![](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F16438831%2F323ac8c4b8dcc4d3bb7e030ee6d058a0%2Feeg.PNG?generation=1706873182811844&alt=media\")
Hi, kagglers!
\nThere are few data in patient, eeg, spectrogram[it seems like to need Augmentation]
\nImage augmentation is good strategy to score higher
\nAlso, there are good Data Augmentations
[noise addition, generative adversarial networks, sliding windows, sampling, Fourier transform,\nrecombination of segmentation, and others]
\nIn this paper: https://www.sciencedirect.com/science/article/abs/pii/S0165027020303083
\n🛑\"Noise Addition and sliding windows provied the highest accuract boost\"
\nAnd then when studying chris notebook[eeg->spectrogram], i have a idea with new chain in spectrogram
\n
We need to classify Lateral or General, so it seems like making with [T3, C3, T4, C4] is not that good option
FL,FR,PL,PR is also 4 part with [LL, LP, RP, RR], But it is divided by front & back
And We have classify others, and new chain can helpful to detect others
🚩 It is just my own idea, and i'm not sure it will be effective. Feel free to your opinion!
", "messages": [ { "id": "2632487", "postDate": "02/02/2024 11:33:35", "content": "Hi, kagglers!
\nThere are few data in patient, eeg, spectrogram[it seems like to need Augmentation]
\nImage augmentation is good strategy to score higher
\nAlso, there are good Data Augmentations
[noise addition, generative adversarial networks, sliding windows, sampling, Fourier transform,\nrecombination of segmentation, and others]
\nIn this paper: https://www.sciencedirect.com/science/article/abs/pii/S0165027020303083
\n🛑\"Noise Addition and sliding windows provied the highest accuract boost\"
\nAnd then when studying chris notebook[eeg->spectrogram], i have a idea with new chain in spectrogram
\n
We need to classify Lateral or General, so it seems like making with [T3, C3, T4, C4] is not that good option
FL,FR,PL,PR is also 4 part with [LL, LP, RP, RR], But it is divided by front & back
And We have classify others, and new chain can helpful to detect others
🚩 It is just my own idea, and i'm not sure it will be effective. Feel free to your opinion!
", "rawMarkdown": "Hi, kagglers!\n\nThere are few data in patient, eeg, spectrogram[it seems like to need **Augmentation**]\n\nImage augmentation is good strategy to score higher\nAlso, there are good Data Augmentations\n\n**[noise addition, generative adversarial networks, sliding windows, sampling, Fourier transform,\nrecombination of segmentation, and others]**\n\nIn this paper: https://www.sciencedirect.com/science/article/abs/pii/S0165027020303083\n\n🛑**\"Noise Addition and sliding windows provied the highest accuract boost\"**\n\nAnd then when studying chris notebook[eeg->spectrogram], i have a idea with new chain in spectrogram\n\n\n\n- We need to classify `Lateral` or `General`, so it seems like making with [T3, C3, T4, C4] is not that good option\n\n- `FL,FR,PL,PR` is also 4 part with [LL, LP, RP, RR], But it is divided by `front & back` \n\n- And We have classify `others`, and new chain can helpful to detect `others`\n \n\n🚩 It is just my own idea, and i'm not sure it will be effective. Feel free to your opinion!", "votes": null }, { "id": "2632510", "postDate": "02/02/2024 11:54:39", "content": "My Notebook[New chain without denosing]
\nhttps://www.kaggle.com/code/seoyunje/hms-new-chain-with-none
\nMy Notebook[New chain with denosing(demy)]
\nhttps://www.kaggle.com/code/seoyunje/hms-new-chain-with-dmey
", "rawMarkdown": "My Notebook[New chain without denosing]\n\nhttps://www.kaggle.com/code/seoyunje/hms-new-chain-with-none\n\nMy Notebook[New chain with denosing(demy)]\n\nhttps://www.kaggle.com/code/seoyunje/hms-new-chain-with-dmey", "votes": null }, { "id": "2634067", "postDate": "02/03/2024 13:23:04", "content": "I liked the idea, I believe it could be interesting, even if it doesn't lower the score, it could be another way of generating diversity in the model.
\nOne point that I believe can be tested is the use of Fz, Cz, Pz, possibly they can provide additional information, what do you think about that?
\nAnother point is that [T3, C3] are sharing two sides FL and PL, this could harm the final analysis, what is your opinion?
