{ "id": 524194, "title": "Medical SAM 2 - (Segment Anything Model 2) - is allowed to use in this competition?", "url": "/competitions/rsna-2024-lumbar-spine-degenerative-classification/discussion/524194", "author_name": "", "post_date": "2024-08-05T04:09:33.914591300Z", "votes": 4, "comment_count": 7, "views": 0, "content": "

Medical SAM 2: Segment Medical Images As Video Via Segment Anything Model 2

\n

Paper

\n

Code

\n

\"\"

\n
\n

In this paper, we introduce Medical SAM 2 (MedSAM-2), an advanced segmentation model that utilizes the SAM 2 framework to address both 2D and 3D medical image segmentation tasks. By adopting the philosophy of taking medical images as videos, MedSAM-2 not only applies to 3D medical images but also unlocks new One-prompt Segmentation capability. That allows users to provide a prompt for just one or a specific image targeting an object, after which the model can autonomously segment the same type of object in all subsequent images, regardless of temporal relationships between the images. We evaluated MedSAM-2 across a variety of medical imaging modalities, including abdominal organs, optic discs, brain tumors, thyroid nodules, and skin lesions, comparing it against state-of-the-art models in both traditional and interactive segmentation settings. Our findings show that MedSAM-2 not only surpasses existing models in performance but also exhibits superior generalization across a range of medical image segmentation tasks. Our code will be released at: https://github.com/MedicineToken/Medical-SAM2

\n
", "messages": [ { "id": "2947219", "postDate": "08/05/2024 04:09:33", "content": "

Medical SAM 2: Segment Medical Images As Video Via Segment Anything Model 2

\n

Paper

\n

Code

\n

\"\"

\n
\n

In this paper, we introduce Medical SAM 2 (MedSAM-2), an advanced segmentation model that utilizes the SAM 2 framework to address both 2D and 3D medical image segmentation tasks. By adopting the philosophy of taking medical images as videos, MedSAM-2 not only applies to 3D medical images but also unlocks new One-prompt Segmentation capability. That allows users to provide a prompt for just one or a specific image targeting an object, after which the model can autonomously segment the same type of object in all subsequent images, regardless of temporal relationships between the images. We evaluated MedSAM-2 across a variety of medical imaging modalities, including abdominal organs, optic discs, brain tumors, thyroid nodules, and skin lesions, comparing it against state-of-the-art models in both traditional and interactive segmentation settings. Our findings show that MedSAM-2 not only surpasses existing models in performance but also exhibits superior generalization across a range of medical image segmentation tasks. Our code will be released at: https://github.com/MedicineToken/Medical-SAM2

\n
", "rawMarkdown": "# Medical SAM 2: Segment Medical Images As Video Via Segment Anything Model 2\n### [Paper](https://arxiv.org/abs/2408.00874)\n### [Code](https://github.com/MedicineToken/Medical-SAM2)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2F818ebc047b23541a0594187b373e7611%2Fposter1.png?generation=1722830955559500&alt=media)\n\n> In this paper, we introduce Medical SAM 2 (MedSAM-2), an advanced segmentation model that utilizes the SAM 2 framework to address both 2D and 3D medical image segmentation tasks. By adopting the philosophy of taking medical images as videos, MedSAM-2 not only applies to 3D medical images but also unlocks new One-prompt Segmentation capability. That allows users to provide a prompt for just one or a specific image targeting an object, after which the model can autonomously segment the same type of object in all subsequent images, regardless of temporal relationships between the images. We evaluated MedSAM-2 across a variety of medical imaging modalities, including abdominal organs, optic discs, brain tumors, thyroid nodules, and skin lesions, comparing it against state-of-the-art models in both traditional and interactive segmentation settings. Our findings show that MedSAM-2 not only surpasses existing models in performance but also exhibits superior generalization across a range of medical image segmentation tasks. Our code will be released at: [https://github.com/MedicineToken/Medical-SAM2](https://github.com/MedicineToken/Medical-SAM2)", "votes": null }, { "id": "2947289", "postDate": "08/05/2024 05:43:37", "content": "

As you can see in the GitHub repo you linked, Medical-SAM2 is licensed under Apache License 2.0. This is a license allowed by Kaggle. E.g. Public Kaggle Notebooks are under Apache License 2.0.

", "rawMarkdown": "As you can see in the GitHub repo you linked, Medical-SAM2 is licensed under Apache License 2.0. This is a license allowed by Kaggle. E.g. Public Kaggle Notebooks are under Apache License 2.0.", "votes": null }, { "id": "2950813", "postDate": "08/07/2024 23:38:36", "content": "

🥁🥁🥁🥁🥁 👍

", "rawMarkdown": "🥁🥁🥁🥁🥁 👍", "votes": null }, { "id": "2984741", "postDate": "09/09/2024 23:37:44", "content": "

does it work with kaggle notebook? anyone tried it?

