{ "id": 253305, "title": "Only case level labels - how to handle?", "url": "/competitions/rsna-miccai-brain-tumor-radiogenomic-classification/discussion/253305", "author_name": "Aman Arora", "post_date": "2021-07-16T00:47:02.663000", "votes": 4, "comment_count": 0, "views": 0, "content": "

In past competitions:

\n
    \n
  1. RSNA Intracranial Hemorrhage Detection
    2. \n
  2. RSNA STR Pulmonary Embolism Detection
    3. \n
\n

We've always had image and study level labels. But it seems like in this case we only have study-level labels - what's the best way to work with these labels? Would sequence models still be handy?

", "messages": [ { "id": 1389691, "postDate": "2021-07-16T00:47:02.663Z", "content": "

In past competitions:

\n
    \n
  1. RSNA Intracranial Hemorrhage Detection
    2. \n
  2. RSNA STR Pulmonary Embolism Detection
    3. \n
\n

We've always had image and study level labels. But it seems like in this case we only have study-level labels - what's the best way to work with these labels? Would sequence models still be handy?

", "rawMarkdown": "In past competitions: \n1. [RSNA Intracranial Hemorrhage Detection](https://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection/data)\n2. [RSNA STR Pulmonary Embolism Detection](https://www.kaggle.com/c/rsna-str-pulmonary-embolism-detection/data?select=train.csv)\n\nWe've always had image and study level labels. But it seems like in this case we only have study-level labels - what's the best way to work with these labels? Would sequence models still be handy?", "votes": 3 } ], "comments": [], "raw_markdown_by_id": { "1389691": "In past competitions: \n1. [RSNA Intracranial Hemorrhage Detection](https://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection/data)\n2. [RSNA STR Pulmonary Embolism Detection](https://www.kaggle.com/c/rsna-str-pulmonary-embolism-detection/data?select=train.csv)\n\nWe've always had image and study level labels. But it seems like in this case we only have study-level labels - what's the best way to work with these labels? Would sequence models still be handy?" } }