Friday, September 27, 2019

Blockchain: What is it, how will it be used in healthcare and medical imaging?

Woojin Kim, MD, is a diagnostic radiologist with fellowship training in musculoskeletal radiology and imaging informatics. As an entrepreneur, he co-founded Montage Healthcare Solutions, a healthcare search and analytics company that provided enterprise search and data mining within healthcare systems for clinical decision support, research, business intelligence & business analytics, and clinical quality analytics. When the company was acquired by Nuance Communications in 2016, Kim became Chief Medical Information Officer for Nuance Communications. A recognized expert in imaging informatics with a focus on healthcare search, data mining, business intelligence and analytics, clinical quality analytics, blockchain, and machine learning/deep learning. He serves on the Board of Directors for the Society of Imaging Informatics in Medicine (SIIM) and is a much sought-after speaker for imaging informatics and blockchain technology with more than 170 abstracts, presentations, and talks. He spoke with PARCA eNews by phone about blockchain in healthcare.

Q. What is blockchain? Everybody's heard about it, but I'm not sure everyone really understands it.

A. Yes, I think so. I believe it is important to describe what it is and what it isn't, so we're all on the same page from the beginning. So first let me tell you what it is not. People sometimes equate blockchain with Bitcoin because cryptocurrency uses blockchain technology, but they are not the same. Cryptocurrency is one of the use cases of blockchain. Bitcoin did help to bring the concept of blockchain to popular awareness with the press and media and the masses.

So the next question is, what is blockchain? Well, blockchain is a decentralized, distributed, and immutable public ledger.

I'll talk about these terms individually. Blockchain has a decentralized network, which means it does not rely on a central point of control. It does not have a single point of failure because there are multiple redundancies.

It's also distributed meaning that the information is stored not by a single entity, but by many different individuals or "nodes" as they are often referred to in blockchains.

Finally, there is the immutable aspect, where once a record is written in a blockchain, it cannot be changed.

Blockchain Resources
There are pretty good online resources people can go to get a visual sense of what blockchain is and how they work (See Blockchain Resources). But that is a 30,000-foot view of what blockchain is. 

Q. To bring it down a little bit to just health care, how would it be used in healthcare and specifically in imaging? 

A. Yes, that's what I wanted to do today, to really focus on the use cases of blockchain within medical imaging and, specifically for your audience, imaging archives. 

Before implementing blockchain, it is essential to know what it can and cannot do. You also want to ask: what problems do I want to solve? Can I solve this without using blockchain? Because you may not need to use blockchain to solve your problem.  

Now that may sound obvious, but I often hear about people who learn about the latest technology whether it is AI or something else like blockchain, and it's like they have found a brand-new hammer and are out there looking for nails.

While I see a lot of potential in blockchain technology, I think we have to go through this mental exercise carefully to make sure there is an actual use case.

So let me go into blockchain for medical imaging archive. As far as I am aware, various proposed blockchain ideas for medical images do not have the patient images transferred or stored in a blockchain - the data is just too big. In fact, with blockchain, the general guidance is to store only minimal data while keeping sensitive data off-chain for better speed and scalability.

Thinking that the images are stored in blockchains is a common misconception people have when they first hear blockchain and medical imaging. Only the hash values are stored in the blockchain to ensure there hasn't been tampering of the images. A hash value is a long alphanumeric string created based on the content of the data. Because hash values are like digital fingerprints of these images, even a tiny change - either a pixel or metadata - can result in a different hash.

As a result, blockchain would simply add a new workflow that records the data source in a digital ledger with hashes verifying that the data source is unchanged. One thing I wanted to emphasize to your readers about hash values is that you cannot recreate the original data content just by looking at the hash value.

And so, the question is, why would you use blockchain in conjunction with imaging archives? For that, I think we need first to highlight some of the problems that exist with current imaging archives. Let’s start by looking at the issues that we have today and go through some of them.

