{
  "fetched_at": "2026-10-19T08:36:30Z",
  "source_url": "https://www.kaggle.com/competitions/glioma-mcd-2025/data",
  "pages": [
    {
      "name": "data-description",
      "content": "The target cohort for the Glioma-MDC 2025 challenge consists of patients diagnosed with glioma, specifically those who have undergone surgical resection or biopsy. The data used in the final biomedical application would be acquired from human glioma tissue samples, typically stained with Hematoxylin and Eosin (H&E) for histopathological examination.\n\nThe challenge cohort for the Glioma-MDC 2025 challenge consists of histopathological image patches obtained from glioma tissue samples. These samples were sourced from patients treated at Beijing Tiantan Hospital,  Sanbo Brain Hospital and supplemented with data derived from The Cancer Genome Atlas (TCGA) database.\n\n**All of the data provided for the Glioma-MDC 2025 challenge is licensed under CC BY-NC-SA (Attribution-NonCommercial-ShareAlike)**.\n\nParticipants should read the data description before starting:\n\n## Files\n\n*   **Glioma_MDC_2025_training.zip** - compressed zip file contains following two types of training data, 1,439 annotated images in total\n*   *training[4-digit-ID].jpg* - the training HE slide image with`512` by `512` pixels\n*   *training[4-digit-ID].json* - the training annotation with HE slide ID correlated to image\n*   **Glioma_MDC_2025_testing.zip** - compressed zip file contains following two types of testing data, 200 annotated images in total\n*   *testing[4-digit-ID].jpg* - the testing HE slide image with`512` by `512` pixels\n*   *testing[4-digit-ID].json* - the testing annotation with HE slide ID correlated to image\n*   **Submission_sample.csv** - a sample submission file in the correct format\n\n## Columns of `Submission_sample.csv`\n\n*   `Image ID` - Image ID for the HE slide\n*   `Label ID` - Location ID for the HE slide\n*   `Prediction` - Prediction of labels in specific location of an HE slide. If the cell is identified as mitosis, a 1 is filled in the corresponding position in the result table. If it is \"Non-mitosis\", a 0 is filled in the corresponding position.\n\n## Important notes\n\n* Each image may have multiple ROIs (Region of Interests), and the number are not always the same. We have annotated all of these ROIs so you don't have to, the task of this challenge is to identify the label of every ROIs: `Mitosis` or `Non-mitosis`\n* The ROI is circled by polygon shapes. For example in the `training0001.json` file you will see:\n```\n\"shapes\": [\n        {\n            \"label\": \"Non-mitosis\",\n            \"points\": [\n                [\n                    262.3829787234043,\n                    187.56028368794324\n                ],\n                [\n                    269.4751773049645,\n                    187.56028368794324\n                ],\n                ...\n            ],\n            \"group_id\": null,\n            \"description\": \"\",\n            \"shape_type\": \"polygon\",\n            \"flags\": {},\n            \"mask\": null\n        }\n    ]\n```\nThis location is labeled as `Non-mitosis` circled by the points in x-axis and y-axis coordinates. To view the annotation on top of the image, you can download open-source software [*labelme*](https://github.com/wkentaro/labelme). The following is an example view of `training0001.jpg` with annotation `training0001.json`. ![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F22565116%2F279d08a490084983024036bb800e0394%2Fexample.jpg?generation=1728854127994477&alt=media)\n* The label in the annotation files of the **testing** set is `Blank[position-ID]`, your task is to predict the label. For example in the `testing0001.json` file you will see:\n```\n \"shapes\": [\n        {\n            \"label\": \"Blank1\",\n            \"points\": [\n                [\n                    303.23756906077347,\n                    229.75690607734808\n                ],\n              ...\n            ],\n            \"group_id\": null,\n            \"description\": \"\",\n            \"shape_type\": \"polygon\",\n            \"flags\": {},\n            \"mask\": null\n        },\n        {\n            \"label\": \"Blank2\",\n            \"points\": [\n                [\n                    164.0110497237569,\n                    243.56906077348066\n                ],\n         ...\n            ],\n            \"group_id\": null,\n            \"description\": \"\",\n            \"shape_type\": \"polygon\",\n            \"flags\": {},\n            \"mask\": null\n        },\n        ...\n    ],\n```\nYou will see there will be `Blank1-6` in the JSON file, which means there are **6** ROIs to be predict in total. If the cell is identified as mitosis, a 1 is filled in the corresponding position in the result table. If it is \"Non-mitosis\", a 0 is filled in the corresponding position. **Make sure filling the prediction accordingly in the `Submission_template.csv` file**.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "🏆🎉 Congratulations, Competitors! 🎊👏",
