The training data has only 2 labels Malignant and Benign, whereas Benign is really made up of several sub-categories such as Nevus, Lentigo, SK (Seborrheic Keratosis), BCC (Basal Cell Carcinoma), SCC (Squamous Cell Carcinoma) and others. This kind of information is present in external (older) data.
\n\n\"Diagnosis\" column of train.csv has these sub-category labels such as Lentigo, SK, etc.\n\"Diagnosis\" column is not present in test.csv (can itself be considered as a target for prediction). Moreover this column has more to do with images rather than patient_id context (patient specific details).
\n\nWhile using strictly image processing, my question is whether using diagnosis column will cause the submission to become ineligible for the Special Prize Without Context - $5,000 (Top-scoring model without using any patient-level contextual information)?
", "messages": [ { "id": "919875", "postDate": "07/08/2020 06:46:09", "content": "The training data has only 2 labels Malignant and Benign, whereas Benign is really made up of several sub-categories such as Nevus, Lentigo, SK (Seborrheic Keratosis), BCC (Basal Cell Carcinoma), SCC (Squamous Cell Carcinoma) and others. This kind of information is present in external (older) data.
\n\n\"Diagnosis\" column of train.csv has these sub-category labels such as Lentigo, SK, etc.\n\"Diagnosis\" column is not present in test.csv (can itself be considered as a target for prediction). Moreover this column has more to do with images rather than patient_id context (patient specific details).
\n\nWhile using strictly image processing, my question is whether using diagnosis column will cause the submission to become ineligible for the Special Prize Without Context - $5,000 (Top-scoring model without using any patient-level contextual information)?
", "rawMarkdown": "The training data has only 2 labels Malignant and Benign, whereas Benign is really made up of several sub-categories such as Nevus, Lentigo, SK (Seborrheic Keratosis), BCC (Basal Cell Carcinoma), SCC (Squamous Cell Carcinoma) and others. This kind of information is present in external (older) data.\n\n\"Diagnosis\" column of train.csv has these sub-category labels such as Lentigo, SK, etc.\n\"Diagnosis\" column is **not present in test.csv** (can itself be considered as a target for prediction). Moreover this column has more to do with images rather than patient_id context (patient specific details).\n\nWhile using strictly image processing, my question is whether using **diagnosis column** will cause the submission to **become ineligible** for the Special Prize **Without Context** - $5,000 (Top-scoring model without using any patient-level contextual information)?", "votes": null }, { "id": "920900", "postDate": "07/08/2020 22:36:53", "content": "Will the diagnosis column be used in any way that clusters patient_id's or considers the context of a patient's overall set of images? If not, then use of diagnosis is considered \"without context.\"
", "rawMarkdown": "Will the diagnosis column be used in any way that clusters patient_id's or considers the context of a patient's overall set of images? If not, then use of diagnosis is considered \"without context.\"", "votes": null }, { "id": "921008", "postDate": "07/09/2020 02:41:33", "content": "Thank you for your quick reply @juliaelliott.
\n\nNo, \"diagnosis\" column does not contain any information about the patient_id, just labels such as Lentigo, SK, etc
", "rawMarkdown": "Thank you for your quick reply @juliaelliott.\n\nNo, \"diagnosis\" column does not contain any information about the patient_id, just labels such as Lentigo, SK, etc", "votes": null } ], "comments": [ { "id": 920900, "author_name": "juliaelliott", "author_url": "", "post_date": "07/08/2020 22:36:53", "content": "Will the diagnosis column be used in any way that clusters patient_id's or considers the context of a patient's overall set of images? If not, then use of diagnosis is considered \"without context.\"
", "votes": null, "replies": [ { "id": 921008, "author_name": "sirishks", "author_url": "", "post_date": "07/09/2020 02:41:33", "content": "Thank you for your quick reply @juliaelliott.
\n\nNo, \"diagnosis\" column does not contain any information about the patient_id, just labels such as Lentigo, SK, etc
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "919875": "The training data has only 2 labels Malignant and Benign, whereas Benign is really made up of several sub-categories such as Nevus, Lentigo, SK (Seborrheic Keratosis), BCC (Basal Cell Carcinoma), SCC (Squamous Cell Carcinoma) and others. This kind of information is present in external (older) data.\n\n\"Diagnosis\" column of train.csv has these sub-category labels such as Lentigo, SK, etc.\n\"Diagnosis\" column is **not present in test.csv** (can itself be considered as a target for prediction). Moreover this column has more to do with images rather than patient_id context (patient specific details).\n\nWhile using strictly image processing, my question is whether using **diagnosis column** will cause the submission to **become ineligible** for the Special Prize **Without Context** - $5,000 (Top-scoring model without using any patient-level contextual information)?", "920900": "Will the diagnosis column be used in any way that clusters patient_id's or considers the context of a patient's overall set of images? If not, then use of diagnosis is considered \"without context.\"", "921008": "Thank you for your quick reply @juliaelliott.\n\nNo, \"diagnosis\" column does not contain any information about the patient_id, just labels such as Lentigo, SK, etc" }, "source": "meta" }