Project AiAi

AiAi.care project is teaching computers to "see" chest X-rays and interpret them how a human Radiologist would. We are using 700,000 labeled Chest X-Rays + Deep Learning to build an FDA 💊 approved, open-source screening tool for Tuberculosis and Lung Cancer. After an MRMC clinical trial, AiAi CAD will be distributed for free to emerging nations, charitable hospitals, and organizations like WHO 🌏

Our secondary goal is to open-source 'Medical Imagenet' pretrained models and weights. Medical data is hard to obtain, so many current Radiology-AI papers rely on transfer-learning ImageNet weights. There are significant differences between ImageNet images (color, low-res, high-contrast) and Radiology images (grayscale, high-res, low-contrast), so we believe that our Medical Imagenet weights will improve sensitivity and specificity across the board for future research.

If you are looking for a non-technical introduction to Project AiAi, please click here to visit https://AiAi.care website. In case you were wondering about the project's name, AiAi stands for AI Augmented Imaging.

Project Milestones

Here are major milestones for the project and target completion dates:

1. :white_check_mark: ~200,000 X-Ray and CT Data Loading, Cleaning (Completed Feb 1, 2017)
2. :white_check_mark: Deep Learning Docker Image (PyTorch, HIPAA) (Completed Feb 28, 2017 )
3. :white_check_mark: Data Augmentation Tests : March 30, 2017 (Completed May 2017)
4. :white_check_mark: Level 1 Models : April 12, 2017 (Completed May 2017)
5. :white_check_mark: HIPAA IT Audit and Validation : December 30, 2017 (Internal Audit Completed 2017)
6. :white_check_mark: We got :sparkles:500,000+:sparkles: additional X-ray images! Merging this with original data. (Completed May 2020)
7. :soon: Experiements with DL architectures, activations, and augmentation: (In Progress, ETA Q1 2021)
8. Level 2/3 model ensembles and differential privacy (Detection, Classification) : (In Progress, ETA Q2 2021)
9. Build mobile-friendly front-end for AiAi CAD (Computer Aided Detection) : Q2 2021
10. PACS / VNA / DICOM / HL7 / EHR ingestion engine Q3 2021
11. MRMC clinical validation for FDA application : Q3 2021

Donate your DL / FHIR / PACS Expertise

If you are a Machine Learning maestro, Kaggle king, or HL7 hacker then please check out our KANBAN project tracker here. You can donate your time and expertise by contributing to any of the issues/tasks pinned on the KANBAN board.

Contribute to Legal Strategy Wiki

Project AiAi is the first effort of its kind to donate an open-source, medical algorithm to the world. This presents some interesting legal challenges, so we have set up a wiki page where lawyers can donate their time and advice for Project AiAi.