Healthcare faces ethical challenges when it comes to handling patients' medical records. They contain sensitive information about patients that can be prejudicial if leaked. Researchers go through time consuming procedures to request and use
medical data-sets, often at the expense of efficiently advancing research. This situation is exacerbated for data scientists who use large, and heterogeneous data-sets scattered across different hospitals, research labs and pharmaceuticals.
At the forefront of this data ethical challenge is data privacy, transparency and usage consent. We are proposing a framework that collects data from different sources and trains a global deep learning model using blockchain based federated
learning in a decentralized manner, that build trust and encourages industry collaboration without data silos. Here, blockchain authenticates the data and federated learning trains the model globally while preserving the privacy of the
organization. This technology holds great promises for precision medicine – tailor made treatments for patient sub-groups, which can only progress through large and multi-modal datasets. Our solution creates full transparency in a trustless
environment and unlocks opportunities for collaborative research in life science.
Electronic Health Records (EHR)
Blockchain has been heavily applied in the medical domain whereby the properties it possesses (decentralization, immutability, and transparency). One such application which has been studied extensively is Electronic Health Records
(EHR) management. EHR is inherently decentralized as stakeholders are distributed between patients, medical institutions and government institutions in some cases. Blockchain-based solutions, allow the different stakeholders to
manager EHR transparently while guaranteeing fairness and usage (records access) consent
CT Scan of Covid-19 patients
We believe blockchain integrated federated learning can support the detection of Covid-19 patients using lung screening as hospitals can share their private data in training the models. This solution will empower academic and pharmaceutical
industry researchers together in a federated research environment , in order to gain better insights from the breakthrough in the COVID-19 research network that will ultimately result in better treatments for patients, that is
accelerated and produced at a lower cost. Biotechnology and pharmaceutical companies will have better understanding on why drug efficacy for COVID-19 varies from patient to patient, enhance the drug development process and quickly
identify the best drug for the right patient at the right time.
Accelerating drug discovery
Our Terracuda.Ai solutions (federated learning + blockchain orchestration platform) is seeking to provide a modeling platform that can quickly and accurately predict promising drug discovery development, all without sacrificing the
data privacy of the participating companies. In the hope that this technology will advance the way industry discovers , develops and manufacturers medicines. Federated learning + blockchain integration approach will create a win-win
scenario by enabling secure participant collaboration with local, competitive data in order to create global knowledge that benefits humanity
Data are trained on distributed nodes (across hospitals) without compromising use privacy
Models are combined from different hospitals with a collaborative task. The goal is to utilize federated learning to share the data among the hospitals with privacy leakage
We train the global model without privacy leakage, then we alter a small part of the randomized mechanism through (a) Random sampling (b) Distorting. The random sub-sampling model will receive the final weight and share the data
locally, we then distribute the weights across the hospitals. The updated weights will be processed across the decentralized blockchain network with every round
Real data is stored at the hospitals, when a new hospital provides the data, it stores a transaction in the block to verify the owner of the data
This solution allows for data sharing retrieval requests. Multiple hospitals can collaboratively share the data and train the model to predict optimal results.
This solution supports the acceleration of precision medicine for drug discovery, through a privacy preserved data collaboration, across multiple hospitals
Terracuda.Ai integrated federated learning + blockchain for life science in healthcare
Privacy & Security
Datasets need not to be transferred to a centralized location for training the algorithm when federated learning is applied. Data privacy is guaranteed
Blockchain enables traceability of data models, results, validations and auditing to build trust across healthcare value chain organizations as new use cases becomes feasible
Higher quality of data (high volume of data can become selective and apply data provenance on blockchain to filter quality data). Significantly lowers the inference error rate
With both privacy & security solution in place, this unlocks unprecedented opportunities for collaborative research that supports time to market for drug discovery