One of the challenges that we're facing today is that access to medical images is being demanded more and more by both patients and providers. Instead of having multiple sets of imaging records for a given patient, imagine – using an old term - having one patient jacket. So instead of having a record of 30 x-rays in hospital A, two ultrasounds in Hospital B, and three MRIs in imaging center C, all of the patient’s imaging exams would become part of a single blockchain ledger for that patient. Imagine the efficiency and cost-savings benefits, such as those that come from the reduction of unnecessary repeat imaging.

Now let’s take a patient who wants to share her imaging data with an ED physician in an area the patient was vacationing or with a new oncologist. Or imagine, after removing identifiers, she wants to share a subset of images with a group of researchers. While image sharing platforms of today can address some of these needs, blockchain as a technology has the potential to provide greater security, access control, and audit trails while giving greater granular control to the patients over their images. It allows for a patient-centric sharing, and some also have proposed this type of use case to protect imaging data from Protected Health Information (PHI) hacks.

Speaking of cryptocurrencies, there are proposals to use token economics to incentivize patients to participate where you would get some kind of a token as a reward for sharing your information with researchers or AI startups. Admittedly, this is quite controversial, but some startups have even argued for the use of blockchain and its token economics to allow hospitals to sell anonymized and annotated patient data, including imaging data for profit. What's more, all of this can be done through smart contracts without any middlemen.

So what is a smart contract? You can think of it as a computer program on the blockchain network that automatically executes specific terms and conditions as a part of a contract between both parties. Think of it as an automatic digital agreement that can be performed without involving any middlemen.

Q. Backing up a bit, if a patient wants to give her oncologist access to all of her imaging, and it's stored on, Hospital A and her Clinic C or you know some other servers; would she then just give him the key to the blockchain and then the oncologist would sign something electronically and then have access to all of those? Is that what we're talking about?

A. For example, the blockchain would store the entire historical record of a patient chronologically; it would be like having a single patient jacket. The images themselves wouldn't be on the blockchain, but the information stored on the blockchain would be like pointers – pointing to where the images are stored. Now, let's say Jane Smith has just seen a new oncologist, who needs to see all her mammograms and breast MRIs. Jane can give the oncologist permission just to look at those images or, for greater context, her entire medical record. This permission is also time-limited, meaning the access may be limited to two years, after which, either the access is terminated or is renewed by Jane or automatically based on her preference.

That is the kind of granular-level control that is possible with blockchain that a patient would have in terms of who gets access to what and for how long. Now you can potentially do that with modern-day image sharing platforms, but it is more limited. For the proponents of blockchain in healthcare, they believe blockchain can give the control back to the patients where it is the patients who own the health records, including medical imaging exams.

Let's look at this from another angle. Today, most likely, your medical record is spread across many different institutions. Unless you have been getting all your medical care from only one healthcare system since the day you were born, there is no single repository for all your health records. Because your healthcare data is scattered all over the place, this means that your healthcare data is not portable.

With blockchain, because the whole thing is on the digital ledger and you own it, you can move from New York City to Los Angeles, and you'd still have access to the same set of data. The data goes where you go, and this is why people are advocating for this technology.

Q. In other words blockchain is a way to get to where I think we're trying to get to in healthcare in this country, to a place that wherever you want to see a doctor, they can have access to all of your health records and imaging.
A. Yes, that's right. One of the most significant potential benefits of using blockchain is for that exact purpose. Let’s take a radiology example, which I give at some of my talks. How many of our patients go to a particular imaging center and have to tell the staff what their contrast allergy is or whether they have metals in their bodies? That patient might have to repeat that information three or four times in a single visit. You might be telling it to the front desk person, then when you go to the preparation room, then again to the nurse, and then to the technologist just before you walk into the scanner, right?

Yes, we do this for patient safety, but isn't there a better way? Not just this example, but generally, why do you have to fill out new forms every time you see a doctor? It's because all of this data is in disparate systems, and they don't talk to each other. However, if such information were stored on a blockchain, I would just give permission, and anybody that I'm about to encounter in that imaging center will have access to my contrast allergy history or metal-in-the-body history where no one has to ask those questions to me repeatedly. In turn, this would save time and costs and would increase patient satisfaction. 

