Response to “Comments on ‘Novel real-time tumor-contouring method using deep learning to prevent mistracking in X-ray fluoroscopy”’
To the editor,
First, we wish to discuss their suggestion on the requirement for using actual fluoroscopic images in the test stage. The necessity to validate our method using clinical fluoroscopic imaging was already mentioned in “Abstract”, “Discussion”, and “Conclusion” . The sentences in “Discussion” include the following: “we understand that our results were obtained from preliminary simulated fluoroscopic images, and we must validate this method using real clinical fluoroscopy. The anticipated primary difficulty is the different image qualities between the DRRs and the clinical fluoroscopy images. However, we expect that this problem can be solved by improving the DRR quality to be similar to the quality of clinical fluoroscopy images, or by creating a wide contrast variation in the training images for the input data set of deep learning” . The last sentence might be too...
Compliance with ethical standards
Conflict of interest
The authors declare that they have no conflicts of interest.
This article does not contain any studies performed on human participants and animals.
- 1.Mori S, Endo M Comments on “Novel real-time tumor-contouring method using deep learning to prevent mistracking in X-ray fluoroscopy” by Terunuma et al. Radiol Phys Technol 2018. https://doi.org/10.1007/s12194-018-0447-4.
- 4.Zhong Z, Zheng L, Kang G, et al. Random Erasing Data Augmentation. 2017Google Scholar
- 5.Vries TD, Taylor GW. Improved regularization of convolutional neural networks with cutout. 2017Google Scholar