A Blockchain for Medical Imaging
Today we have a guest post by Dr. Lance Reinsmith about blockchain for radiology. He is a self described Python and deep learning enthusiast. He lives in San Antonio, Texas and works as a diagnostic radiologist at South Texas Radiology Group.
A lot of progress has been made recently for the healthcare blockchain with companies like Optum, Humana, PokitDok and Hashed Health leading the way. However, most of the focus thus has been on securing and sharing financial transactions and electronic health records.
I believe blockchain technology could also play an integral role in storage and distribution of medical images and eventually in the creation of a marketplace for the interpretation of medical imaging studies. This technology has the potential to solve problems of medical image storage and availability, patient health information security, and timely interpretation of medical imaging studies.
Blockchain: The Trust Machine
The blockchain is the next disruptive leap in global informatics. Simply put, blockchain is the construction of a decentralized and trustless means of interrelating in a world previously characterized by oversight and lack of trust. Thus, it is a HUGE paradigm shift from the way people have interacted and work together currently and in the past. Since its inception as the backbone of a global peer-to-peer autonomous and trustless currency, blockchain has demonstrated the potential to create decentralized contracts, applications, and entire organizations vital to global communication and commerce.
Bitcoin and Crypto-currencies
The first and most well-known iteration of the blockchain put into practice was the currency, Bitcoin. This is a decentralized currency not backed by the faith and credit of a nation. Rather, it is backed by the trust its users have in the cryptographic algorithms involved which are tested and shown to withstand any and all realistic attempts to crack them. I say “realistic” because someone could try, but such a brute force attempt would be so difficult that basically ANY other means of getting currency (including legitimately EARNING it) would be far easier. In fact, most or all of the nefarious “hacks” involving Bitcoin to this point have been the result of human error rather than a problem with the system itself.
Bitcoin transactions are notable for being secure, completely verifiable, and not controlled by any authoritative organization. Contrary to public misconception, Bitcoin is not anonymous or “secretive.” Instead, all of the information about EVERY Bitcoin transaction is not only viewable to the general public but creates a public ledger for auditing.
The second iteration of the blockchain are Smart Contracts. This is the new buzzword I feel we will hear more and more. At its heart, the blockchain is really just a publicly viewable ledger of transactions. These transactions are not limited to peer-to-peer currency exchanges. Information about the ownership of any property can be documented in the blockchain. The same verification algorithms used to verify Bitcoin transactions can thus be used to prove ownership or agreements between two or more parties.
Since it is possible to create an autonomous currency like Bitcoin, then it is also possible to use the same technology to create binding contracts, escrow transactions, third-party arbitration, or multiparty signatures that fulfill themselves instead of being enforced by a bank or governing body. Examples of smart contracts include a term life policy that automatically pays a death benefit on verification that a person dies or an auto loan that autonomously transfers a title to the owner when the principal is paid off.
In a properly designed smart contract, all property becomes smart property — it is encoded and bound to the blockchain in such a way that, using a unique identifier, it can be tracked, controlled, or exchanged.
Revolution Beyond Blockchain
The final element of the blockchain as it is envisioned today are Decentralized Autonomous Applications (DApps) and Decentralized Autonomous Organizations (DAOs). Today, these are mostly conceptual. However, the end goal would be a completely autonomously operating corporation that interacts with human clients just as a traditional corporation would.
Obviously, if not structured extremely well, DApps and DAOs would be doomed to failure because, short of supremely artificially intelligent computers, there is no prescient entity to foresee problems or correct errors in the system. This is exactly what happened with the DAO (a decentralized autonomous organization unfortunately named identically to the acronym for “Decentralized Autonomous Organization”), a venture fund that was supposed to have no employees and function democratically by the votes of its investors. In this case, an error in the design allowed a party to siphon money from the fund, and nothing could be done until it was too late.
Blockchain in Medical Imaging
In 2006, there were over 400 million radiology studies performed in the US alone. This number has increased, perhaps 10 fold. In addition, government regulations require most studies be saved for 10 years, pediatric studies for 10–18 years, and mammograms forever. Conservatively, there is a need to store at least 10 billion studies in the US alone, and this number is growing annually. If each study is 200MB (a low estimate), that is at least 2 petabytes of storage. Since studies are redundantly stored, the total amount of actual storage necessary is in the multiples of petabytes. If you consider that the United States comprises only 12% of worldwide imaging volume, the global storage requirement jumps 8-fold.
While this is a large number of radiologic images to store and audit, medical imaging is not only confined to diagnostic radiology; disciplines such as pathology, ophthalmology, dermatology, internal medicine, surgical sub-specialties, and dentistry are increasingly creating and storing digital images and videos.
At the same time that storage requirements increase, there is also rising demand for rapid accessibility of medical imaging data. This is due to government edicts and societal demands for quick access to information. Transferring image data on printed or optical media has become insufficient for today’s needs. Furthermore, the wide dissemination and accessibility of imaging data makes sense for an efficient medical ecosystem and sharing between healthcare organizations.
