Thank you for releasing the DICDNet code‑base — it has been very helpful for my research.
While working through the repo I noticed that the paper validates the model on two clinical data sets that contain metal (SpineWeb and CLINIC‑metal, though I am mostly interested in the CLINIC-metal images for my research.) but I could not find the corresponding script(s) for running inference on clinical images.
Because those data do not provide ground‑truth metal‑free CT, the test pipeline needs to be different from the one used for the paired synthetic data (e.g. no PSNR/SSIM computation, optional saving of intermediate outputs, etc.). Without such a script it’s impossible (for me) to reproduce the clinical‑validation part of the paper.
Could you please consider:
adding the inference script (or notebook) that you used for the CLINIC‑metal experiments;
briefly documenting any expected folder structure / preprocessing steps (mask threshold, spacing, windowing) that differ from the synthetic pipeline;
indicating how you visualized results in Fig. 8 and Fig. 9 of the paper for fair comparison.
Having these files would let the community fully replicate the work and build on it.
Thanks again for the great contribution, and looking forward to your reply!
Thank you for releasing the DICDNet code‑base — it has been very helpful for my research.
While working through the repo I noticed that the paper validates the model on two clinical data sets that contain metal (SpineWeb and CLINIC‑metal, though I am mostly interested in the CLINIC-metal images for my research.) but I could not find the corresponding script(s) for running inference on clinical images.
Because those data do not provide ground‑truth metal‑free CT, the test pipeline needs to be different from the one used for the paired synthetic data (e.g. no PSNR/SSIM computation, optional saving of intermediate outputs, etc.). Without such a script it’s impossible (for me) to reproduce the clinical‑validation part of the paper.
Could you please consider:
adding the inference script (or notebook) that you used for the CLINIC‑metal experiments;
briefly documenting any expected folder structure / preprocessing steps (mask threshold, spacing, windowing) that differ from the synthetic pipeline;
indicating how you visualized results in Fig. 8 and Fig. 9 of the paper for fair comparison.
Having these files would let the community fully replicate the work and build on it.
Thanks again for the great contribution, and looking forward to your reply!