Advances in artificial intelligence in health are today hampered by the difficulty of accessing data, which is necessary to drive deep learning algorithms. A transparent and secure database, the Blockchain could offer a way to centralize health information while respecting the anonymity of patients.
Since IBM decided to use its Watson supercomputer in cancer detection, a lot of ink has leaked around the potential of artificial intelligence in health. On the model currently being tested by IBM Watson in several hospitals, image recognition can be used to help radiologists detect abnormalities and potential diseases on an MRI. In a research article published by the American College of Radiology , researchers at the University of Texas, Austin, predict that artificial intelligence will soon become an integral part of the daily lives of radiologists, and will make their work more efficient, accurate and satisfactory.
Artificial intelligence can also scan patients’ medical records and genomes for information that can help make a diagnosis or treatment easier. Medical research could also greatly benefit from the contribution of artificial intelligence: by browsing the latest scientific publications, it would help build bridges between different works and identify the most promising discoveries, see to automate the process peer review.Artificial intelligence can also facilitate the discovery of new drugs and their compliance with health standards in force, to accelerate their entry into the market. Add the virtual assistants used to monitor patients remotely or the use of robots in rehabilitation, and we understand that the technology is generating some enthusiasm.
Most of the recent advances in artificial intelligence are based on deep learning, a branch of the discipline that allows machines to improve with experience, and is based on deep neural networks. , inspired by the functioning of the human brain. The deep learning lies behind AlphaGo , the all-round go game champion developed by Google Deepmind, but also behind the software that drives autonomous cars , or the image recognition techniques that allow Watson to analyze an MRI. If its performance is staggering, deep learning needs an immense amount of data to express its potential. In the age of the Internet and connected objects, health data are not lacking. However, these are often fragmented, difficult to access and exploitable.
Data fragmentation, a brake on innovation
The problem is particularly acute in the United States, where the health system is complex and involves different stakeholders. Thus, while the 1996 Health Insurance Portability and Accountability Act (HIPAA) gives patients the right to access and control their health data, in practice this access is often limited, as shown by the Reuters in an article published in 2016. Patients remain unaware of the content and value of their medical data. In addition, patient data is stored by their health provider on a private server, which prevents their circulation and exploitation by other actors (for example, to train an artificial intelligence program to recognize the symptoms of a sickness).
Note also that health data is not limited to information collected by physicians about their patients. In the era of new technologies, data from medical imaging, laboratory tests (such as blood tests), genetic tests, and those generated by wearables also need to be taken into account . The data is thus stored in different silos (doctors, hospitals, insurers, laboratories, start-ups …) with very little exchange between them. On the other hand, given the sensitive and private nature of this information, practicing a policy of full openness and free flow of data is, of course, not an option.
As a result, young health-intensive AIs are spending a large part of their budget buying data sets to train their algorithms. Confronted with a limited budget, they can often afford only a limited amount of data, which limits the possibilities of their solutions. In other words: the data exist, just as there is the technology to exploit them, but a common database is missing, capable of circulating them among the various actors, while respecting their confidential nature. Until recently, there was no technology that could disseminate information in an open, collaborative and secure manner. But with the advent of the Blockchain, things are changing.
The blockchain to secure health data
A giant medical database enabled by the Blockchain
A data center
The Blockchain is a new way of storing information. This is a database that is not centralized, but distributed. Each participant has a copy on his computer, which he can freely consult and modify. The Blockchain is therefore also transparent. If everyone has access to the database, how do you prevent a little clever person from changing the information to serve his interests? In the case of Bitcoin, for example, how do you prevent an ill-intentioned person from artificially increasing the amount of Bitcoin he has?
To counter this pitfall, each transaction must be verified and approved by some network participants, called “miners”, using complex cryptographic techniques. Security is thus provided by the individuals themselves rather than by a central entity. Once back, the information is visible to all participants. Simple and devilishly effective. Marc Andreessen , Silicon Valley’s historic investor, sees Blockchain as the most important invention from the Internet.
Healthcare providers, insurers and pharmacies store all their information in the same register for a given patient, and everyone can access the data entered by others.
This double hat, both open and secure, makes the Blockchain the ideal technology to solve the dilemma that is slowing the progress of artificial intelligence in health today. “The Blockchain is a digital registry that records and then shares transactions and interactions in chronological order, bringing security and interoperability to healthcare providers and their patients. In this area, each visit of a patient, each diagnosis, prescribed treatment and other key data included in the electronic health record constitute the transactions. Healthcare providers, insurers and pharmacies store all their information in the same register for a given patient, and everyone can access the data entered by others. “ Describes Sandy Hathaway , entrepreneur and venture capitalist, in an article on Medium. “Blockchains hosting health data can and must be created as ecosystems based on a principle of permission, with trusted collaborators. Additional controls must also protect the information that makes their owner identifiable while allowing the sharing of anonymous health data. She adds.