{
  "id": 510925,
  "title": "Seemingly Discrepancies in train.csv and train_label_coordinates.csv",
  "url": "/competitions/rsna-2024-lumbar-spine-degenerative-classification/discussion/510925",
  "author_name": "",
  "post_date": "2024-06-08T10:21:41.902019300Z",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Your response is highly encouraged and is invaluable.</p>\n<p>1) One thing that is not clear to me is that when one sees the train_label_coordinates.csv file, it shows a condition of spinal canal stenosis in the patient (study_id) 4003253, but train.csv shows that everything is normal/mild with that person. This seems contradictory to me. So, am I interpreting both files in the right way, or does train_label_coordinates.csv say, \"Take the study 4003253, series 702807833, instance (dcm) 8, and check if there is a Spinal Canal Stenosis at level L1/L2; then, after predicting, use train.csv to test your prediction,\" or something else?</p>\n<p>2)What do the coordinates x and y in the train_label_coordinates.csv mean? The data page says that [x/y] - The x/y coordinates for the center of the area that defined the label. But I can't understand what the statement means by 'label'. Is it the level given in that row? I have tried to check, but I do not have a firm opinion.</p>\n<p>3) Most importantly, there are missing column values in train.csv, and correspondingly their entries are missing in train_label_coordinates.csv. For example, the answer of right_subarticular_stenosis_l1_l2 for 'study_id 46494080' is missing in train.csv, and correspondingly there is no entry of right_subarticular_stenosis_l1_l2 for 'study_id 46494080' in train_label_coordinates.csv. By looking at all 25 entries of the sample_submission.csv, I concluded that for each study_id, each type of stenosis, and each level of the intervertebral disc, three probabilities for normal/mild, moderate, and severe have to be predicted. But how can anyone predict the values, especially if entries for axial (top-down) views are missing in the most expected file 'test_label_coordinates.csv', because even if a model correctly identifies the type of stenosis, which level will the model put it in? I do not know how the evaluation metric handles it. Hence, I need a lot of help.</p>",
  "messages": [
    {
      "id": "2861699",
      "postDate": "06/08/2024 10:21:41",
      "content": "<p>Your response is highly encouraged and is invaluable.</p>\n<p>1) One thing that is not clear to me is that when one sees the train_label_coordinates.csv file, it shows a condition of spinal canal stenosis in the patient (study_id) 4003253, but train.csv shows that everything is normal/mild with that person. This seems contradictory to me. So, am I interpreting both files in the right way, or does train_label_coordinates.csv say, \"Take the study 4003253, series 702807833, instance (dcm) 8, and check if there is a Spinal Canal Stenosis at level L1/L2; then, after predicting, use train.csv to test your prediction,\" or something else?</p>\n<p>2)What do the coordinates x and y in the train_label_coordinates.csv mean? The data page says that [x/y] - The x/y coordinates for the center of the area that defined the label. But I can't understand what the statement means by 'label'. Is it the level given in that row? I have tried to check, but I do not have a firm opinion.</p>\n<p>3) Most importantly, there are missing column values in train.csv, and correspondingly their entries are missing in train_label_coordinates.csv. For example, the answer of right_subarticular_stenosis_l1_l2 for 'study_id 46494080' is missing in train.csv, and correspondingly there is no entry of right_subarticular_stenosis_l1_l2 for 'study_id 46494080' in train_label_coordinates.csv. By looking at all 25 entries of the sample_submission.csv, I concluded that for each study_id, each type of stenosis, and each level of the intervertebral disc, three probabilities for normal/mild, moderate, and severe have to be predicted. But how can anyone predict the values, especially if entries for axial (top-down) views are missing in the most expected file 'test_label_coordinates.csv', because even if a model correctly identifies the type of stenosis, which level will the model put it in? I do not know how the evaluation metric handles it. Hence, I need a lot of help.</p>",