", "rawMarkdown": "I liked the idea, I believe it could be interesting, even if it doesn't lower the score, it could be another way of generating diversity in the model.\n\nOne point that I believe can be tested is the use of Fz, Cz, Pz, possibly they can provide additional information, what do you think about that?\n\nAnother point is that [T3, C3] are sharing two sides FL and PL, this could harm the final analysis, what is your opinion?", "votes": null }, { "id": "2636436", "postDate": "02/05/2024 05:03:58", "content": "Yes, i'm also focusing on generating diversity in the model, Because i think the potential for characteristics to vary significantly among patients is very high!
\nThere're still a lot of experiments in spectrogram😀 Keep Customizing!
", "rawMarkdown": "Yes, i'm also focusing on generating diversity in the model, Because i think the potential for characteristics to vary significantly among patients is very high! \n\n- I think To try make montage about [Fz, Cz, Pz] is good idea, I'm just focusing on making montage by left or right \n- Sharing with [T3, C3] in FL & PL could harm the final analysis, But, I think it's not that good option to make [T3, C3, Cz, T4, C4]\n\nThere're still a lot of experiments in spectrogram😀 Keep Customizing!", "votes": null }, { "id": "2637045", "postDate": "02/05/2024 13:22:09", "content": "Nice, tks from responding Peter, let's do yours experiments
", "rawMarkdown": "Nice, tks from responding Peter, let's do yours experiments", "votes": null }, { "id": "2658567", "postDate": "02/19/2024 09:20:11", "content": "Thank you for sharing your idea! I have two questions regarding the [Fz, Cz, Pz] montage—I'm not sure if it's relevant. Is there a specific medical rationale or clinical justification behind choosing this particular montage for EEG data analysis, or is it more of a methodological consideration tied to the classification of brain activity into seizure, generalized, lateralized, or other classes? If you experimented with this montage [Fz, Cz, Pz], what kind of brain activity can it effectively detect?
", "rawMarkdown": "Thank you for sharing your idea! I have two questions regarding the [Fz, Cz, Pz] montage—I'm not sure if it's relevant. Is there a specific medical rationale or clinical justification behind choosing this particular montage for EEG data analysis, or is it more of a methodological consideration tied to the classification of brain activity into seizure, generalized, lateralized, or other classes? If you experimented with this montage [Fz, Cz, Pz], what kind of brain activity can it effectively detect?", "votes": null }, { "id": "2658911", "postDate": "02/19/2024 13:47:42", "content": "Playing around with using the centerlines myself - but a key point to remember - we are modeling to predict what the expert's say, so likely we should be using the montages that they use. Read the ACNS Guidelines, visit the web site and watch the 4 videos.
", "rawMarkdown": "Playing around with using the centerlines myself - but a key point to remember - we are modeling to predict what the expert's say, so likely we should be using the montages that they use. Read the [ACNS Guidelines](https://www.acns.org/UserFiles/file/ACNSStandardizedCriticalCareEEGTerminology_rev2021.pdf), visit the web site and watch the 4 videos.", "votes": null } ], "comments": [ { "id": 2632510, "author_name": "seoyunje", "author_url": "", "post_date": "02/02/2024 11:54:39", "content": "My Notebook[New chain without denosing]
\nhttps://www.kaggle.com/code/seoyunje/hms-new-chain-with-none
\nMy Notebook[New chain with denosing(demy)]
\nhttps://www.kaggle.com/code/seoyunje/hms-new-chain-with-dmey
", "votes": null, "replies": [] }, { "id": 2634067, "author_name": "rafaelzimmermann1", "author_url": "", "post_date": "02/03/2024 13:23:04", "content": "I liked the idea, I believe it could be interesting, even if it doesn't lower the score, it could be another way of generating diversity in the model.
\nOne point that I believe can be tested is the use of Fz, Cz, Pz, possibly they can provide additional information, what do you think about that?
\nAnother point is that [T3, C3] are sharing two sides FL and PL, this could harm the final analysis, what is your opinion?
", "votes": null, "replies": [ { "id": 2636436, "author_name": "seoyunje", "author_url": "", "post_date": "02/05/2024 05:03:58", "content": "Yes, i'm also focusing on generating diversity in the model, Because i think the potential for characteristics to vary significantly among patients is very high!