", "rawMarkdown": "does it work with kaggle notebook? anyone tried it?", "votes": null }, { "id": "2984875", "postDate": "09/10/2024 05:09:19", "content": "

Try this Kaggle notebook: Automatic Spine Cord Segmentation: SAM 2

", "rawMarkdown": "Try this Kaggle notebook: [Automatic Spine Cord Segmentation: SAM 2](https://www.kaggle.com/code/claverru/automatic-spine-cord-segmentation-sam-2)", "votes": null }, { "id": "2984888", "postDate": "09/10/2024 05:22:17", "content": "

I already did. It is sam2 doing 2d segmentation. Medical SAM 2 is 3d segmentation using sam2 doing video segmentation.

", "rawMarkdown": "I already did. It is sam2 doing 2d segmentation. Medical SAM 2 is 3d segmentation using sam2 doing video segmentation.", "votes": null }, { "id": "2984892", "postDate": "09/10/2024 05:25:34", "content": "

And it appears using medical sam2 is quite difficult compared to just sam2.

", "rawMarkdown": "And it appears using medical sam2 is quite difficult compared to just sam2.", "votes": null }, { "id": "2986686", "postDate": "09/11/2024 23:39:59", "content": "

phew, i managed to run train_3d.py example of Medical-SAM2. so yes, it can work with kaggle notebook.

", "rawMarkdown": "phew, i managed to run train_3d.py example of Medical-SAM2. so yes, it can work with kaggle notebook.", "votes": null } ], "comments": [ { "id": 2947289, "author_name": "coderrkj", "author_url": "", "post_date": "08/05/2024 05:43:37", "content": "

As you can see in the GitHub repo you linked, Medical-SAM2 is licensed under Apache License 2.0. This is a license allowed by Kaggle. E.g. Public Kaggle Notebooks are under Apache License 2.0.

", "votes": null, "replies": [] }, { "id": 2950813, "author_name": "cyrilbourgeois", "author_url": "", "post_date": "08/07/2024 23:38:36", "content": "

🥁🥁🥁🥁🥁 👍

", "votes": null, "replies": [] }, { "id": 2984741, "author_name": "ibinti", "author_url": "", "post_date": "09/09/2024 23:37:44", "content": "

does it work with kaggle notebook? anyone tried it?

", "votes": null, "replies": [ { "id": 2984875, "author_name": "coderrkj", "author_url": "", "post_date": "09/10/2024 05:09:19", "content": "

Try this Kaggle notebook: Automatic Spine Cord Segmentation: SAM 2

", "votes": null, "replies": [ { "id": 2984888, "author_name": "ibinti", "author_url": "", "post_date": "09/10/2024 05:22:17", "content": "

I already did. It is sam2 doing 2d segmentation. Medical SAM 2 is 3d segmentation using sam2 doing video segmentation.

", "votes": null, "replies": [ { "id": 2984892, "author_name": "ibinti", "author_url": "", "post_date": "09/10/2024 05:25:34", "content": "

And it appears using medical sam2 is quite difficult compared to just sam2.

", "votes": null, "replies": [] } ] } ] }, { "id": 2986686, "author_name": "ibinti", "author_url": "", "post_date": "09/11/2024 23:39:59", "content": "

phew, i managed to run train_3d.py example of Medical-SAM2. so yes, it can work with kaggle notebook.

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "2947219": "# Medical SAM 2: Segment Medical Images As Video Via Segment Anything Model 2\n### [Paper](https://arxiv.org/abs/2408.00874)\n### [Code](https://github.com/MedicineToken/Medical-SAM2)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2F818ebc047b23541a0594187b373e7611%2Fposter1.png?generation=1722830955559500&alt=media)\n\n> In this paper, we introduce Medical SAM 2 (MedSAM-2), an advanced segmentation model that utilizes the SAM 2 framework to address both 2D and 3D medical image segmentation tasks. By adopting the philosophy of taking medical images as videos, MedSAM-2 not only applies to 3D medical images but also unlocks new One-prompt Segmentation capability. That allows users to provide a prompt for just one or a specific image targeting an object, after which the model can autonomously segment the same type of object in all subsequent images, regardless of temporal relationships between the images. We evaluated MedSAM-2 across a variety of medical imaging modalities, including abdominal organs, optic discs, brain tumors, thyroid nodules, and skin lesions, comparing it against state-of-the-art models in both traditional and interactive segmentation settings. Our findings show that MedSAM-2 not only surpasses existing models in performance but also exhibits superior generalization across a range of medical image segmentation tasks. Our code will be released at: [https://github.com/MedicineToken/Medical-SAM2](https://github.com/MedicineToken/Medical-SAM2)", "2947289": "As you can see in the GitHub repo you linked, Medical-SAM2 is licensed under Apache License 2.0. This is a license allowed by Kaggle. E.g. Public Kaggle Notebooks are under Apache License 2.0.", "2950813": "🥁🥁🥁🥁🥁 👍", "2984741": "does it work with kaggle notebook? anyone tried it?", "2984875": "Try this Kaggle notebook: [Automatic Spine Cord Segmentation: SAM 2](https://www.kaggle.com/code/claverru/automatic-spine-cord-segmentation-sam-2)", "2984888": "I already did. It is sam2 doing 2d segmentation. Medical SAM 2 is 3d segmentation using sam2 doing video segmentation.", "2984892": "And it appears using medical sam2 is quite difficult compared to just sam2.", "2986686": "phew, i managed to run train_3d.py example of Medical-SAM2. so yes, it can work with kaggle notebook." }, "source": "meta" }