Figure 5.
Study population. After identifying labeled pairs, we manually under-sampled the pairs to balance the numbers. In this test, the stable group was much larger than the aggravation group, so pairs in the stable group were under-sampled to match the numbers better. We used these labeled pairs for supervised learning; the final numbers are 15,422 for the stable group and 15,388 for the aggravation group. The characteristics of the study population are shown in Table S2.