{
  "fetched_at": "2026-10-19T08:34:28Z",
  "source_url": "https://www.kaggle.com/competitions/breast-density-prediction/data",
  "pages": [
    {
      "name": "data-description",
      "content": "Your challenge in this competition is to predict the breast density from the mammogram images. \n\n## Files\n\n* **'train' folder** - Comprises three sub-folders: 'images', 'breast_masks', and 'dense_masks'. The 'images' sub-folder houses the original images. The 'breast_masks' and 'dense_masks' sub-folders contain the ground truth segmentation masks for the breast area and dense tissue segmentation, respectively. All images are in JPG format. All masks and the corresponding images have the same dimension.\n\n* **'test' folder** - Contains the images for test set in JPG format. No ground truths are provided for the test set.\n\n## File lists\n*   **train.csv** - The training set filelist consists of two columns. The first column is the ‘Filename’, and the second column is the ‘Density', the ground truth for the breast density prediction task.\n\n*   **test.csv** - The test set filelist contains the filenames of the test sets. \n\n*   **sample_submission.csv** - A sample submission file in the correct format. Note that the density column has arbitrary numbers!\n",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "Breast density, a measure of the proportion of dense tissue to the total breast area, is a significant risk factor for breast cancer. Precise segmentation of dense tissues and the overall breast area is vital for calculating breast density. While semi-automatic and fully automatic computer-aided design tools have been developed to estimate breast density percentages in mammograms, these tools often fall short. They are typically limited to specific mammogram views and struggle with complete delineation of the pectoral muscle. Furthermore, these tools often perform poorly in cases of data variability and frequently require an experienced radiologist to adjust the segmentation threshold for dense tissue within the breast area. This challenge calls for the development of a new deep learning architecture that can automatically estimate the area-based breast percentage density from mammograms, overcoming the limitations of current approaches.\n\n## How to solve the problem\n\nThe task of predicting breast density can be approached in two ways. One approach is to treat it as a regression problem, where the goal is to develop a model that, given an input mammogram, outputs a density value in the range of 0-100:\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F17866327%2F44610acb81f4e92a3cf907681176e12c%2Fregression_model.png?generation=1701085628213273&alt=media)\n\nThe other, and often preferred, approach is to treat it as a segmentation task. In this case, two separate segmentation models need to be developed. The first model is used to segment the breast area, and the second is used to segment the dense tissues. The breast density is then calculated by dividing the sum of the dense-tissue segmentation binary mask by the sum of the predicted breast-area segmentation mask, and multiplying the result by 100. This gives the percentage of the breast area that is composed of the dense tissues:\n\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F17866327%2Feb159ee1aae90a0b2ad39216060d75ad%2Flagrida_latex_editor.png?generation=1701084917247253&alt=media)\n\n'N' refers to the total number of pixels present in each of the predicted segmentation masks for the breast area and dense tissues, individually, containing the white pixels only.\n\nOne way to implement the segmentation approach is to use a multi-task learning approach, where a model is trained to predict both the breast area and the dense tissues simultaneously, as proposed by [Gudhe et. al](https://www.nature.com/articles/s41598-022-16141-2), depicted below.  The architecture consists of an encoder, a bottleneck, and task-specific decoders. The encoder extracts low- and high-level imaging features from the input mammogram, which are then processed by the bottleneck unit to extract more complex and spatial information at different resolutions. The task-specific decoders simultaneously segment the breast area and dense tissues. The loss function, focal Tversky loss, is computed using the predicted and ground-truth segmentations of the breast area and dense tissues. A modified weight adaptive multi-task loss function is used for the segmentation task, computing the combined loss of the breast-area and dense-tissue segmentation tasks. The predicted segmentation outputs are overlaid on the original mammogram, with the predicted segmented breast area shown as a red contour line and the predicted fibroglandular tissues within the breast area represented by solid green pixels.\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F17866327%2F466b94eb715fd115671c1c20dd1a4e46%2F41598_2022_16141_Fig4_HTML.webp?generation=1701083384940700&alt=media)\n\n# Important Note\n\nIn this challenge, participants will encounter differences between ground truth density labels and the values computed using the dense tissue/breast area formula. If they choose a segmentation approach to predict density values, they should consider this discrepancy. To address it, participants can modify the loss function to include three terms:\n\n1. **L_dense_tissue_segmentation**: Loss related to dense tissue segmentation.\n2. **L_breast_area_segmentation**: Loss related to breast area segmentation.\n3. **L_(Y-\\hat Y)**: Loss corresponding to the difference between predicted and actual breast density value.\n\nHowever, if participants opt for a regression-based method—where they predict density values directly from input mammograms—there is no need to account for segmentation losses.