Skip to main content
. 2023 Mar 21;10:1143986. doi: 10.3389/fvets.2023.1143986

Figure 6.

Figure 6

Manual ground truth and automatic deep learning generated gross tumor volume contours in four CT image slices from one canine test set patient (nasal cavity tumor). (Left column) Manual ground truth contours (green). (Middle column) Auto-segmentation generated by model S9 (magenta; model trained from scratch on canine data only). (Right column) Auto-segmentation generated by model T2 (blue; model trained using transfer learning). The two models resulted in Sørensen-Dice similarity coefficients of 0.72 (model S9) and 0.85 (model T2) for the given patient (calculated over all 173 image slices).