      "rawMarkdown": "Your response is highly encouraged and is invaluable.\n\n1) One thing that is not clear to me is that when one sees the train_label_coordinates.csv file, it shows a condition of spinal canal stenosis in the patient (study_id) 4003253, but train.csv shows that everything is normal/mild with that person. This seems contradictory to me. So, am I interpreting both files in the right way, or does train_label_coordinates.csv say, \"Take the study 4003253, series 702807833, instance (dcm) 8, and check if there is a Spinal Canal Stenosis at level L1/L2; then, after predicting, use train.csv to test your prediction,\" or something else?\n\n2)What do the coordinates x and y in the train_label_coordinates.csv mean? The data page says that [x/y] - The x/y coordinates for the center of the area that defined the label. But I can't understand what the statement means by 'label'. Is it the level given in that row? I have tried to check, but I do not have a firm opinion.\n\n3) Most importantly, there are missing column values in train.csv, and correspondingly their entries are missing in train_label_coordinates.csv. For example, the answer of right_subarticular_stenosis_l1_l2 for 'study_id 46494080' is missing in train.csv, and correspondingly there is no entry of right_subarticular_stenosis_l1_l2 for 'study_id 46494080' in train_label_coordinates.csv. By looking at all 25 entries of the sample_submission.csv, I concluded that for each study_id, each type of stenosis, and each level of the intervertebral disc, three probabilities for normal/mild, moderate, and severe have to be predicted. But how can anyone predict the values, especially if entries for axial (top-down) views are missing in the most expected file 'test_label_coordinates.csv', because even if a model correctly identifies the type of stenosis, which level will the model put it in? I do not know how the evaluation metric handles it. Hence, I need a lot of help.",
      "votes": null
    },
    {
      "id": "2861997",
      "postDate": "06/08/2024 13:32:34",
      "content": "<p><a href=\"https://www.kaggle.com/pankajdebroy\" target=\"_blank\">@pankajdebroy</a> Thank you for your notebook <a href=\"https://www.kaggle.com/code/pankajdebroy/rsna-visualisation-training-submission\" target=\"_blank\">https://www.kaggle.com/code/pankajdebroy/rsna-visualisation-training-submission</a> .I finally got the answers of (1) and (2)</p>",
      "rawMarkdown": "pankajdebroy Thank you for your notebook https://www.kaggle.com/code/pankajdebroy/rsna-visualisation-training-submission .I finally got the answers of (1) and (2)",
      "votes": null
    },
    {
      "id": "2861998",
      "postDate": "06/08/2024 13:33:06",
      "content": "<p>And to some extend also to 3</p>",
      "rawMarkdown": "And to some extend also to 3",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2861997,
      "author_name": "devsya",
      "author_url": "",
      "post_date": "06/08/2024 13:32:34",
      "content": "<p><a href=\"https://www.kaggle.com/pankajdebroy\" target=\"_blank\">@pankajdebroy</a> Thank you for your notebook <a href=\"https://www.kaggle.com/code/pankajdebroy/rsna-visualisation-training-submission\" target=\"_blank\">https://www.kaggle.com/code/pankajdebroy/rsna-visualisation-training-submission</a> .I finally got the answers of (1) and (2)</p>",
      "votes": null,
      "replies": [
        {
          "id": 2861998,
          "author_name": "devsya",
          "author_url": "",
          "post_date": "06/08/2024 13:33:06",
          "content": "<p>And to some extend also to 3</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2861699": "Your response is highly encouraged and is invaluable.\n\n1) One thing that is not clear to me is that when one sees the train_label_coordinates.csv file, it shows a condition of spinal canal stenosis in the patient (study_id) 4003253, but train.csv shows that everything is normal/mild with that person. This seems contradictory to me. So, am I interpreting both files in the right way, or does train_label_coordinates.csv say, \"Take the study 4003253, series 702807833, instance (dcm) 8, and check if there is a Spinal Canal Stenosis at level L1/L2; then, after predicting, use train.csv to test your prediction,\" or something else?\n\n2)What do the coordinates x and y in the train_label_coordinates.csv mean? The data page says that [x/y] - The x/y coordinates for the center of the area that defined the label. But I can't understand what the statement means by 'label'. Is it the level given in that row? I have tried to check, but I do not have a firm opinion.\n\n3) Most importantly, there are missing column values in train.csv, and correspondingly their entries are missing in train_label_coordinates.csv. For example, the answer of right_subarticular_stenosis_l1_l2 for 'study_id 46494080' is missing in train.csv, and correspondingly there is no entry of right_subarticular_stenosis_l1_l2 for 'study_id 46494080' in train_label_coordinates.csv. By looking at all 25 entries of the sample_submission.csv, I concluded that for each study_id, each type of stenosis, and each level of the intervertebral disc, three probabilities for normal/mild, moderate, and severe have to be predicted. But how can anyone predict the values, especially if entries for axial (top-down) views are missing in the most expected file 'test_label_coordinates.csv', because even if a model correctly identifies the type of stenosis, which level will the model put it in? I do not know how the evaluation metric handles it. Hence, I need a lot of help.",
    "2861997": "pankajdebroy Thank you for your notebook https://www.kaggle.com/code/pankajdebroy/rsna-visualisation-training-submission .I finally got the answers of (1) and (2)",
    "2861998": "And to some extend also to 3"
  },
  "source": "meta"
}