A part of the reason why you can trust the information contained with the blockchain comes from its immutable property. You can add data to a blockchain, but you cannot modify or delete what's already on the blockchain. The reason why it is called a "blockchain" is that each block is connected to the next block by the previous hash value – hence, each block forming a "chain." Each block will contain the hash value that represents whatever data it contains, but it also includes the hash value of the previous block. This means that if someone were to tamper with the data and change something from one of the blocks, it would change the hash value of that block which no longer matches the previous hash value stored in the subsequent block, and this invalidates the rest of the blockchain.

You can potentially hack it, but remember the beauty of blockchain is that it is also distributed, which means an exact copy of that entire blockchain is stored by multiple nodes, potentially thousands of computers. This means, if you wanted to hack a particular block in a blockchain, not only do you have to somehow hack that block and the rest of that entire blockchain before a new block is added on, but you would have to do this to at least 51 percent of all the nodes out there at the same time. And, that's practically impossible with today's computing power.

Because of this immutability, the proponents of blockchain believe it is secure. This allows the users to trust its content, which means the person who's looking at the contrast allergy information could trust that information without having to ask the patient each time. With blockchain, you have a trusted system without any middleman. That allows for a lot of things to happen automatically without going through multiple different intermediary steps, which can lead to cost savings and increased efficiency.

Q. Okay, so just for my understanding when you go to a new doctor, and she wants to add another image, is that adding another block to your blockchain, or is it adding another blockchain?

A. You would be adding another block to an existing blockchain that contains your medical record. If you get a new MRI today, that gets added to the blockchain.

Q. Do you have examples of hospitals or organizations that are implementing blockchain for imaging?

A. Let me first share a few examples outside medical imaging. Change Healthcare, a company I think your readers are familiar with, has a blockchain technology for claims management. A company called Hashed Health uses blockchain for physician credentialing. As a radiologist, I can tell you I have to fill out a very thick stack of papers each time I get credentialed at a hospital. I have to write down precisely what licenses I've had in the past, when I got them, when they expired, what medical school I went to, what residency and fellowship programs I did and when, etc. The thing is I have to do this for every hospital I want to get credentialed at, and all that paperwork is cumbersome. After I send them in, it would take months for the other side to review and verify all the information before allowing me to work at that hospital.

The thing is all these hospitals, granted they all use different forms, are asking for the same information. Imagine now, putting all of that information on a blockchain. Every time I got a new medical license, that license would get added to the blockchain, which has been verified. Then, all I have to do to get credentialed at a new hospital is say, "Hey, I’ll give you access to the blockchain that contains that information." The credentialing board would have instantaneous access to all my records. Since the information stored on the blockchain can be trusted, the verification process would also be immediate. In fact, the entire process can be completed in seconds, if not minutes.

According to the Hashed Health's website, it is "estimated that for every day a physician's employment or contracting is delayed, the organization forfeits $7,500 in net revenues." Therefore, this is the kind of cost savings blockchain could potentially mean to hospitals.

So then, back to your original question, are people using this for medical imaging? I can tell you that there have been a number of startups trying to use blockchain with medical images, whether for creating a marketplace for teaching materials, datasets for AI training, and datasets for research and clinical trials, adding tokens to incentivize, performing teleradiology, and so on. However, many of them have already disappeared or changed business models. So I can tell you that this isn't easy. Blockchain is still nascent in medical imaging.

Having said that, one possible area where blockchain may be useful is related to AI. As suggested before, blockchain has the potential to provide an additional layer of trust and security on top of an existing image archive.

Imagine you're live with multiple AI models running on your teleradiology platform. Now, a malicious hacker can use something called an adversarial attack to alter the transmitted images to the AI models for analysis. With the adversarial attack, you can change the original image just slightly, such as by adding a little bit of noise, which would be invisible to the naked eye. However, the AI models would make wrong decisions due to this subtle manipulation of the images.

For example, let’s say you have a positive pneumothorax on a chest x-ray, but after the adversarial attack, the AI would call it negative. A little bit of noise that's invisible to you and me would alter the decision of the AI model.