\n\n# Github\nThe source code for the segmentation approach (only considering the segmentation losses) is available in the provided [Github repository](https://github.com/uefcancer/Deepdensity). Participants are encouraged to utilize and enhance this model for the challenge density estimation task.\n",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "\nSubmissions are evaluated on [Mean Absolute Error](https://en.wikipedia.org/wiki/Mean_absolute_error) between the predicted breast density and the observed density value. The lower the error, the better the model.\n\n\n## Submission File\nFor each Filename in the test set, you must predict a Density value between 0-100. The file should contain a header and have the following example format:\n\nFilename,Density\n03ea8b1445b94e3084e5442c49a3bbff.jpg,15\n03fd56d2b8ad85b9f1eddc9d07fa5c3c.jpg,35\n03ffb4a465bbf5ed4fcd1f3984a22e4c.jpg,15\n04a3e79996d3ea13d264d995e25c91c1.jpg,55\netc",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "rules",
      "content": "Sure, here are some suggested rules for your Kaggle competition:\n\n1. **Eligibility**: The competition is open to participants worldwide, except where prohibited by law. Participants must be at least 18 years old.\n\n2. **Data Usage**: The data provided for the competition should only be used for the purpose of this competition. Any other usage is strictly prohibited  unless a proper citation to the dataset and our studies are given., \n\n3. **Submission**: Each participant or team can submit a maximum of 5 entries per day. The final submission that will be considered for evaluation will be the last one submitted before the deadline.\n\n4. **Collaboration**: Participants are allowed to form teams. However, once a team is formed, it cannot be dissolved. Team mergers are not allowed.\n\n5. **External Data**: Participants are permitted to use external mammogram datasets for training their models. However, it's important to note that these datasets must be publicly available and accessible and mentioned by the participants.\n\n6. **Reproducibility**: The top 3 winning solutions must be reproducible. Participants should provide enough documentation and code for the reproducibility of their solution.\n\n8. **Code Sharing**: After the competition, the top 3 winners are encouraged to share their code and approach in the Kaggle Kernels for the benefit of the community.\n\n9. **Disqualification**: Any participant found violating these rules will be disqualified from the competition.\n\nHave fun!",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "abstract",
      "content": "Welcome to the **Breast Density Estimation challenge**! This competition calls on participants to create a model capable of accurately predicting breast density from a dataset of mammogram images.\n\nThe goal of this challenge is to push the boundaries of medical imaging technology and enhance the precision of breast cancer screening by providing a more accurate assessment of breast density. Breast density is a significant risk factor for breast cancer, the most commonly occurring cancer in women. However, due to the shortage of expert radiologists and the extensive time and effort required to examine mammograms on a large scale, there is a need for an automated solution. Furthermore, the assessments of breast density by radiologists can be subjective, leading to variability between and within readers.\n\nBy participating in this challenge, you will be contributing to a solution that could potentially lead to earlier detection of breast cancer and prevention. You will be training your models on a dataset of 569 mammogram images and testing their performance on a separate set of 149 images. The challenge welcomes a range of approaches, including both regression and segmentation methods.\n\nThis challenge presents a great opportunity for participants to apply their deep learning and image analysis skills to a real-world problem with significant public health implications. Best of luck to all participants!",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Winners ",
      "content": "We extend an invitation to the top three teams on the Leaderboard to collaborate with us in writing a manuscript. This will detail the results of the breast density estimation challenge and will be submitted to a highly-ranked conference.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Use of external datasets",
      "content": "Participants are permitted to use external mammogram datasets for training their models. However, it's important to note that these datasets must be publicly available and accessible and described with the submission.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Citing studies",
      "content": "If you utilize the data created for this challenge in your research, please cite the following studies:\n\n1. Gudhe, N.R., Behravan, H., Sudah, M. et al. (2022). Area-based breast percentage density estimation in mammograms using weight-adaptive multitask learning. Scientific Reports, 12, 12060. [DOI: 10.1038/s41598-022-16141-2]\n\n2. Behravan, H., Gudhe, N.R., Okuma, H., & Mannermaa, A. (2023). Mammogram Density Assessment Dataset. Mendeley Data, V2. [DOI: 10.17632/tdx3h2fn9v.2]\n\n3. Nguyen, H.T. et al. (2022). A large-scale benchmark dataset for computer-aided diagnosis in full-field digital mammography. [DOI: 10.1101/2022.03.07.22272009] ",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Acknowledgements",
      "content": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F17866327%2F536218490c31801ccffa5fd335a85b6a%2Flogos.png?generation=1701353892200710&alt=media)",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}