\n
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1887174%2Ff008de43268cf0a5819976652788eecc%2Fjdfkasjdflkajsdkfjasldfk.jpg?generation=1597173557636402&alt=media\")
\nI feel like the W shape followed by downward slope is pretty common along all the patients. I have no idea why that might be. Please correct me if I'm wrong. I might just be seeing things LOL.
Is it just me or there is a common shape in the plots of FVC vs Weeks plot for the patients?
\n
\nI feel like the W shape followed by downward slope is pretty common along all the patients. I have no idea why that might be. Please correct me if I'm wrong. I might just be seeing things LOL.
Is it just me or there is a common shape in the plots of FVC vs Weeks plot for the patients?
\n
\nI feel like the W shape followed by downward slope is pretty common along all the patients. I have no idea why that might be. Please correct me if I'm wrong. I might just be seeing things LOL.
Nice One Dude !! Keep sharing those !!
", "rawMarkdown": "Nice One Dude !! Keep sharing those !!", "votes": null }, { "id": "967319", "postDate": "08/12/2020 06:47:38", "content": "I think there FVC has a high amount of variability and the trend is somewhat drowned out by noise in the measurements. Since the measurements tend to be group towards the start of the time period it may cause zig-zagging early on. The graph below shows FVC decline grouped into blocks of weeks.
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4209361%2F0b63364a5edbbf86cd32152bb5cacf47%2Fdev.png?generation=1597214815804571&alt=media\")
@Jonykarki I was also interested to notice this pattern.
\nAn experienced doctor demonstrated to me what a patient needs to do during a FVC test - the patient literally has to do as close to a COMPLETE exhale as possible into the spirometer to give a proper reading of the total lung volume. During the demonstration, towards the end of the exhale it looked very uncomfortable - which is understandable as it's not a normal activity to completely empty your lungs.
\nI think it is not unreasonable to think that a patient may not breath out fully due to discomfort, potentially making the initial FVC readings lower than true FVC. In the example figure you posted here, the first 2 data points look like they could be too low, then the next few after that get higher - almost as if the patient is gradually mastering the technique. There are many other patient FVC vs Week plots that look similar.
\nIf this is a genuine issue, I think it poses difficulties in processing of the data before model training (ie. dropping data points we believe to be too low?!). Looking at the time plots for each patient, I can't see a simple mathematical way of isolating these inaccurate points. Maybe there is one though - or maybe we'd need a doctor to take a look! :)
\nThe comment by @lgrieg is also interesting - the FVC measurements seem to be more frequent at the beginning of the measurement period for many patients, possibly explaining why we see the effects of expected noise even more earlier on …
", "rawMarkdown": "Jonykarki I was also interested to notice this pattern. \n\nAn experienced doctor demonstrated to me what a patient needs to do during a FVC test - the patient literally has to do as close to a COMPLETE exhale as possible into the spirometer to give a proper reading of the total lung volume. During the demonstration, towards the end of the exhale it looked very uncomfortable - which is understandable as it's not a normal activity to completely empty your lungs.\n\nI think it is not unreasonable to think that a patient may not breath out fully due to discomfort, potentially making the initial FVC readings lower than true FVC. In the example figure you posted here, the first 2 data points look like they could be too low, then the next few after that get higher - almost as if the patient is gradually mastering the technique. There are many other patient FVC vs Week plots that look similar.\n\nIf this is a genuine issue, I think it poses difficulties in processing of the data before model training (ie. dropping data points we believe to be too low?!). Looking at the time plots for each patient, I can't see a simple mathematical way of isolating these inaccurate points. Maybe there is one though - or maybe we'd need a doctor to take a look! :)\n\nThe comment by @lgrieg is also interesting - the FVC measurements seem to be more frequent at the beginning of the measurement period for many patients, possibly explaining why we see the effects of expected noise even more earlier on ...", "votes": null }, { "id": "971828", "postDate": "08/16/2020 00:42:58", "content": "You have a very interesting point.
