RSNA Pneumonia Detection Challenge
Can you build an algorithm that automatically detects potential pneumonia cases?
New Kaggle competition! We are excited to share the dataset for this challenge, created on the MD.ai platform in collaboration with the Radiological Society of North America (RSNA), the Society of Thoracic Radiology (STR), the US National Institutes of Health (NIH), and Kaggle. Good luck!
Curating and annotating high-quality labeled datasets for machine learning training and validation
NLP-based Dataset Creation
Use powerful natural language processing based interactive querying to create datasets from unstructured archives and collections. Through our data partners such as NTT Data, we can provide access to more data than any single institution.
Label and annotate datasets collaboratively using our web-based app, with quality control and project management features. DICOM studies are managed using our custom-built mini-PACS, allowing labels and annotations to be scoped to the exam, series, or image level. Soon, you will be able to use previously trained models directly in the annotation workflow.
Collaborative Data Annotation
⬤ Alice created bounding box annotation
⬤ Bob created freeform annotation
The Cancer Genome Atlas - Lung Adenocarcinomapublic
From The Cancer Imaging Archive (TCIA): the Cancer Genome Atlas Lung Adenocarcinoma data collection is part of a larger effort to build a research community focused on connecting cancer phenotypes to genotypes by providing clinical images matched to subjects from The Cancer Genome Atlas (TCGA). Clinical, genetic, and pathological data resides in the Genomic Data Commons (GDC) Data Portal while the radiological data is stored on The Cancer Imaging Archive.
Working with data from organizations such as The Cancer Imaging Archive and others, we are building and annotating large public datasets with the goal of accelerating research and the application of AI in medicine.
Training and deploying clinically-relevant machine learning models
Deep Learning Model Training
Send annotated datasets for docker-based GPU training of deep learning models, using TensorFlow, Keras, or other frameworks. These trained models can be used further in the annotation workflow, analyzed for additional rigorous clinical validation, or prepared for productization.
Validate trained models on additional datasets. Being connected to the data creation and annotation process enables rapid model iteration and improvement.
Web-Based or Docker-Based
Deploy your trained models to the web or docker containers. Certain models can be deployed on our platform to be run entirely in the web browser -- study images do not need to be uploaded to any server for processing. Models may also be deployed to docker containers, with data transfer and security trade-offs to consider.