Table 2 Predictive performance metrics in ablation experiments
From: Development of an AI model for DILI-level prediction using liver organoid brightfield images
Metrics | HLO without spatial | HLO using D0–D1 images | HLO | |
|---|---|---|---|---|
Recall | vNo-DILI-Concern | 0.9016 | 0.7049 | 0.9016 |
vLess-DILI-Concern | 0.6200 | 0.7500 | 0.7500 | |
vMost-DILI-Concern | 0.6986 | 0.6849 | 0.7808 | |
Specificity | vNo-DILI-Concern | 0.7835 | 0.7835 | 0.8557 |
vLess-DILI-Concern | 0.9626 | 0.9043 | 0.9552 | |
vMost-DILI-Concern | 0.8706 | 0.7882 | 0.9059 | |
Precision | vNo-DILI-Concern | 0.7237 | 0.6719 | 0.7971 |
vLess-DILI-Concern | 0.7500 | 0.6923 | 0.7500 | |
vMost-DILI-Concern | 0.8095 | 0.7353 | 0.8769 | |
F1 score | vNo-DILI-Concern | 0.8029 | 0.6880 | 0.8462 |
vLess-DILI-Concern | 0.6818 | 0.7199 | 0.7500 | |
vMost-DILI-Concern | 0.7500 | 0.7092 | 0.8261 | |
Accuracy | Â | 0.7610 | 0.7025 | 0.8234 |
