Table 1.
List of studies that used AI on image data to predict or rank embryos based on pregnancy outcome. The reported information only concerns evaluation of pregnancy-related outcomes. Therefore, if a study includes additional tasks such as blastocyst prediction, these are not included in the table
Reference | Input | Outcome | Embryo populationa | Human vs. AIb | Metricsc |
---|---|---|---|---|---|
[20] | Static image | Fetal heartbeat | *-D5-blastocyst | ✓ | Accuracy, AUC |
[21] | Static image, patient age | Beta-HCG | *-D5/D6-blastocyst | - | Accuracy, sensitivity, specificity, PPV, NPV, FPR, FNR, F1, AUC |
[13] | Static image, patient age, blastocyst age, lab settings | Ploidy/beta-HCG | *-D5/D6-blastocyst | ✓ | Accuracy, sensitivity, specificity, PPV, AUC, NDCG |
[22] | Time-lapse video | Fetal heartbeat | ICSI-D3-*, ICSI-D5-* | Sensitivity, PPV, AUC | |
[11] | Time-lapse video | Fetal heartbeat2 | *-D5-* | - | AUC |
[12] | Static image | Fetal heartbeat | *-D5-blastocyst | ✓ | Accuracy, sensitivity, specificity |
[23] | Static image | “pregnancy” | *-*-blastocyst | - | Accuracy, sensitivity, specificity |
[24] | Static image | Live birth | *-D5/D6-blastocyst | Accuracy, sensitivity, specificity, AUC | |
[25] | Static image | Live birth | *-D5-blastocyst1 | - | Accuracy, sensitivity, specificity, PPV, NPV, AUC |
[26] | Static image, annotations, patient info (age, BMI, ...) | Live birth | *-D5/D6-blastocyst | Accuracy, sensitivity, specificity, informedness, AUC | |
[27] | Static image, annotations, patient info (age, BMI, ...) | Live birth | *-D5/D6-blastocyst | - | Accuracy, sensitivity, specificity, informedness, PPV, NPV, AUC |
[14] | Time-lapse video | Fetal heartbeat | *-*-* | ✓ | PPV, NPV, AUC |
[15] | Time-lapse video | Fetal heartbeat | *-D5/D6-* | AUC |
a The notation for Embryo population is explained in “Data foundation” and visualized in Fig. 1
b Human vs. AI comparisons are discussed in “Bias in model comparisons”
c All metrics are explained in detail in “Evaluation metrics: which performance measure to use?”
1 Only aneuploid miscarriages (confirmed with genetic testing of chorionic villus samples) were included as negative live births
2 Negative fetal heartbeat was assumed for all non-transferred embryos that had “failed or abnormal fertilization, grossly abnormal morphology or aneuploidy from preimplantation genetic testing”