", "rawMarkdown": "You have a very interesting point.", "votes": null } ], "comments": [ { "id": 967319, "author_name": "lgreig", "author_url": "", "post_date": "08/12/2020 06:47:38", "content": "I think there FVC has a high amount of variability and the trend is somewhat drowned out by noise in the measurements. Since the measurements tend to be group towards the start of the time period it may cause zig-zagging early on. The graph below shows FVC decline grouped into blocks of weeks.
\n
@Jonykarki I was also interested to notice this pattern.
\nAn experienced doctor demonstrated to me what a patient needs to do during a FVC test - the patient literally has to do as close to a COMPLETE exhale as possible into the spirometer to give a proper reading of the total lung volume. During the demonstration, towards the end of the exhale it looked very uncomfortable - which is understandable as it's not a normal activity to completely empty your lungs.
\nI think it is not unreasonable to think that a patient may not breath out fully due to discomfort, potentially making the initial FVC readings lower than true FVC. In the example figure you posted here, the first 2 data points look like they could be too low, then the next few after that get higher - almost as if the patient is gradually mastering the technique. There are many other patient FVC vs Week plots that look similar.
\nIf this is a genuine issue, I think it poses difficulties in processing of the data before model training (ie. dropping data points we believe to be too low?!). Looking at the time plots for each patient, I can't see a simple mathematical way of isolating these inaccurate points. Maybe there is one though - or maybe we'd need a doctor to take a look! :)
\nThe comment by @lgrieg is also interesting - the FVC measurements seem to be more frequent at the beginning of the measurement period for many patients, possibly explaining why we see the effects of expected noise even more earlier on …
", "votes": null, "replies": [ { "id": 971828, "author_name": "jonykarki", "author_url": "", "post_date": "08/16/2020 00:42:58", "content": "You have a very interesting point.
", "votes": null, "replies": [] } ] }, { "id": 967224, "author_name": "sudarshanpatil", "author_url": "", "post_date": "08/12/2020 05:31:59", "content": "Nice One Dude !! Keep sharing those !!
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "966926": "Is it just me or there is a common shape in the plots of FVC vs Weeks plot for the patients?\n\nI feel like the W shape followed by downward slope is pretty common along all the patients. I have no idea why that might be. Please correct me if I'm wrong. I might just be seeing things LOL.", "967224": "Nice One Dude !! Keep sharing those !!", "967319": "I think there FVC has a high amount of variability and the trend is somewhat drowned out by noise in the measurements. Since the measurements tend to be group towards the start of the time period it may cause zig-zagging early on. The graph below shows FVC decline grouped into blocks of weeks.\n\n", "971789": "Jonykarki I was also interested to notice this pattern. \n\nAn experienced doctor demonstrated to me what a patient needs to do during a FVC test - the patient literally has to do as close to a COMPLETE exhale as possible into the spirometer to give a proper reading of the total lung volume. During the demonstration, towards the end of the exhale it looked very uncomfortable - which is understandable as it's not a normal activity to completely empty your lungs.\n\nI think it is not unreasonable to think that a patient may not breath out fully due to discomfort, potentially making the initial FVC readings lower than true FVC. In the example figure you posted here, the first 2 data points look like they could be too low, then the next few after that get higher - almost as if the patient is gradually mastering the technique. There are many other patient FVC vs Week plots that look similar.\n\nIf this is a genuine issue, I think it poses difficulties in processing of the data before model training (ie. dropping data points we believe to be too low?!). Looking at the time plots for each patient, I can't see a simple mathematical way of isolating these inaccurate points. Maybe there is one though - or maybe we'd need a doctor to take a look! :)\n\nThe comment by @lgrieg is also interesting - the FVC measurements seem to be more frequent at the beginning of the measurement period for many patients, possibly explaining why we see the effects of expected noise even more earlier on ...", "971828": "You have a very interesting point." }, "source": "meta" }