![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1025985%2F945d7dacb946c55154746ec39a9280bd%2F2019-09-20%2011.17.35.png?generation=1568949525391202&alt=media\")
\n We built a CT system in our lab when I was in college. The left projection is something I scanned. And the right image is what it looks like after reconstruction. My master thesis is about research on the calibration method and interior tomography of high-resolution CT.
\n\nThe introduction on CT reconstruction algorithms:\nhttps://en.wikipedia.org/wiki/Tomographic_reconstruction
\n\nCan reconstruction algorithms help in this challenge？\n-- Maybe data augmentation. If we have the complete 3D head data of a patient.
\n\n========================================================================\nSome interesting pics:\n\n We built a CT system in our lab when I was in college. The left projection is something I scanned. And the right image is what it looks like after reconstruction.
===This is a toothpick！===
", "messages": [ { "id": 629766, "postDate": "2019-09-19T06:38:01.283Z", "content": "My master thesis is about research on the calibration method and interior tomography of high-resolution CT.
\n\nThe introduction on CT reconstruction algorithms:\nhttps://en.wikipedia.org/wiki/Tomographic_reconstruction
\n\nCan reconstruction algorithms help in this challenge？\n-- Maybe data augmentation. If we have the complete 3D head data of a patient.
\n\n========================================================================\nSome interesting pics:\n\n We built a CT system in our lab when I was in college. The left projection is something I scanned. And the right image is what it looks like after reconstruction.
===This is a toothpick！===
", "rawMarkdown": "My master thesis is about research on the calibration method and interior tomography of high-resolution CT. \n\nThe introduction on CT reconstruction algorithms:\nhttps://en.wikipedia.org/wiki/Tomographic_reconstruction\n\n1. Fourier-Domain Reconstruction Algorithm\n2. Back Projection Algorithm\n3. Iterative Reconstruction Algorithm\n\nCan reconstruction algorithms help in this challenge？\n-- Maybe data augmentation. If we have the complete 3D head data of a patient.\n\n========================================================================\nSome interesting pics:\n\n We built a CT system in our lab when I was in college. The left projection is something I scanned. And the right image is what it looks like after reconstruction. \n\n===This is a toothpick！===", "votes": 15 }, { "id": 630132, "postDate": "2019-09-19T19:23:06.387Z", "content": "Your master thesis is looking extremely interesting! The main problem with the reconstruction algorithms that you mention is they are usually applied to the raw machine data. In this competition the dicom files are already processed or reconstructed by the different CT machines into humanly readable axial images. Consequently, we don't have access to the raw data. Even in research we don't easily have access to the raw machine data because it's usually walled in the CT machine software.
\n\nBut applying specific reconstruction algorithms to make the physical machine data more readable by deep learning models is definitely an interesting and promising area of research.
", "rawMarkdown": "Your master thesis is looking extremely interesting! The main problem with the reconstruction algorithms that you mention is they are usually applied to the raw machine data. In this competition the dicom files are already processed or reconstructed by the different CT machines into humanly readable axial images. Consequently, we don't have access to the raw data. Even in research we don't easily have access to the raw machine data because it's usually walled in the CT machine software.\n\nBut applying specific reconstruction algorithms to make the physical machine data more readable by deep learning models is definitely an interesting and promising area of research.", "votes": 3, "replies": [ { "id": 630279, "postDate": "2019-09-20T02:54:33.243Z", "content": "Yes! I mean we can forward project the 3D data with different spatial params to create raw projections (in this setting, we can add some noise, artifacts and etc.), and then apply the reconstruction algorithms to get new 3D data --> data augmentation. I think it might be better to do this kind of data augmentation following the CT imaging pipeline.
", "rawMarkdown": "Yes! I mean we can forward project the 3D data with different spatial params to create raw projections (in this setting, we can add some noise, artifacts and etc.), and then apply the reconstruction algorithms to get new 3D data --> data augmentation. I think it might be better to do this kind of data augmentation following the CT imaging pipeline.", "votes": 4 }, { "id": 630291, "postDate": "2019-09-20T03:31:47.173Z", "content": "I think this may be a very interesting approach. Some reconstruction methods have a straightforward inverse, which allows us to play with it. Actually, any kind of transformation to a different space could be tested. I am curious to see if training on these domains could help. My bet is that it does not, but using your approach to do data augmentation looks promising imho.
