Hei!
\n\nI browsed through topics and I haven't found this resource, so I will leave it here and apologies if it has been already posted.
\n\n\n \n\n\n
It is a comprehensive analysis on DICOM images that has a useful function that applies a mask on the lungs, so out of the full CT image you remain only with the lungs. Might be useful 😎
\n\nCheers and good luck!
", "messages": [ { "id": 943922, "postDate": "2020-07-24T17:13:30.757Z", "content": "Hei!
\n\nI browsed through topics and I haven't found this resource, so I will leave it here and apologies if it has been already posted.
\n\n\n \n\n\n
It is a comprehensive analysis on DICOM images that has a useful function that applies a mask on the lungs, so out of the full CT image you remain only with the lungs. Might be useful 😎
\n\nCheers and good luck!
", "rawMarkdown": "Hei!\n\nI browsed through topics and I haven't found this resource, so I will leave it here and apologies if it has been already posted.\n\n> [dicom-processing-segmentation-visualization-in-python](https://www.raddq.com/dicom-processing-segmentation-visualization-in-python/)\n\nIt is a comprehensive analysis on DICOM images that has a useful function that applies a mask on the lungs, so out of the full CT image you remain only with the lungs. Might be useful 😎\n\nCheers and good luck!", "votes": 53 }, { "id": 944415, "postDate": "2020-07-25T05:27:23.967Z", "content": "Hi there!
\n\nI found out that Marker-Controlled Watershed transformation works really well on medical images and preserves much of the information.
\n\nHere's more about Watershed Transformation:\n- CMM\n- Mathworks Article
\n\nI made a kernel which segments lung images using the above approach, feel free to check it out here:
\n\n\n\n\nLung Segmentation using Marker-Controlled Watershed Transformation
\n
I would love someone to make use of my kernel in their models. Good luck!
\n\nIf you consider this as an advertisement, I'll kindly remove this comment :)
", "rawMarkdown": "Hi there! \n\nI found out that **Marker-Controlled Watershed transformation** works really well on medical images and preserves much of the information.\n\nHere's more about Watershed Transformation:\n- [CMM](http://www.cmm.mines-paristech.fr/~beucher/wtshed.html)\n- [Mathworks Article](https://www.mathworks.com/company/newsletters/articles/the-watershed-transform-strategies-for-image-segmentation.html)\n\nI made a kernel which segments lung images using the above approach, feel free to check it out here:\n\n> [Lung Segmentation using Marker-Controlled Watershed Transformation](https://www.kaggle.com/aadhavvignesh/lung-segmentation-by-marker-controlled-watershed)\n\nI would love someone to make use of my kernel in their models. Good luck!\n\nIf you consider this as an advertisement, I'll kindly remove this comment :)", "votes": 9, "replies": [ { "id": 945329, "postDate": "2020-07-25T18:30:59.640Z", "content": "Very interesting approach! Thanks for sharing. 😄
\n\n(Indeed, it is hard sometimes to recommend your work without making it look like you're trying to get yourself an upvote; however, sharing is caring, especially when it's such a well documented notebook like this one)
", "rawMarkdown": "Very interesting approach! Thanks for sharing. 😄\n\n(Indeed, it is hard sometimes to recommend your work without making it look like you're trying to get yourself an upvote; however, sharing is caring, especially when it's such a well documented notebook like this one)", "votes": 2 }, { "id": 947274, "postDate": "2020-07-27T06:56:28.600Z", "rawMarkdown": "", "votes": 1, "isDeleted": true } ] }, { "id": 1006359, "postDate": "2020-09-11T07:51:53.487Z", "content": "Hi @andradaolteanu , many thanks for sharing this material - it's very useful! By the way, I enjoy your notebooks very much. Keep on the good work and best of luck in this and future competitions!
