
| T_N | 0.493 - 0.457 - 0.570 | 0.512 - 0.658 - 0.632 | 0.577 - 0.507 - 0.760 | 0.342 - 0.462 - 0.370 |
| T_O | 0.948 - 0.830 - 0.840 | 0.788 - 0.602 - 0.754 | 0.532 - 0.484 - 0.598 | 0.550 - 0.431 - 0.615 |