", "rawMarkdown": "I think this may be a very interesting approach. Some reconstruction methods have a straightforward inverse, which allows us to play with it. Actually, any kind of transformation to a different space could be tested. I am curious to see if training on these domains could help. My bet is that it does not, but using your approach to do data augmentation looks promising imho. ", "votes": 2 } ] }, { "id": 647327, "postDate": "2019-10-12T13:14:57.643Z", "content": "I don't think it will help in this competition.\nit's meaningless.
", "rawMarkdown": "I don't think it will help in this competition.\nit's meaningless." }, { "id": 1627480, "postDate": "2021-12-24T00:17:20.790Z", "content": "Did you try any learning based methods?
", "rawMarkdown": "Did you try any learning based methods?" }, { "id": 651484, "postDate": "2019-10-17T14:30:19.073Z", "content": "I am doing a subject about 3D reconstruction based on mri.Would you mind sharing your paper with me ?Thank you very much.
", "rawMarkdown": "I am doing a subject about 3D reconstruction based on mri.Would you mind sharing your paper with me ?Thank you very much." } ], "comments": [ { "id": 630132, "author_name": "Alexandre Cadrin-Chênevert", "author_url": "", "post_date": "2019-09-19T19:23:06.387000", "content": "Your master thesis is looking extremely interesting! The main problem with the reconstruction algorithms that you mention is they are usually applied to the raw machine data. In this competition the dicom files are already processed or reconstructed by the different CT machines into humanly readable axial images. Consequently, we don't have access to the raw data. Even in research we don't easily have access to the raw machine data because it's usually walled in the CT machine software.
\n\nBut applying specific reconstruction algorithms to make the physical machine data more readable by deep learning models is definitely an interesting and promising area of research.
", "votes": 3, "replies": [ { "id": 630279, "author_name": "SeuTao", "author_url": "", "post_date": "2019-09-20T02:54:33.243000", "content": "Yes! I mean we can forward project the 3D data with different spatial params to create raw projections (in this setting, we can add some noise, artifacts and etc.), and then apply the reconstruction algorithms to get new 3D data --> data augmentation. I think it might be better to do this kind of data augmentation following the CT imaging pipeline.
", "votes": 4, "replies": [] }, { "id": 630291, "author_name": "FelipeKitamura, MD, PhD", "author_url": "", "post_date": "2019-09-20T03:31:47.173000", "content": "I think this may be a very interesting approach. Some reconstruction methods have a straightforward inverse, which allows us to play with it. Actually, any kind of transformation to a different space could be tested. I am curious to see if training on these domains could help. My bet is that it does not, but using your approach to do data augmentation looks promising imho.
", "votes": 2, "replies": [] } ] }, { "id": 647327, "author_name": "pupil3", "author_url": "", "post_date": "2019-10-12T13:14:57.643000", "content": "I don't think it will help in this competition.\nit's meaningless.
", "votes": 0, "replies": [] }, { "id": 1627480, "author_name": "Md Muztaba Ahbab", "author_url": "", "post_date": "2021-12-24T00:17:20.790000", "content": "Did you try any learning based methods?
", "votes": 0, "replies": [] }, { "id": 651484, "author_name": "QingChengYi", "author_url": "", "post_date": "2019-10-17T14:30:19.073000", "content": "I am doing a subject about 3D reconstruction based on mri.Would you mind sharing your paper with me ?Thank you very much.
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "629766": "My master thesis is about research on the calibration method and interior tomography of high-resolution CT. \n\nThe introduction on CT reconstruction algorithms:\nhttps://en.wikipedia.org/wiki/Tomographic_reconstruction\n\n1. Fourier-Domain Reconstruction Algorithm\n2. Back Projection Algorithm\n3. Iterative Reconstruction Algorithm\n\nCan reconstruction algorithms help in this challenge？\n-- Maybe data augmentation. If we have the complete 3D head data of a patient.\n\n========================================================================\nSome interesting pics:\n\n We built a CT system in our lab when I was in college. The left projection is something I scanned. And the right image is what it looks like after reconstruction. \n\n===This is a toothpick！===", "630132": "Your master thesis is looking extremely interesting! The main problem with the reconstruction algorithms that you mention is they are usually applied to the raw machine data. In this competition the dicom files are already processed or reconstructed by the different CT machines into humanly readable axial images. Consequently, we don't have access to the raw data. Even in research we don't easily have access to the raw machine data because it's usually walled in the CT machine software.\n\nBut applying specific reconstruction algorithms to make the physical machine data more readable by deep learning models is definitely an interesting and promising area of research.", "647327": "I don't think it will help in this competition.\nit's meaningless.", "1627480": "Did you try any learning based methods?", "651484": "I am doing a subject about 3D reconstruction based on mri.Would you mind sharing your paper with me ?Thank you very much." } }