Table 2.
Vessel-level analysis of end-diastolic frame estimations of the 1st- and 2nd-analyst and of the CIB- and DL-methodologies in the NIRS-IVUS sequences acquired at 30fps
| Studied vessel | Segmentation approach | End-diastolic frames corresponding to ECG estimations | End-diastolic frames not corresponding to ECG estimations | Missed cardiac cycles | P |
|---|---|---|---|---|---|
| All studied vessels | 1st-analyst 1st estimation | 3034 (92.6%) | 110 (3.4%) | 133 (4.1%) | |
| 1st-analyst 2nd estimation | 3014 (93.1%) | 72 (2.2%) | 153 (4.7%) | ||
| 2nd-analyst | 3069 (93.5%) | 117 (3.6%) | 98 (3.0%) | < 0.001 | |
| CIB-methodology | 3052 (93.5%) | 96 (2.9%) | 115 (3.5%) | ||
| DL-methodology | 2987 (91.6%) | 95 (2.9%) | 180 (5.5%) | ||
| Left anterior descending artery | 1st-analyst 1st estimation | 1141 (94.7%) | 23 (1.9%) | 42 (3.4%) | |
| 1st-analyst 2nd estimation | 1136 (94.4%) | 22 (1.8%) | 46 (3.8%) | ||
| 2nd-analyst | 1140 (94.7%) | 22 (1.8%) | 42 (3.5%) | < 0.001 | |
| CIB-methodology | 1119 (93.9%) | 10 (0.8%) | 63 (5.3%) | ||
| DL-methodology | 1062 (88.8%) | 14 (1.2%) | 120 (10.0%) | ||
| Left circumflex artery | 1st-analyst 1st estimation | 976 (92.8%) | 41 (3.9%) | 35 (3.3%) | |
| 1st-analyst 2nd estimation | 968 (94.3%) | 16 (1.6%) | 43 (4.2%) | ||
| 2nd-analyst | 976 (92.6%) | 43 (4.1%) | 35 (3.3%) | < 0.001 | |
| CIB-methodology | 972 (94.0%) | 23 (2.2%) | 39 (3.8%) | ||
| DL-methodology | 971 (94.5%) | 16 (1.6%) | 40 (3.9%) | ||
| Right coronary artery | 1st-analyst 1st estimation | 917 (89.9%) | 46 (4.5%) | 57 (5.6%) | |
| 1st-analyst 2nd estimation | 910 (90.3%) | 34 (3.4%) | 64 (6.3%) | ||
| 2nd-analyst | 953 (92.9%) | 52 (5.1%) | 21 (2.0%) | < 0.001 | |
| CIB-methodology | 961 (92.7%) | 63 (6.1%) | 13 (1.3%) | ||
| DL-methodology | 954 (91.8%) | 65 (6.3%) | 20 (1.9%) |
CIB conventional image-based, DL deep learning, ECG electrocardiogram