Hi everyone\nI am new to this competition. I want to know that what is the ground truth here like; ISUP score or Gleason score? when I visualize the label image it shows me 5 different values that don't make sense to me.\nAlso as 100 mask values are missing so can we remove that rows in a dataframe. And further, we use focal loss for dealing with unbalancing with data, is this a good approach?\nCan anyone please help me in this regard...
", "messages": [ { "id": 874686, "postDate": "2020-06-05T08:03:06.830Z", "content": "The ISUP score is based on the Gleason score (one-to-many mapping, see description of the competition), the Gleason score is based on the detected Gleason growth patterns (the individual labels in the mask). In other words, the mask values show more detail, and the Gleason score and ISUP score show the image-level label. For the test set you need to predict the ISUP score, but you can of course first predict the Gleason score and map this to the ISUP score.
", "rawMarkdown": "The ISUP score is based on the Gleason score (one-to-many mapping, see description of the competition), the Gleason score is based on the detected Gleason growth patterns (the individual labels in the mask). In other words, the mask values show more detail, and the Gleason score and ISUP score show the image-level label. For the test set you need to predict the ISUP score, but you can of course first predict the Gleason score and map this to the ISUP score.", "votes": 1, "replies": [ { "id": 874902, "postDate": "2020-06-05T11:02:21.657Z", "content": "Thank you for your response, So what if we directly use isup score as you said first to predict the gleason score and map it to isup. As we already have isup scores so why we need mapping... still confused
", "rawMarkdown": "Thank you for your response, So what if we directly use isup score as you said first to predict the gleason score and map it to isup. As we already have isup scores so why we need mapping... still confused" }, { "id": 875015, "postDate": "2020-06-05T13:03:52.920Z", "content": "Note that there are three different things at play:
\n\nYou can read more here: https://www.kaggle.com/c/prostate-cancer-grade-assessment/overview/additional-resources
", "rawMarkdown": "Note that there are three different things at play: \n\n1. The individual Gleason growth patterns (3, 4, 5)\n2. The Gleasons core (for example 3+5).\n3. The ISUP score (1-5).\n\nYou can read more here: https://www.kaggle.com/c/prostate-cancer-grade-assessment/overview/additional-resources", "votes": 1 }, { "id": 875038, "postDate": "2020-06-05T13:27:53.513Z", "content": "Thank you @wouterbulten for your time. I will look into the link for further clarification.
", "rawMarkdown": "Thank you @wouterbulten for your time. I will look into the link for further clarification." } ] }, { "id": 874357, "postDate": "2020-06-04T21:42:30.117Z", "content": "To answer your question: isup_grade.\nYou can check some kernels for more detail.
Hi everyone\nI am new to this competition. I want to know that what is the ground truth here like; ISUP score or Gleason score? when I visualize the label image it shows me 5 different values that don't make sense to me.\nAlso as 100 mask values are missing so can we remove that rows in a dataframe. And further, we use focal loss for dealing with unbalancing with data, is this a good approach?\nCan anyone please help me in this regard...
", "rawMarkdown": "Hi everyone\nI am new to this competition. I want to know that what is the ground truth here like; ISUP score or Gleason score? when I visualize the label image it shows me 5 different values that don't make sense to me.\nAlso as 100 mask values are missing so can we remove that rows in a dataframe. And further, we use focal loss for dealing with unbalancing with data, is this a good approach?\nCan anyone please help me in this regard..." } ], "comments": [ { "id": 874686, "author_name": "Wouter Bulten", "author_url": "", "post_date": "2020-06-05T08:03:06.830000", "content": "The ISUP score is based on the Gleason score (one-to-many mapping, see description of the competition), the Gleason score is based on the detected Gleason growth patterns (the individual labels in the mask). In other words, the mask values show more detail, and the Gleason score and ISUP score show the image-level label. For the test set you need to predict the ISUP score, but you can of course first predict the Gleason score and map this to the ISUP score.
", "votes": 1, "replies": [ { "id": 874902, "author_name": "Mudasser Afzal", "author_url": "", "post_date": "2020-06-05T11:02:21.657000", "content": "Thank you for your response, So what if we directly use isup score as you said first to predict the gleason score and map it to isup. As we already have isup scores so why we need mapping... still confused
", "votes": 0, "replies": [] }, { "id": 875015, "author_name": "Wouter Bulten", "author_url": "", "post_date": "2020-06-05T13:03:52.920000", "content": "Note that there are three different things at play:
\n\nYou can read more here: https://www.kaggle.com/c/prostate-cancer-grade-assessment/overview/additional-resources
", "votes": 1, "replies": [] }, { "id": 875038, "author_name": "Mudasser Afzal", "author_url": "", "post_date": "2020-06-05T13:27:53.513000", "content": "Thank you @wouterbulten for your time. I will look into the link for further clarification.
", "votes": 0, "replies": [] } ] }, { "id": 874357, "author_name": "BachT", "author_url": "", "post_date": "2020-06-04T21:42:30.117000", "content": "To answer your question: isup_grade.\nYou can check some kernels for more detail.