Using another example, a hacker can use something called a generative adversarial network to make a fake lung nodule on a chest CT or delete lung cancer from a chest CT, fooling the radiologists. Granted, these are theoretical at this point; the reason I am mentioning these examples is that researchers have shown such attacks are possible. What is scary is that as far as I know, I am not aware of any AI or PACS companies that can deal with such cyberattacks.

So how can blockchain help with all this? Well, blockchain with their hash values can potentially provide enhanced cybersecurity for image sharing. Take teleradiology, for example, blockchain can ensure that the images have not been altered after they were obtained, during transmission, and after arrival on the viewer. Why? Because as soon as the images get altered by these hackers, the hash values would change, and the systems would know about it immediately. In essence, by using blockchain, you are creating immutable images.

With that, let me discuss the challenges and limitations, and why we are not seeing widespread use of blockchain in medical imaging.

The first of many challenges and limitations is operational. Adoption of blockchain means changes in business processes. Both the employees and management need to adapt to the new workflow. However, at this point, a lot of people don't even know what blockchain is. There is a knowledge gap, and once you know what it is, you realize blockchain requires a different way of operating and a different way of thinking.

Second, skills and experience in blockchain technology are currently limited so even if you wanted to bring in a blockchain technology, who's going to staff it and manage it?

Third, there are technical limitations with blockchain as it is far from perfect. It has performance and scalability limitations. Blockchain is slower, and it's harder to scale than say traditional databases. Also, it's not easy to switch between different blockchain technologies.

So it's not unlike the early days of PACS where if you wanted to change from PACS company A to PACS Company B, the data migration was painful. With blockchain, it is even harder because blockchain technologies are evolving fast, and you can't simply switch from one blockchain to another Adding to the technical limitations, we also lack the technical infrastructure for data interoperability when it comes to healthcare.

Fourth, there are cultural limitations. I shared this at this year's SIIM Annual Meeting. Let's face it; there is a reluctance to share data in healthcare. While many people cite the reasons of privacy and security as the main reasons, and that is true to some extent, we do have a culture and tendency in healthcare, whether you're a hospital or a technology vendor, to not want to share data. Honestly, I see cultural challenges to be one of the biggest hurdles for the adoption of blockchain in healthcare.

Fifth, even though one of the potential strengths of blockchain is transparency because you can see every transaction history when you look at a public blockchain, using a public blockchain would raise patient privacy issues even if public/private keys are used. Because of this, many healthcare blockchain efforts use what's called a private or permissioned blockchain.

The sixth challenge relates to the token economics I mentioned earlier. On the surface, the token economics looks like a great idea to incentivize people, but I have yet to see a token that works in radiology.

Seventh is the use cases. Even that hypothetical AI use case I shared earlier, people may say, "Oh yeah, I can see how blockchain would be beneficial for that." However, you can counter-argue to use checksum instead of a complex blockchain backend to ensure the images have not been altered.

Finally, there are regulatory challenges. For example, I don't know whether using random public keys combined with hashed exam identifiers are good enough for de-identification, or patient's digital signatures on the blockchain are sufficient for PHI release under HIPAA.

In the EU, the challenges are even higher for blockchain in healthcare as personal data cannot leave the EU, according to the GDPR (General Data Protection Regulation). Also, there is a section called "right to be forgotten" with the right to erasure. This means you cannot really put personal data on blockchain because the blockchain is by nature immutable and with this right to erasure, well, that's sort of hard to accomplish because with blockchain, by design, we cannot delete things, right?

These are just some of the things that cause challenges to adoption of blockchain in healthcare. Yes, the technology is very interesting and has many potentials, but at the same time, you’ve got to take a step back and say, I have to understand the limitations and challenges. They need to be addressed before blockchain can be widely adopted.

All of this is why I think this conversation is important, and I appreciate you reaching out to me. I believe it is important for radiology professionals, such as PACS administrators, to become more familiar with this technology, and perhaps, more importantly, understand when to use this technology and when not to.

No comments:

Post a Comment