", "rawMarkdown": "Hi @andradaolteanu , many thanks for sharing this material - it's very useful! By the way, I enjoy your notebooks very much. Keep on the good work and best of luck in this and future competitions!", "votes": 1 }, { "id": 967283, "postDate": "2020-08-12T06:12:29.740Z", "content": "Extracting metadata or as done by @aadhavvignesh the HU converted images can also be used for modeling purposes
", "rawMarkdown": "Extracting metadata or as done by @aadhavvignesh the HU converted images can also be used for modeling purposes", "votes": 1 }, { "id": 949908, "postDate": "2020-07-29T04:12:40.293Z", "content": "Manually, here are some nice ways to view the segmentation from Guido Zuidhof's notebook: https://www.kaggle.com/gzuidhof/full-preprocessing-tutorial
\n\nIt also might be worth looking into past works from the LUNA16 competition - not sure how well this manual segmentation method stacks up against what you found, it would be worth comparing both.
", "rawMarkdown": "Manually, here are some nice ways to view the segmentation from Guido Zuidhof's notebook: https://www.kaggle.com/gzuidhof/full-preprocessing-tutorial\n\nIt also might be worth looking into past works from the LUNA16 competition - not sure how well this manual segmentation method stacks up against what you found, it would be worth comparing both.", "votes": 2 }, { "id": 1006622, "postDate": "2020-09-11T12:25:46.937Z", "content": "Thanks for sharing this excellent material :)
", "rawMarkdown": "Thanks for sharing this excellent material :)" }, { "id": 986983, "postDate": "2020-08-26T22:26:56.683Z", "content": "does masking lungs help accuracy much? I thought a CNN can easily learn the difference between the lungs and anything else. And unless there is some information that can help the prediction it can easily learn to ignore it. My hypothesis is that masking lungs can speed up convergence but not much else. Please feel free to correct me on this one.
", "rawMarkdown": "does masking lungs help accuracy much? I thought a CNN can easily learn the difference between the lungs and anything else. And unless there is some information that can help the prediction it can easily learn to ignore it. My hypothesis is that masking lungs can speed up *convergence* but not much else. Please feel free to correct me on this one." }, { "id": 966431, "postDate": "2020-08-11T12:18:01.337Z", "content": "This is great. Thank you!
", "rawMarkdown": "This is great. Thank you!" }, { "id": 950823, "postDate": "2020-07-29T16:41:23.100Z", "content": "Thanks for sharing!!
", "rawMarkdown": "Thanks for sharing!!" }, { "id": 945663, "postDate": "2020-07-26T04:26:11.163Z", "content": "Very Useful, Thanks a lot for sharing!
", "rawMarkdown": "Very Useful, Thanks a lot for sharing!" } ], "comments": [ { "id": 944415, "author_name": "Aadhav Vignesh", "author_url": "", "post_date": "2020-07-25T05:27:23.967000", "content": "Hi there!
\n\nI found out that Marker-Controlled Watershed transformation works really well on medical images and preserves much of the information.
\n\nHere's more about Watershed Transformation:\n- CMM\n- Mathworks Article
\n\nI made a kernel which segments lung images using the above approach, feel free to check it out here:
\n\n\n\n\nLung Segmentation using Marker-Controlled Watershed Transformation
\n
I would love someone to make use of my kernel in their models. Good luck!
\n\nIf you consider this as an advertisement, I'll kindly remove this comment :)
", "votes": 9, "replies": [ { "id": 945329, "author_name": "Andrada", "author_url": "", "post_date": "2020-07-25T18:30:59.640000", "content": "Very interesting approach! Thanks for sharing. 😄
\n\n(Indeed, it is hard sometimes to recommend your work without making it look like you're trying to get yourself an upvote; however, sharing is caring, especially when it's such a well documented notebook like this one)
", "votes": 2, "replies": [] }, { "id": 947274, "author_name": "", "author_url": "", "post_date": "2020-07-27T06:56:28.600000", "content": "", "votes": 1, "replies": [] } ] }, { "id": 1006359, "author_name": "Douglas K.G. Araujo", "author_url": "", "post_date": "2020-09-11T07:51:53.487000", "content": "Hi @andradaolteanu , many thanks for sharing this material - it's very useful! By the way, I enjoy your notebooks very much. Keep on the good work and best of luck in this and future competitions!
