MedDAO

The scarcity of medical images for training AI models poses a significant barrier to the progress of medical research and the development of precise AI-driven healthcare solutions. Without access to diverse and representative datasets, AI models struggle to accurately diagnose diseases, analyze medical conditions, and provide personalized treatment plans. This shortage impedes researchers' ability to unlock the full potential of AI in the healthcare domain, hindering advancements in diagnostics, treatment, and patient outcomes. MedDAO is an innovative decentralized autonomous organization that aims to address this critical issue of the global shortage of medical images required for training AI models in the healthcare domain. By revolutionizing the paradigm of patient-researcher interactions and transforming the landscape of medical data accessibility, MedDAO provides a comprehensive solution to these challenges. Let's discuss the features of MedDAO step-by-step:

• Incentivization: MedDAO incentivizes patients to contribute their medical data by offering a fair compensation mechanism. Patients receive monetary rewards for selling their medical files, encouraging their active participation in building comprehensive datasets. Patients are able to see the compensation price per submission which they can claim after their submission is approved by the aggregator. This incentivization model fosters the creation of diverse and representative datasets, enabling researchers to train AI models effectively and achieve significant advancements in medical research and treatment methodologies.

• Aggregation: MedDAO incorporates a system of aggregators who play a crucial role in regulating the DAO ecosystem. These aggregators analyze the medical markets, assess researchers' demands, and optimize requests for medical files to form data sets or collections. By dynamically adjusting pricing structures based on factors such as data complexity and clinical relevance, aggregators ensure fairness and transparency in the marketplace. They also validate the quality and authenticity of files submitted to different datasets, maintaining the integrity of the platform. Researchers also have the option to schedule a decentralized video-call with aggregators to discuss custom dataset requests.

• Privacy and Security: MedDAO places utmost importance on protecting patient privacy and ensuring data security. To maintain confidentiality, researchers and companies are never granted complete access to the medical files. Instead, our platform offers a secure Bacalhau endpoint, where researchers can submit their algorithms or code for execution on the specified dataset. The endpoint securely executes the code and provides researchers with the results, without exposing the underlying medical images. Apart from this all data on MedDAO is end-to-end encrypted. These robust privacy-preserving mechanisms prevent unauthorized access and ensure data integrity.

The application is built on state of the art tech stack. The front end is built on nextjs framework. Patient, aggregator and researchers are authenticated via metamask. ENS is integrated here to show the equivalent ens name of the authenticated user. The smart contracts are deployed on ethereum testnet in the absence of a working and compatible filecoin test net. The smart contracts are integrated with the front end using etherjs.

Files uploaded by patients is encrypted by lit protocol and pushed to spheron. All notifications are implemented via integration with push protocol. The decentralized compute platform is offered by bacalhau.

Bacalhau offers a client that runs on an user's system to interact with the bacalhau network. Unfortunately they do not have a Js sdk. Therefore, for the sake of the hackathon and to have an always available programmable endpoint, we took the standalone client and wrapped it in express js to provide an accessible API endpoint.

The patient uploads his/her medical images and incentivised by the aggregator. The marketplace section of the app connects the researcher with the aggregarors's offering. The platform makes running AI jobs simple and affordable.

The researcher needs to wrap his/her algorithm / code in a docker container and supply the absolute command to the platform. The command is executed on the bacalhau network by reading the content from an input directory pointed to by the cid of the data set and stores the results in the output dir. Bacalhau copies over the output data to ipfs and returns a cid for access by the researcher. The bacalhau network is currently public hence the invocable compute client is hosted on Google cloud.

The programmable endpoint is basically the glue that connects the pieces of the puzzle together forming a completely decentralized compute platform for medical research.

Finally, the entire app is hosted on spheron.