Name of Project
Healthcare_data_guild (Secure, private guild / bubble for aggregate healthcare data stakeholders)
Proposal in one sentence
Allow a private dataset, resembling aggregated healthcare data, to be stored on Filecoin Plus, enclose it in a privacy ‘bubble’, publish metadata on the Ocean marketplace, and provide the output of a relevant compute-to-data algorithm run on this dataset to an authorized Ocean marketplace customer.
Description of the project and what problem is it solving
Problem to be solved
Under US law (HIPAA, Hitech Act), healthcare companies involved in direct patient care are required to maintain data for every patient of theirs. Storing certain kinds of medical data indefinitely / for long periods of time is expensive, especially if it generates no returns for money spent on storage. Furthermore, this data is locked within individual company data silos. This is true for several companies that are part of the intra-operative neuromonitoring (IONM) industry in which the project lead (ASG) has worked for several years. Each individual company has a dataset that by itself is inadequate for machine learning algorithms, but several such private datasets together can provide valuable insights. Yet these companies are unable to share data between themselves since this might provide crucial business intelligence to their competitors.
Solution
We propose an access-controlled guild or bubble, based on self-sovereign IDs, of various stakeholders of aggregated healthcare data. Within the bubble, fine grained permissions will allow private data storage, analytics and monetization of individual company owned private datasets. The datasets will remain private, yet be subject to ML algorithms provided by authorized ML algorithm vendors and algorithm outputs will be consumed by authorized consumers all of whom are part of the bubble. This will allow a central focus on patient privacy, data provenance and traceable, controlled compute-to-data within the Ocean marketplace.
A similar federated learning solution called Melloddy, but based on centralized data storage at AWS, is currently being piloted for large pharmaceutical companies.
Grant Deliverables
[ ] upload example dataset (e.g. containing upto 14 features) to Filecoin Plus
[ ] enclose dataset in an additional privacy bubble smart contract controlled by data provider. The privacy bubble smart contract will be provided by Datona Labs Ltd (UK).
[ ] publish the dataset’s discoverable metadata on the Ocean Marketplace
[ ] publish a relevant compute-to-data algorithm (e.g. a supervised learning algorithm to predict insurance status of patient) to the Ocean marketplace
[ ] enable approval, by healthcare_dataguild or the dataset provider, of a third party customer / algorithm provider through the privacy bubble.
[ ] run compute-to-data on the dataset and provide output to the approved Ocean marketplace customer
Which category best describes your project?
Build / improve applications or integrations to Ocean
What is the final product?
A proof-of-concept of a storage and monetization paradigm using compute-to-data on private aggregate datasets within a specialized guild / bubble of healthcare data providers, analysts and consumers.
Question on “value add” criteria: which one or more of the criteria will your project focus on? Why do you believe your team will do well on those criteria?
Usage of Ocean - A long-term goal of ours is to unlock value in US healthcare data silos with Ocean Protocol based technology. As a first step, we wish to create a proof-of-concept of a paradigm to store, analyze and monetize these datasets for their aggregators and ultimately provide control to each patient over their own data.
The largest neurophysiology companies provide services in ~ 100,000 surgeries a year, and the combined case coverage for companies based in the US is ~ 750,000 cases. Although the total case volume is large, the types of cases (e.g. neck, thorax, low back, brain aneurysms, tumors etc) being covered are many and the patient population is varied. Hence an individual company lacks the ability to collect sufficient data for machine learning.If a successful data guild of stakeholders around these datasets can be formed, multiple data providers, analysts and consumers will be onboarded to the Ocean Marketplace contributing to its success.
Community - By establishing this paradigm of a data guild of stakeholders for private, regulated, healthcare data on the Ocean Marketplace, we will enable the Ocean community to develop similar solutions in other verticals including other areas of healthcare as well as other regulated and privacy sensitive industries (e.g. legal). We expect the Ocean, as well as, the larger AI/ML community will develop several novel algorithms for this guild and drive more value to the Ocean marketplace.
Using their experience and connections in the surgical neurophysiology field, the built-out core team will network with large and small surgical neurophysiology company executives to deploy the service as a long-term goal. The core team has a deep understanding of the vertical allowing them to proactively respond to the customer needs and specifications. The US Northeast, where the project lead is based, is at the forefront of this field and sets the ‘standard of service’ for other regions of the country.
Funding Requested
3000 USD
Proposal Wallet Address
0x4fDCCF789B8631110A942AD1A8663cE054846e21
Have you previously received an OceanDAO Grant: No
-Discord Handle (if applicable): @mnkyntigr
-Project lead Contact Email: nomesg2020+OceanDAO@gmail.com
-Country of Residence: United States
Part 2 - Team
Core Team
Ambarish S. Ghatpande, PhD, CNIM
- Role: project lead / coordinator
Relevant Credentials:
-
Github: https://github.com/aghatpande
Experience and Background
-
Statistical analysis and machine learning using R/RStudio 2013 - present
-
Surgical neurophysiologist III with SpecialtyCare 2016 - 2021
-
Surgical neurophysiologist Sentient Medical Systems 2015 - 2016
-
Senior Research Scientist Lupin Ltd 2014
-
Research positions at University of Colorado, Monell Chemical Senses Center and University of Maryland (2000-2013)
-
PhD, dissertation in ion channel biophysics 1993 - 2000
Datona Labs Ltd / Principal: David N. Potter
- Role: privacy solutions expert / coder
Relevant Credentials:
Background/Experience
A senior software engineer with a history of developing and assuring safety critical software in the rail and aviation industries. I’ve been developing blockchain solutions since 2013 and am currently building a decentralised privacy platform with associated DApps using linux, javascript, solidity, web3.js, figma, react and styled components and expertise in applied cryptography.
