Fig. 5
From: Multi-view secondary input collaborative deep learning for lung nodule 3D segmentation
Comparison of model structures and different inputs. Columns from left to right are the original CT image, radiologist marker image, MV-SIR result, and 2D segmentation effect map. The top to bottom rows sequentially present the results of the MV-CNN, MV-I-CNN, the MV-SIR with VH input, the MV-SIR with SH input, the MV-SIR with combined VH and SH input, and secondary input MV-SIR. The 2D segmentation map can intuitively show that our model has achieved the best segmentation effect