Request for Code Access – Cyclical Self-Training for Lung Registration

[niusha-gh27]

Hello,
I am a master's student of Biomedical Engineering in Tehran University of biomedical sciences, currently working on lung keypoint registration for my thesis. I recently came across your impressive work on cyclical self-training for lung registration and found it highly relevant to my research.

However, I noticed that the code for the keypoint registration part of your article is not publicly available on your GitHub repository. I would be extremely grateful if you could share the code with me, as I am very interested in testing and incorporating your method into my thesis project. With your permission, I would be glad to cite your work appropriately in my final publication.

If sharing the code is possible, please feel free to contact me at: n-ghadesi@razi.tums.ac.ir
Thank you very much for your time and contribution to the field.

Best regards,
Niuosha Ghadesi
MSc Student, Biomedical Engineering

[mattiaspaul]

Hi Niuoasha, you can find the keypoint registration code here: https://github.com/multimodallearning/Lung250M-4B/blob/main/corrfield/corrfield.py or https://github.com/multimodallearning/deep-geo-reg/blob/main/eval.ipynb
Hope it helps. Mattias

[niusha-gh27]

Thank you very much for your response. I’ve already read the articles you shared, and while they are indeed valuable, I believe they focus on supervised approaches. In contrast, the point registration method described in your article *“Unsupervised 3D Registration through Optimization-Guided Cyclical Self-Training”* is unsupervised, which is exactly what I’m looking for. I’m particularly interested in experimenting with this method for my thesis, so I would greatly appreciate it if the code for that work could be shared. Additionally, I have a question regarding *GraphRegNet*, which I find to be an excellent contribution, especially given its strong TRE performance. I understand that the paper describes a two-stage registration process using different sets of keypoints. I was able to run the code for a single stage, but the results were not as promising. I suspect that implementing the second stage, as mentioned in the paper, might improve the performance. However, I couldn’t find clear documentation or code for the second stage either in the paper or the repository. I would be very grateful if you could provide some guidance on how to implement or access the second stage of GraphRegNet. Thank you again for your time and support. Best regards, Niuosha Ghadesi MSc Student, Biomedical Engineering

…

On Sun, Jul 27, 2025 at 9:08 PM mattiaspaul ***@***.***> wrote: *mattiaspaul* left a comment (multimodallearning/reg-cyclical-self-train#1) <#1 (comment)> Hi Niuoasha, you can find the keypoint registration code here: https://github.com/multimodallearning/Lung250M-4B/blob/main/corrfield/corrfield.py or https://github.com/multimodallearning/deep-geo-reg/blob/main/eval.ipynb Hope it helps. Mattias — Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BGYLRQKC3YMWLEVMKQZEZFL3KUFC7AVCNFSM6AAAAACCO7U4RSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTCMRUGU3DAMZYGI> . You are receiving this because you authored the thread.Message ID: ***@***.*** com>