      "content": "We hope you are enjoying the competition!  \n\nWe are excited to announce that this section will be hosted at the upcoming **ISBI 2025 conference**, taking place from **April 14–17** in **Houston, TX, USA**. You can find more details about the conference [here](https://biomedicalimaging.org/2025/), and we look forward to seeing you there!\n\n📝 We also encourage you to submit a **challenge paper** to ISBI. For submission guidelines and details, please visit the [challenge submission page](https://biomedicalimaging.org/2025/challenges/).\n\n📢 To help us stay in touch regarding your submission and potential trip, kindly fill out this survey:  \n👉 [https://forms.gle/u4AH18GK8yhHA8Ly5](https://forms.gle/u4AH18GK8yhHA8Ly5)\n\n🏆As part of the final evaluation, we are hosting an additional one-shot testing phase to assess your best-performing model on a held-out test set. The results will help determine another rankings and prize eligibility.\n👉 [https://www.kaggle.com/competitions/glioma-mdc-2025-one-shot-testing/overview](https://www.kaggle.com/competitions/glioma-mdc-2025-one-shot-testing/overview)\n\nThank you, and we look forward to your participation! 🚀",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "⚠️Important Announcement ⚠️",
      "content": "We have updated our dataset to version **Data_122824**. Please use the most up-to-date version to train your model.\nWhat has been changed in this version:\n1. Re-annotated unclear images for better training results.\n2. Fixed the unmatched JSON and JPG files.\n3. Added one more test sample (you may need to use the most up-to-date *Submission_template.csv*).\n\n**The data format and evaluation metric remained the same. You don't have to worry about changing your model structure.**",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "rules",
      "content": "For the Glioma-MDC 2025 challenge, only fully automatic methods are allowed. All submitted algorithms must operate autonomously without any user intervention during the detection and classification process. From image input to final output, the entire workflow should be handled by the algorithm itself, ensuring consistency and fairness in the evaluation of different approaches. This requirement is in place to advance the development of robust, fully automated solutions in digital pathology. You are required to opensource your model once the competition is completed.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "abstract",
      "content": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F22565116%2F3685e89765051e84801ec24c05c12e1e%2FGBM.jpg?generation=1733852704895305&alt=media)\n\nThe Glioma-MDC 2025 challenge aims to advance the field of digital pathology by developing robust algorithms for the detection and classification of mitotic figures in glioma tissue samples. Gliomas, the most common and aggressive primary brain tumors, have a dire prognosis, making accurate and timely diagnosis crucial for effective treatment. A key indicator of glioma aggressiveness is the rate of cellular proliferation, which is often measured by identifying and counting mitotic figures in histopathological images. \n\nFrom a biomedical perspective, detecting mitotic figures, especially those that are abnormal, provides essential information for tumor grading and prognostication. Traditional methods rely on manual counting by pathologists, which can be time-consuming and prone to variability. Automating this process using advanced image analysis techniques could significantly enhance diagnostic accuracy, consistency, and efficiency, ultimately leading to better patient outcomes. \n\nFrom a technical perspective, the challenge involves the development and validation of machine learning and computer vision algorithms capable of accurately identifying mitotic figures in high-resolution images of H&E-stained glioma tissue. Participants will work with a dataset of annotated image patches, focusing on the detection and classification of mitotic figures. The challenge will test the ability of these algorithms to generalize across different image