![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2051313%2Ffcc26416104c2c4f72b0d6e7e049300d%2Fdownload.png?generation=1609560955345733&alt=media\")
Hi All,
\nI've generated heatmaps where on an idealized chest x-ray we would expect to see each class of findings. Output below:
\n
The kernel to generate that visualization is here:
\n\nMy thinking is that heatmap data can be used to help bolster confidence ratings in bounding boxes. Additionally, the heatmap data can help us to focus on certain areas of the chest x-ray images where we would expect to see certain clinical findings (i.e. generate a classifier that is good at recognizing Aortic enlargement and focus only on the areas of the chest x-ray where we are likely to see it.
\nLet me know what you think!
", "messages": [ { "id": 1135241, "postDate": "2021-01-02T04:20:54.123Z", "content": "Hi All,
\nI've generated heatmaps where on an idealized chest x-ray we would expect to see each class of findings. Output below:
\n
The kernel to generate that visualization is here:
\n\nMy thinking is that heatmap data can be used to help bolster confidence ratings in bounding boxes. Additionally, the heatmap data can help us to focus on certain areas of the chest x-ray images where we would expect to see certain clinical findings (i.e. generate a classifier that is good at recognizing Aortic enlargement and focus only on the areas of the chest x-ray where we are likely to see it.
\nLet me know what you think!
", "rawMarkdown": "Hi All,\n\nI've generated heatmaps where on an idealized chest x-ray we would expect to see each class of findings. Output below:\n\n\n\nThe kernel to generate that visualization is here:\n\n* https://www.kaggle.com/craigmthomas/localization-of-findings\n\nMy thinking is that heatmap data can be used to help bolster confidence ratings in bounding boxes. Additionally, the heatmap data can help us to focus on certain areas of the chest x-ray images where we would expect to see certain clinical findings (i.e. generate a classifier that is good at recognizing Aortic enlargement and focus only on the areas of the chest x-ray where we are likely to see it.\n\nLet me know what you think!", "votes": 5 }, { "id": 1137431, "postDate": "2021-01-03T23:29:53.610Z", "content": "Good job !
\nIt is fun to see that most of the lesions occurs in the right lung (see pneumothorax or lung opacity for exemple) !
", "rawMarkdown": "Good job ! \n\nIt is fun to see that most of the lesions occurs in the right lung (see pneumothorax or lung opacity for exemple) !", "votes": 1, "replies": [ { "id": 1137460, "postDate": "2021-01-04T00:41:02.890Z", "content": "Thanks! That is actually quite an interesting observation. I wonder if that's an honest observation due to the fact that the right lung is larger, or bias in how things are read. Something to look into more.
", "rawMarkdown": "Thanks! That is actually quite an interesting observation. I wonder if that's an honest observation due to the fact that the right lung is larger, or bias in how things are read. Something to look into more." } ] } ], "comments": [ { "id": 1137431, "author_name": "Tanguy Perennec", "author_url": "", "post_date": "2021-01-03T23:29:53.610000", "content": "Good job !
\nIt is fun to see that most of the lesions occurs in the right lung (see pneumothorax or lung opacity for exemple) !
", "votes": 1, "replies": [ { "id": 1137460, "author_name": "Craig Thomas", "author_url": "", "post_date": "2021-01-04T00:41:02.890000", "content": "Thanks! That is actually quite an interesting observation. I wonder if that's an honest observation due to the fact that the right lung is larger, or bias in how things are read. Something to look into more.
", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "1135241": "Hi All,\n\nI've generated heatmaps where on an idealized chest x-ray we would expect to see each class of findings. Output below:\n\n\n\nThe kernel to generate that visualization is here:\n\n* https://www.kaggle.com/craigmthomas/localization-of-findings\n\nMy thinking is that heatmap data can be used to help bolster confidence ratings in bounding boxes. Additionally, the heatmap data can help us to focus on certain areas of the chest x-ray images where we would expect to see certain clinical findings (i.e. generate a classifier that is good at recognizing Aortic enlargement and focus only on the areas of the chest x-ray where we are likely to see it.\n\nLet me know what you think!", "1137431": "Good job ! \n\nIt is fun to see that most of the lesions occurs in the right lung (see pneumothorax or lung opacity for exemple) !" } }