Figure 2 | Scientific Reports

Figure 2

From: Predicting long-term time to cardiovascular incidents using myocardial perfusion imaging and deep convolutional neural networks

Figure 2

MPI AI model performance in the training/cross-validation stage (n = 1540). (A) The receiver operating characteristic curves (ROC) of 5 models generated in the fivefold analysis, for classifying patients with or without events regardless of the time of event occurrence. (B) The ROC of detecting events occurred within the first 1000 days of observation. (C) The ROC of detecting events occurred within the first 2000 days of observation. (D) The ROC of detecting events occurred within the first 3000 days of observation. (E) The ROCs after the patients are divided into three subgroups with different risk levels, using the median risk score and the threshold that separate quartiles 3 and 4. (F) The Kaplan–Meier plots of patient strata, separated into three different risk subgroups.

Back to article page