patches and maintain high accuracy in detecting mitoses, especially in challenging cases with abnormal mitotic figures. \n\nThe envisioned impact of the Glioma-MDC 2025 challenge is twofold: biomedically, it will contribute to more precise glioma grading, aiding in the personalization of treatment plans and improving prognostic accuracy. Technically, it will push the boundaries of current image analysis methodologies, fostering innovations that can be applied to other domains within digital pathology and beyond. ",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "In the Glioma-MDC 2025 Challenge, a *case* is defined as a single histopathological image patch of glioma tissue, with multiple cases provided for analysis. Each *case* includes **varied locations** represented by coordinates and their **corresponding labels**. The objective is to develop a model capable of accurately predicting labels in these pre-annotated regions.\n\n## Training Cases\nEach training case comprises a stained image patch accompanied by a JSON file detailing the coordinates identifying specific cell locations. These cells are labeled as either *\"Non-mitosis\"* or *\"Mitosis\"*, providing the ground truth necessary for model training.\n\n## Test Cases\nEach test case also includes a stained image patch and an associated JSON file with multiple coordinate locations. However, the labels for these coordinates are left *blank*. Challenge participants must analyze these unlabeled coordinates and classify each one as either *\"Non-mitosis\"* or *\"Mitosis\"*.\n\n\n### Example of imaging\nShowing training0001.jpg. Each H&E slide is standardized as 512-by-512 pixels `jpg` image.\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F22565116%2F0819052a29a3dc7f0752c39dea804af4%2Ftraining0001.jpg?generation=1728408278142408&alt=media)\n\n### Example of annotation in JSON file\nShowing training0001.json annotation file.\n- `label` is the classification label\n- `points` are the polygon points in pixel coordinate\n```\n{\n  \"version\": \"5.4.1\",\n  \"flags\": {},\n  \"shapes\": [\n    {\n      \"label\": \"Non-mitosis\",\n      \"points\": [\n        [\n          262.3829787234043,\n          187.56028368794324\n        ],\n        [\n          269.4751773049645,\n          187.56028368794324\n        ],\n        ...\n      ],\n      \"group_id\": null,\n      \"description\": \"\",\n      \"shape_type\": \"polygon\",\n      \"flags\": {},\n      \"mask\": null\n    }\n  ],\n  \"imagePath\": \"training0001.jpg\",\n  \"imageHeight\": 512,\n  \"imageWidth\": 512\n}\n```\n",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "The primary metric used to assess the performance of the algorithms in the Glioma-MDC 2025 challenge is the F1 Score. The F1 Score provides a balanced measure of the algorithm's detection accuracy by combining both precision (the proportion of correctly identified mitotic figures out of all detected figures) and recall (the ability to identify all true mitotic figures). It is particularly useful in scenarios where there is an imbalance between the number of true positives and false positives/negatives.\n\n- Precision (Positive Predictive Value): The proportion of true positive mitosis detections among all detected mitoses. \n\n- Recall (Sensitivity): The proportion of true positive mitosis detections among all actual mitoses in the ground truth. \n\n- F1 Score: The harmonic mean of precision and recall, which provides a single measure of the detection accuracy. \n\nThis metric will be the primary determinant in ranking the participating algorithms, reflecting their overall ability to accurately detect and classify mitotic figures within the provided glioma tissue images. Additionally, precision and recall will be analyzed separately to provide further insights into each algorithm's classification accuracy and detection sensitivity, respectively. \n\n\n## Submission File\nWe have provided a test outcome table `Submission_template.csv`, please fill in the column prediction with your predicted labels (Mitosis: 1; Non-mitosis: 0). \n(**Note**: Each image in the test set corresponds to a JSON file, and the JSON file contains cell coordinates marked with \"Blank1\", \"Blank2\" and so on.)\n\ne.g.,\n```\ntesting0001,Blank1,0\ntesting0001,Blank2,1\n...\n```",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Organizer",
      "content": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F22565116%2Fcfca720a0ebb7ad4248299f5793c950c%2Forganizers.png?generation=1729607815506606&alt=media)",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}