", "votes": 1, "replies": [] }, { "id": 967283, "author_name": "Digvijay Yadav", "author_url": "", "post_date": "2020-08-12T06:12:29.740000", "content": "Extracting metadata or as done by @aadhavvignesh the HU converted images can also be used for modeling purposes
", "votes": 1, "replies": [] }, { "id": 949908, "author_name": "Trigram", "author_url": "", "post_date": "2020-07-29T04:12:40.293000", "content": "Manually, here are some nice ways to view the segmentation from Guido Zuidhof's notebook: https://www.kaggle.com/gzuidhof/full-preprocessing-tutorial
\n\nIt also might be worth looking into past works from the LUNA16 competition - not sure how well this manual segmentation method stacks up against what you found, it would be worth comparing both.
", "votes": 2, "replies": [] }, { "id": 1006622, "author_name": "Salman Ibne Eunus", "author_url": "", "post_date": "2020-09-11T12:25:46.937000", "content": "Thanks for sharing this excellent material :)
", "votes": 0, "replies": [] }, { "id": 986983, "author_name": "DarkCube", "author_url": "", "post_date": "2020-08-26T22:26:56.683000", "content": "does masking lungs help accuracy much? I thought a CNN can easily learn the difference between the lungs and anything else. And unless there is some information that can help the prediction it can easily learn to ignore it. My hypothesis is that masking lungs can speed up convergence but not much else. Please feel free to correct me on this one.
", "votes": 0, "replies": [] }, { "id": 966431, "author_name": "Sagnik", "author_url": "", "post_date": "2020-08-11T12:18:01.337000", "content": "This is great. Thank you!
", "votes": 0, "replies": [] }, { "id": 950823, "author_name": "Digvijay Yadav", "author_url": "", "post_date": "2020-07-29T16:41:23.100000", "content": "Thanks for sharing!!
", "votes": 0, "replies": [] }, { "id": 945663, "author_name": "Salman Ibne Eunus", "author_url": "", "post_date": "2020-07-26T04:26:11.163000", "content": "Very Useful, Thanks a lot for sharing!
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "943922": "Hei!\n\nI browsed through topics and I haven't found this resource, so I will leave it here and apologies if it has been already posted.\n\n> [dicom-processing-segmentation-visualization-in-python](https://www.raddq.com/dicom-processing-segmentation-visualization-in-python/)\n\nIt is a comprehensive analysis on DICOM images that has a useful function that applies a mask on the lungs, so out of the full CT image you remain only with the lungs. Might be useful 😎\n\nCheers and good luck!", "944415": "Hi there! \n\nI found out that **Marker-Controlled Watershed transformation** works really well on medical images and preserves much of the information.\n\nHere's more about Watershed Transformation:\n- [CMM](http://www.cmm.mines-paristech.fr/~beucher/wtshed.html)\n- [Mathworks Article](https://www.mathworks.com/company/newsletters/articles/the-watershed-transform-strategies-for-image-segmentation.html)\n\nI made a kernel which segments lung images using the above approach, feel free to check it out here:\n\n> [Lung Segmentation using Marker-Controlled Watershed Transformation](https://www.kaggle.com/aadhavvignesh/lung-segmentation-by-marker-controlled-watershed)\n\nI would love someone to make use of my kernel in their models. Good luck!\n\nIf you consider this as an advertisement, I'll kindly remove this comment :)", "1006359": "Hi @andradaolteanu , many thanks for sharing this material - it's very useful! By the way, I enjoy your notebooks very much. Keep on the good work and best of luck in this and future competitions!", "967283": "Extracting metadata or as done by @aadhavvignesh the HU converted images can also be used for modeling purposes", "949908": "Manually, here are some nice ways to view the segmentation from Guido Zuidhof's notebook: https://www.kaggle.com/gzuidhof/full-preprocessing-tutorial\n\nIt also might be worth looking into past works from the LUNA16 competition - not sure how well this manual segmentation method stacks up against what you found, it would be worth comparing both.", "1006622": "Thanks for sharing this excellent material :)", "986983": "does masking lungs help accuracy much? I thought a CNN can easily learn the difference between the lungs and anything else. And unless there is some information that can help the prediction it can easily learn to ignore it. My hypothesis is that masking lungs can speed up *convergence* but not much else. Please feel free to correct me on this one.", "966431": "This is great. Thank you!", "950823": "Thanks for sharing!!", "945663": "Very Useful, Thanks a lot for sharing!" } }