Purpose: CBCT-based online adaptive radiotherapy (ART) calls for accurate auto-segmentation models to reduce the time cost for physicians to edit contours. However, auto-segmentation of CBCT images is a difficult task, majorly due to low image quality and lack of true labels for training a deep learning model. Meanwhile CBCT auto-segmentation in ART is a unique task compared to other segmentation problems, where manual contours on planning CT (pCT) are available.
Methods: To make use of this prior knowledge, we propose to combine deformable image registration (DIR) and direct segmentation (DS) on CBCT for head and neck patients. First, we use dpCT contours derived from multiple DIR methods between pCT and CBCT as pseudo labels for training. Second, we use deformed pCT (dpCT) contours as bounding box to constrain the region of interest for DS. Meanwhile dpCT contours are used as pseudo labels for training, but are generated from different DIR algorithms from bounding box. Third, we fine-tune the model with bounding box on true labels. We try to compete against DIR-only auto-segmentation.
Results: We found that DS on CBCT trained with pseudo labels and without utilizing any prior knowledge for training has very poor segmentation performance compared to DIR-only segmentation. However, adding dpCT contours as bounding box in the DS network can dramatically improve segmentation performance, comparable to DIR-only segmentation. The DS model with bounding box can be further improved by fine-tuning it with some real labels instead of using pseudo labels for training. Experiments showed that 7 out of 19 structures have at least 0.2 dice similarity coefficient increase compared to DIR-only segmentation.
Conclusion: Utilizing dpCT contours as pseudo labels for training and as bounding box for shape and location feature extraction in a DS model is a good way to combine DIR and DS.