\nThere're still a lot of experiments in spectrogram😀 Keep Customizing!
", "votes": null, "replies": [ { "id": 2637045, "author_name": "rafaelzimmermann1", "author_url": "", "post_date": "02/05/2024 13:22:09", "content": "Nice, tks from responding Peter, let's do yours experiments
", "votes": null, "replies": [ { "id": 2658567, "author_name": "theeventhorizons", "author_url": "", "post_date": "02/19/2024 09:20:11", "content": "Thank you for sharing your idea! I have two questions regarding the [Fz, Cz, Pz] montage—I'm not sure if it's relevant. Is there a specific medical rationale or clinical justification behind choosing this particular montage for EEG data analysis, or is it more of a methodological consideration tied to the classification of brain activity into seizure, generalized, lateralized, or other classes? If you experimented with this montage [Fz, Cz, Pz], what kind of brain activity can it effectively detect?
", "votes": null, "replies": [] } ] } ] } ] }, { "id": 2658911, "author_name": "pcjimmmy", "author_url": "", "post_date": "02/19/2024 13:47:42", "content": "Playing around with using the centerlines myself - but a key point to remember - we are modeling to predict what the expert's say, so likely we should be using the montages that they use. Read the ACNS Guidelines, visit the web site and watch the 4 videos.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2632487": "Hi, kagglers!\n\nThere are few data in patient, eeg, spectrogram[it seems like to need **Augmentation**]\n\nImage augmentation is good strategy to score higher\nAlso, there are good Data Augmentations\n\n**[noise addition, generative adversarial networks, sliding windows, sampling, Fourier transform,\nrecombination of segmentation, and others]**\n\nIn this paper: https://www.sciencedirect.com/science/article/abs/pii/S0165027020303083\n\n🛑**\"Noise Addition and sliding windows provied the highest accuract boost\"**\n\nAnd then when studying chris notebook[eeg->spectrogram], i have a idea with new chain in spectrogram\n\n\n\n- We need to classify `Lateral` or `General`, so it seems like making with [T3, C3, T4, C4] is not that good option\n\n- `FL,FR,PL,PR` is also 4 part with [LL, LP, RP, RR], But it is divided by `front & back` \n\n- And We have classify `others`, and new chain can helpful to detect `others`\n \n\n🚩 It is just my own idea, and i'm not sure it will be effective. Feel free to your opinion!", "2632510": "My Notebook[New chain without denosing]\n\nhttps://www.kaggle.com/code/seoyunje/hms-new-chain-with-none\n\nMy Notebook[New chain with denosing(demy)]\n\nhttps://www.kaggle.com/code/seoyunje/hms-new-chain-with-dmey", "2634067": "I liked the idea, I believe it could be interesting, even if it doesn't lower the score, it could be another way of generating diversity in the model.\n\nOne point that I believe can be tested is the use of Fz, Cz, Pz, possibly they can provide additional information, what do you think about that?\n\nAnother point is that [T3, C3] are sharing two sides FL and PL, this could harm the final analysis, what is your opinion?", "2636436": "Yes, i'm also focusing on generating diversity in the model, Because i think the potential for characteristics to vary significantly among patients is very high! \n\n- I think To try make montage about [Fz, Cz, Pz] is good idea, I'm just focusing on making montage by left or right \n- Sharing with [T3, C3] in FL & PL could harm the final analysis, But, I think it's not that good option to make [T3, C3, Cz, T4, C4]\n\nThere're still a lot of experiments in spectrogram😀 Keep Customizing!", "2637045": "Nice, tks from responding Peter, let's do yours experiments", "2658567": "Thank you for sharing your idea! I have two questions regarding the [Fz, Cz, Pz] montage—I'm not sure if it's relevant. Is there a specific medical rationale or clinical justification behind choosing this particular montage for EEG data analysis, or is it more of a methodological consideration tied to the classification of brain activity into seizure, generalized, lateralized, or other classes? If you experimented with this montage [Fz, Cz, Pz], what kind of brain activity can it effectively detect?", "2658911": "Playing around with using the centerlines myself - but a key point to remember - we are modeling to predict what the expert's say, so likely we should be using the montages that they use. Read the [ACNS Guidelines](https://www.acns.org/UserFiles/file/ACNSStandardizedCriticalCareEEGTerminology_rev2021.pdf), visit the web site and watch the 4 videos." }, "source": "meta" }