General practitioners (or GPs) have one of the hardest jobs in medicine. Based on a limited objective measurements, a set of subjective symptoms and past experience, they must decide whether a patient can be sent home, needs to see a specialist consultant or requires hospital attention. And all this in a very limited time frame (the average GP consultation time in the UK was 10.6 minutes, according to a 2015 survey).
Given these circumstances, any accuracy in diagnosis is amazing. Medical Doctors (MDs) are, perhaps unconsciously, incredibly skilled statisticians, with complex models of how symptoms and the physical (and mental) state of a patient map to possible ailments. These models are almost certainly influenced by experience. Every diagnosis adds to the data set, and their mental models modified to be consistent with the new information.
Artificial intelligence and medicine
The process described above is very similar to the general machine learning scheme. A learning algorithm is provided with training data to produce a model. This is then used to make decisions and predictions from new data. The algorithm can be tweaked on a feedback loop if improvements are needed. This process is iterated until the algorithm is trained to user satisfaction, or continuously to have an ever-adapting service.
It has long been speculated that a powerful enough computer equipped with the right machine learning facilities could help make medical diagnosis more efficient and accurate. In recent years this approach has been shown to be useful in several realms of medicine (haematological or skin cancer diagnosis, among others).
The scientists involved in these developments were focused on isolated or small sets of closely related ailments and diseases. Once a patient has reached the specialist, the list of specific health problems they may have has been significantly reduced from what the GP must initially consider. This makes the machine learning approach much easier. For example, in the case of skin cancer, identification is often achieved by visual inspection. While a difficult task in itself, image recognition is much easier than correlating different vital signs and sensation of a patient with disease.
It’s all in the data: the Internet of Things
The relevant questions are clear. Can machine learning be used to make a GPs job easier?, can artificial intelligence help reduce false-diagnosis rates? More importantly, can these tools help cure or prevent disease? While it is hard to quantify, it is safe to assume that prevention is more efficient than treatment. It is obviously better for quality of life purposes, and it is well known than curing disease is financially expensive.
In-depth analysis, such as detailed blood work, medical imaging or biopsies are expensive and cumbersome. As a patient these tests can also be extremely unpleasant. However, many vital signs can be easily and swiftly measured, and with little discomfort. Heart rate, body temperature, respiratory rate and blood pressure are indicators of general health and bodily function. Other measurements such as blood saturation or glucose levels can give further information. Prolonged monitoring of these vitals, in conjunction with an accurate personal and family medical history can be used to anticipate the need to visit a Doctor, or aid them in determining whether a patient requires more attention or not.
Measurement of many of these vitals can already automated, albeit with expensive equipment. The information is also rarely stored. With the Internet of Things this can change. Accurate sensors can be produced at a reduced cost and distributed in places with a network connection. Upon providing identification, the patient could be examined by these sensors, either as a check of general health or to determine the urgency of their complaint. In both cases the data can be stored in a personal file on the cloud, for immediate and future use. Additionally, the entire medical history of the patient can be accessed anywhere in the globe almost instantly.
Privacy is key
While the benefits of such an approach to medicine could be very beneficial, there are significant patient privacy issues which must be considered. Using IoT in medicine would imply having enormous amounts of personal data on the cloud and out of patient’s control. This information could be used for discrimination, blackmail or other negative actions
Databases are also mutable. Even if they are encrypted, if someone is skilled enough to penetrate the protection, that person will certainly be able to change stored information. Medical histories are considered for health insurance pricing, access to certain professions and, as discussed above, future medical treatment. A medical history could be considered in decisions as crucial as whether a patient is to receive a transplanted organ or not.
In the age of data, the safety of information is of vital importance. In the case of medical information, very literally so. As patients we deserve to receive the best possible treatment. Continuous health monitoring for preventive purposes and artificial intelligence-assisted doctors can provide this. But this cannot be at the price of having our personal information exposed and manipulated.
The role of blockchain
Before medical data can be stored on the cloud, we must ensure that it can only be accessed by the patient, as well whoever he or she authorises, and that it is immutable. The latter of these requirements is satisfied by a blockchain by construction. Blockchains are immutable registers: once data is committed to one, built-in countermeasures ensure that it cannot be modified.
Quantum1Net’s (Q1N) blockchain-in-blockchain (BinB) feature is especially well suited for this task. It allows for smaller blockchains to be embedded in the main one and used for specific purposes. Each patient could have their medical history stored in a personal embedded blockchain, at the ready for whenever it is needed. Q1N’s third-party integration also means that personalised encryption can be used to ensure the data is kept private. Therefore, the first of the requirements above is also satisfied.
An example of how this could work is as follows. Distributed IoT devices would identify a patient and access their personal embedded blockchain. Vital sign measurements would be stored on it whenever they were taken. That way, the data is available for instant use or for later use with other measurements to study how they evolve. If and when the patient required medical assistance, the Doctor in question would request authorisation to access the data stored in the blockchain, protected by the patient’s personal key. If the patient desired continuous monitoring and evaluation, the embedded blockchain could be synchronised with a specific app or cloud service.
You may have identified an overarching theme in the discussion: the power is in the hands of the patient. All operations involving sensitive information, including access, updating and even the measurement process itself, are subject to patient authorization. Yes, the data is stored on a distributed ledger, which means there are copies on storage devices all over the globe. Yes, the patient may need to allow seemingly anonymous machines to take check vital signs. But the decentralized nature of the blockchain implies that the only person with any real control over this information is the patient. No single third-party will control the data nor be able to access or manipulate it without prior consent.
Full Diagnosis: when will a computer be capable of this?
Given the possibility to collect the data and keep it safe, the question remaining is as follows: will there computers powerful enough to collect and process all the necessary data? And given enough power, can algorithms be created which will be able to aid in diagnosing or preventing disease? Luckily, over the past couple of decades the design and optimisation of supercomputers for scientific applications has skyrocketed. Every year more powerful supercomputing facilities are available, with the capacity to manage enormous data sets and complete difficult computations.
It remains to be seen when computers will be able to make general medical decisions. Though when this is achieved it cannot be at the price of patient privacy. On the contrary, emphasis must be placed on patient privacy and safety. Introducing artificial intelligence, IoT-type data collectors and distributed databases into the health sector should further empower patients. It should ensure that their medical data is readily available for use on their command and safe from unwelcome eyes and manipulation. Once the necessary infrastructures are available for this, data collection is a relatively simple task.
To own and control our own medical records is one of the most attractive applications of blockchain due to the many benefits, both to the individual and society as a whole, it presents. In conjunction with other technologies, it could also lead to much improved medical attention, greatly improving our quality of life.
Learn more about the Q1N blockchain, and our BinB concept here