Table 2 Reader statistics and inter-observer variability between human radiologist and DL algorithm
From: Deep learning for semi-automated unidirectional measurement of lung tumor size in CT
Reader | Image Use | Number of Annotated Images | Average Measurement (cm) |
|---|---|---|---|
Radiologist 1 | Test | 244 | 2.99 ± 0.93 (1.57–4.91) |
Radiologist 2 | Training | 734 | 3.17 ± 0.96 (1.51–5.00) |
Validation | 159 | 3.21 ± 0.98 (1.50–4.99) | |
Radiologist 3 | Training | 395 | 2.92 ± 0.88 (1.49–4.94) |
Validation | 85 | 3.35 ± 0.89 (1.56–4.79) | |
DL Algorithm | Test | 244 | 3.07 ± 0.91 (1.37–5.44) |
Radiologist 1 & DL Algorithm ICC: 0.959 (95% CI: 0.947, 0.967) | |||