Skip to main content

Table 3 Median (25% − 75% percentile) DSC and HD for MAASTRO cohort via baseline and facility-specific models.

From: Comparison of deep learning networks for fully automated head and neck tumor delineation on multi-centric PET/CT images

 

Regular Unet

2D Retina Unet

3D Retina Unet

DSC

HDavg (mm)

HD95 (mm)

DSC

HDavg (mm)

HD95 (mm)

DSC

HDavg (mm)

HD95 (mm)

Baseline model

0.62(0.52–0.71)

0.60 (0.22–0.66)

0.57 (0.10–0.68)

3.9 (2.7–6.1)

5.1 (2.8-19.68)

6.0 (2.9–20.6)

13.4 (6.8–20.0)

18.0 (6.8–46.3)

20.3 (7.4–57.2)

Facility-specific model

0.70 (0.56–0.75)

0.68 (0.59–0.78)

0.66 (0.46–0.75)

2.8 (1.9–3.9)

4.9 (2.3–8.1)

4.8 (2.4–12.2)

7.8 (4.7–12.7)

17.8 (8.2–43.1)

16.5 (5.8–34.6)