Summary
Internationally Federated Platform for AI Research and Diagnostics
The development of AI and ML models for neuropathology is often hindered by the scarcity of high-quality, well-annotated training data. Insufficient data makes it challenging to train robust models, resulting in suboptimal performance and impeding progress in research. However, current digital pathology software makes these tasks complicated for researchers. The Digital Slide Archive (DSA) is an open-source, web-based platform for digital pathology. The DSA enables scientists and researchers to manage large collections of tissue images and combine them with clinical and genetic information. Because it’s open-source, the DSA can be improved and expanded to include new features.
The Brain Digital Slide Archive (BDSA) project aims to address these challenges by enabling the integration of AI into neuropathology by utilizing the DSA’s advanced tools. The BDSA is an international collaboration of research institutions, working to design a federated system for sharing neuropathology slide images and annotations. The project focuses on developing tools that facilitate seamless sharing of slide images between partner institutions and streamline the annotation process. These annotated datasets will inform AI models capable of accurately segmenting and detecting anomalies in neuropathology images, thereby accelerating research and innovation in the field.
The methodologies and tools developed through the BDSA project have applications far beyond neuropathology. The same federated systems and AI-powered approaches can be adapted for use in other domains that rely on large-scale image analysis and annotation, such as oncology, radiology, and other areas of medical imaging. Addressing foundational challenges in data sharing and annotation, the BDSA project has the potential to drive AI innovation across a broad spectrum of disciplines.
Datasets/Model
The data used for this project is restricted and not publicly available.
Access
Access to the data used for this project is restricted and not publicly available.
Ownership
This project is still in development.
This research was supported in part by the National Institutes of Health under award numbers U24NS133945-01 and U24NS133949-01. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.
Resources Utilized
CAAI Team members Mitchell Klusty, Vaiden Logan, Mahmut Gokmen, and Dr. Cody Bumgardner worked on BDSA.