-
Founder & CTO Datona Labs 2019 - present
-
Independent software assessor for rail accident investigation team Hong Kong government 2018 - present
-
Technical authority Hitachi 2016 - 2020
-
Founder & CTO OpenSig 2016 - present
-
Independent safety assessor Kawasaki 2015 - 2019
-
Software safety assurance Ebeni 2012 - present
-
Independent safety assessor Lloyd’s Register Beijing - 2011
-
Technical authority Invensys Rail 2003 - 2011
-
Technical consultant EADS 2003
-
Technical consultant Silver Atena 2000 - 2002
-
Software consultant Ultra Electronics 1999
-
Software engineer Westinghouse Rail 1997 - 1999
-
Software engineer Ultra Electronics 1996 - 1997
-
BSc computer science University of York - 1996
We are scouting for academic neurophysiologists and senior executives in neurophysiology / healthcare / AI companies to be members of the core team.
Advisers
Ongoing discussions with Ocean Protocol grantees in the healthcare and AI space.
Proposal Details
Allied health care providers range from anesthesiology service provider companies all the way to environmental services provider companies in the United States. A vast army of these companies provide services to U.S. hospitals. Many different types of companies are involved in direct patient care and are required to store PHI (protected health information) data by law.
Surgical neurophysiology companies are an example of allied health provider companies. Large neurophysiology companies are unique in having nationwide coverage. Surgical neurophysiology companies provide their services for surgeries involving the brain, spinal cord, peripheral nerves and many types of vascular (involving heart, major blood vessels including those in the brain) surgeries. There are approximately 750,000 such surgeries performed annually in the United States and this case coverage is growing every year.
During surgery, the surgical neurophysiologist present in the operating room records neurophysiological data from the patient and runs tests at the request of the surgeon. This allows the surgeon to operate avoiding injury to neural tissue. For example, during brain tumor excision, a surgeon needs the neurophysiologist to guide him by identifying brain areas involved in movement, sensation, vital bodily functions (breathing etc). Large amounts of neurophysiological data are collected and needs to be stored as PHI by the company that employs the neurophysiologist.
The outcome of these types of surgeries are dependent on several features of the patient’s medical history as well as the intra-operative course of the surgery. Potentially, this data when subject to machine learning, will unlock new insights into surgical outcomes. Unfortunately, machine learning models are data hungry, and no individual company can generate sufficient data on the many different types of surgeries performed. They need to collaborate amongst each other but are wary of this due to competition.
We believe two key requirements need to be met to onboard healthcare data stakeholders. A legal mechanism to store and monetize the data, and, a collaborative mechanism that ensures the individual datasets remain private.
Our proposed paradigm includes an inexpensive, decentralized storage solution (e.g. Filecoin), absolute control of data in the hands of the individual data providing company with fine-grained layers of privacy. Furthermore, the providers should be able to publish their dataset attributes / metadata with compute-to-data permissions within the bubble. A data guild- / individual data provider- authorized customer should be allowed to run compute-to-data ML algorithms on these datasets and to access the predictive models / insights outputted by the algorithm.
Demonstrating assured privacy and security for patients, data aggregators, service providers and consumers using Ocean Protocol and Datona Labs technology will be the primary goal for this proposal.
Using inputs from these stakeholders and after consultation / collaboration with the Ocean Protocol developer community, the team will outline legal and realistic solution/s to these problems based on the paradigm described above.
Project Deliverables
Any prior work completed thus far?
The project lead has published two test datasets (vital_signs_1 & 2) which can be used for predictive models using supervised learning. The supervised learning algorithm is also working using R code and is able to predict a patient’s insurance status. We will publish this as an appropriate compute-to-data ML algorithm as part of our deliverables. Running these ‘useful in real-life’ predictive models do not require specialized compute infrastructure (e.g. GPUs) and are appropriate as a proof-of-concept. The project lead is also contributing to the Algovera AI hacking sessions actively to understand the intricacies of compute-to-data. Basic code for privacy bubbles is mature and will be appropriately customised.
What is the project roadmap?
The milestones will be worked on concurrently and delivery dates are an approximate projection
Filecoin Plus storage and retrieval of dataset - End of January 2022
Publishing a private dataset suitable for supervised learning algorithms and a relevant compute-to-data algorithm on Ocean Marketplace - Feb 2022
Custom ‘bubble’ to allow multiple variations of access and permissions for all types of stakeholders - Feb 2022
Authorized customer (Ocean community member for the proof-of-concept) purchases access to the dataset with permission to run the approved algorithm, runs the algorithm and takes delivery of the predictive model described earlier - March 2022
Once we have a proof-of-concept, we will approach neurophysiology companies for outreach and iterate on the paradigm using their inputs.
We hope to deploy our service when a good market-fit is achieved, perhaps, requiring multiple iterations of our paradigm along with other necessary developmental processes.
Final delivery of proof-of-concept paradigm
end of March 2022
Please include the team’s future plans and intentions.
We hope to successful implement this paradigm in the neurophysiology field as our major goal by end of 2022. This should kindle interest amongst other types of health provider companies (e.g. anesthesia providers that typically work closely with surgical neurophysiologists in the operating room) and stimulate their adoption of Ocean Protocol.
Our mission is to become a nationally accredited, specialized service providing healthcare companies with storage, analytics and monetization of their data on the Ocean Marketplace, while ensuring patient rights and privacy. Funds requested in this proposal will allow us to develop a proof-of-concept of such a service paradigm.