While storage and access to medical imaging data is a difficult enough technical problem, privacy regulations such as HIPAA have added additional complexity to any company developing medical imaging applications. And as data breaches at larger corporations such as Target, eBay, JP Morgan Chase, and even the Internal Revenue Service have eroded public trust in centralized storage of sensitive data, healthcare organizations are increasingly relying on their own hybrid cloud environments.
In summary, there is a need to store and disseminate large amounts of medical imaging data but bureaucratic and logistical obstacles as well as lack of incentives prevent an organization or organizations from doing so…until the blockchain.
Decentralized Autonomous Application for Medical Imaging
I propose to create a decentralized autonomous application for the storage and permission control of medical imaging studies using blockchain technology. Since diagnostic radiology is my area of expertise, I will focus on radiologic data, but the proposed DApp could be translated to any type of medical imaging.
Radiologic organizations (hospital radiology departments, outpatient imaging centers, etc.) generate asymmetric encryption key pairs, store their private keys, and disseminate their public keys. As the radiologic organizations (RO) create radiologic information (RI) including study images, scanned documents, reports, etc., the RI is encrypted using the RO’s private key.
Each study is given a randomly assigned ID, and the ownership of the study (the RI) is encoded as a transaction in blockchain. This is analogous to mining a Bitcoin; the chain must be checked to see if the study ID exists on the chain before adding it. (The Bitcoin blockchain itself is not likely appropriate due to block size conflicts and the burdensome 10 minute block time. Ethereum or a new chain would be more ideal.)
While the clinical information in the RI is owned by the patient in a legal sense, patients will be encouraged to allow the ROs to be the custodians of their RI much the same way it is done currently. Of course, a patient could revoke ownership from the RO, removing it from the radiology blockchain. But, there is no incentive to do this; the patient would lose the benefits of portability and security.
Once encoded in the chain, the encrypted RI becomes smart property that is disseminated in a BitTorrent-like fashion to the entire network in the same way the block ledger is. ROs may choose to keep a copy of their RI on their servers — though this is not obligatory. Ideally, RO servers don’t just hold their own studies; they store studies done all over the country — or the world. This is possible without needing to expand storage capacity because studies do not need to be backed up as they are extensively duplicated. Once RI has been disseminated the data is secure so long as the ROs private key is not compromised.
Patients and ROs can direct the flow of the RI by the creation and enforcement of smart contracts on the blockchain. For example, suppose a patient sees a new physician or is seen in an Emergency Room away from home. Assuming these new ROs (nRO) have created their own private/public key pairs, they submit a request to “lease” the patient’s RI for a period of time.
The blockchain ledger specifies the owner of the RI, the originating RO (oRO). The nRO sends its public key to the oRO. The oRO responds by returning a randomly-generated and time stamped symmetric key encrypted using the nRO’s public key. The oRO retrieves the imaging data from the disseminated store, decrypts it with their private key, re-encrypts it with the shared symmetric key, and digitally signs it before sending it to the nRO.
Once the nRO receives the cipher, it verifies the signature, proving authenticity of the data. Next, it decrypts the data using the shared symmetric key which it was able to decrypt using its own private key.
The transactions between the oRO, the nRO, and any intermediaries is documented in the blockchain. This is crucial because there is a specified lease time whereby the time-stamped shared key will cease to function when the lease expires. If a nRO attempts to decrypt a cipher with an expired symmetric key, the oRO is notified as the transaction shows up on the blockchain ledger. At this point, the oRO can chose to renew the lease or report the nRO for inappropriately accessing RI it does not own.
Those entities responsible for storing the RI and the oRO for retrieving and reencrypting the RI are paid a transaction fee — a stipulation outlined in the smart contract. This is done automatically with a cryptocurrency, and the transaction is also encoded in the blockchain.
In this scenario, ROs have an incentive to store and make available RI, and patients have their data widely available and secure. The system also saves money as the transaction costs would be more economical than overnight mail or repeating studies unnecessarily.
The system is not entirely “trustless” as the patient must still trust the RO, but it is an improvement on the system we have today. Only a compromise of the RO’s private key would result in a third party being able to unlawfully access the information. Unfortunately, a leasing RO cannot be prevented from decrypting RI using an expired symmetric key; though, this would be documented in the public ledger, the blockchain, and repeat offenders could be caught easily.
Additionally, demographics can be removed from studies themselves, and patient information could be encoded in the blockchain instead. Thus, a compromise of RI would not necessarily be a breach of protected health information.
Patients may also elect to enter into smart contracts to have their anonymized RI entered into pools made available for research studies or teaching files, receiving compensation without risk of a confidentiality breach.
Ultimately, the radiology blockchain could be expanded to create a decentralized autonomous organization (DAO) which generates smart contracts for rapid, secure results, reporting, and billing. I envision a smart radiology infrastructure to facilitate the interpretation of imaging studies by verified radiologists. Unread studies are fed into a common marketplace where radiologists select examinations to interpret based on an evolving incentive structure defined by timeliness, difficulty, or other factors. Results are securely sent to originating ROs, and payments are made automatically to the interpreting physicians.
The future of radiology looks promising and a DAO based utilizing the blockchain would be the first step to creating an efficient, trustless, and autonomous mechanism for the practice of diagnostic radiology.