I don't get it. Id this eye science is names 'science' they should operate with something countable. If you have 4 stages (why 4?), they have to have exact description. something like - for stage two - swollen blood vessels, then, how they name them \"wool cotton spots\", with area, let say 5% of retina, etc. Something, that we can count and know that we are at the same page with all doctor in the world. And by the way in this case they would not need fancy algorithm, trying to guess doctor preferences, but vision program, which would look for defects, count them, area affected (and something else?) and we will get exact result, without any guessing.\nWhat do you think, guys?
", "messages": [ { "id": "568792", "postDate": "07/05/2019 12:53:58", "content": "I don't get it. Id this eye science is names 'science' they should operate with something countable. If you have 4 stages (why 4?), they have to have exact description. something like - for stage two - swollen blood vessels, then, how they name them \"wool cotton spots\", with area, let say 5% of retina, etc. Something, that we can count and know that we are at the same page with all doctor in the world. And by the way in this case they would not need fancy algorithm, trying to guess doctor preferences, but vision program, which would look for defects, count them, area affected (and something else?) and we will get exact result, without any guessing.\nWhat do you think, guys?
", "rawMarkdown": "I don't get it. Id this eye science is names 'science' they should operate with something countable. If you have 4 stages (why 4?), they have to have exact description. something like - for stage two - swollen blood vessels, then, how they name them \"wool cotton spots\", with area, let say 5% of retina, etc. Something, that we can count and know that we are at the same page with all doctor in the world. And by the way in this case they would not need fancy algorithm, trying to guess doctor preferences, but vision program, which would look for defects, count them, area affected (and something else?) and we will get exact result, without any guessing.\nWhat do you think, guys?", "votes": null }, { "id": "568805", "postDate": "07/05/2019 13:14:01", "content": "It's the subjective diagnosis by a human therefore it will be inconsistent. Three doctors may look at the same image and all give different diagnoses. I think the point of the competition is to remove the human from the diagnosis in order to increase consistency and accuracy.
", "rawMarkdown": "It's the subjective diagnosis by a human therefore it will be inconsistent. Three doctors may look at the same image and all give different diagnoses. I think the point of the competition is to remove the human from the diagnosis in order to increase consistency and accuracy.", "votes": null }, { "id": "568904", "postDate": "07/05/2019 16:23:44", "content": "Yes, that is what I am talking about. Formulate numerical criteria and they will stay the same, doesn't mean, which doctor is working.
", "rawMarkdown": "Yes, that is what I am talking about. Formulate numerical criteria and they will stay the same, doesn't mean, which doctor is working.", "votes": null }, { "id": "569500", "postDate": "07/06/2019 19:07:29", "content": "Hi, @serg132003 there is just more than agreeing in some common classification rules, this article explains why the algorithm was created.\nhttps://ai.googleblog.com/2016/11/deep-learning-for-detection-of-diabetic.html
\n\n\"Interpreting these photographs requires specialized training, and in many regions of the world, there aren’t enough qualified graders to screen everyone who is at risk. '
", "rawMarkdown": "Hi, @serg132003 there is just more than agreeing in some common classification rules, this article explains why the algorithm was created.\nhttps://ai.googleblog.com/2016/11/deep-learning-for-detection-of-diabetic.html\n\n\"Interpreting these photographs requires specialized training, and in many regions of the world, there aren’t enough qualified graders to screen everyone who is at risk. '", "votes": null } ], "comments": [ { "id": 568805, "author_name": "monkeyorman", "author_url": "", "post_date": "07/05/2019 13:14:01", "content": "It's the subjective diagnosis by a human therefore it will be inconsistent. Three doctors may look at the same image and all give different diagnoses. I think the point of the competition is to remove the human from the diagnosis in order to increase consistency and accuracy.
", "votes": null, "replies": [] }, { "id": 568904, "author_name": "serg132003", "author_url": "", "post_date": "07/05/2019 16:23:44", "content": "Yes, that is what I am talking about. Formulate numerical criteria and they will stay the same, doesn't mean, which doctor is working.
", "votes": null, "replies": [] }, { "id": 569500, "author_name": "cv13j0", "author_url": "", "post_date": "07/06/2019 19:07:29", "content": "Hi, @serg132003 there is just more than agreeing in some common classification rules, this article explains why the algorithm was created.\nhttps://ai.googleblog.com/2016/11/deep-learning-for-detection-of-diabetic.html
\n\n\"Interpreting these photographs requires specialized training, and in many regions of the world, there aren’t enough qualified graders to screen everyone who is at risk. '
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "568792": "I don't get it. Id this eye science is names 'science' they should operate with something countable. If you have 4 stages (why 4?), they have to have exact description. something like - for stage two - swollen blood vessels, then, how they name them \"wool cotton spots\", with area, let say 5% of retina, etc. Something, that we can count and know that we are at the same page with all doctor in the world. And by the way in this case they would not need fancy algorithm, trying to guess doctor preferences, but vision program, which would look for defects, count them, area affected (and something else?) and we will get exact result, without any guessing.\nWhat do you think, guys?", "568805": "It's the subjective diagnosis by a human therefore it will be inconsistent. Three doctors may look at the same image and all give different diagnoses. I think the point of the competition is to remove the human from the diagnosis in order to increase consistency and accuracy.", "568904": "Yes, that is what I am talking about. Formulate numerical criteria and they will stay the same, doesn't mean, which doctor is working.", "569500": "Hi, @serg132003 there is just more than agreeing in some common classification rules, this article explains why the algorithm was created.\nhttps://ai.googleblog.com/2016/11/deep-learning-for-detection-of-diabetic.html\n\n\"Interpreting these photographs requires specialized training, and in many regions of the world, there aren’t enough qualified graders to screen everyone who is at risk. '" }, "source": "meta" }