Skychain is aiming to take the trust placed in human healthcare workers and put it into Artificial Intelligence. According to the company, their diagnostics system has recently been tested and seemed to prove more accurate than real life doctors at spotting conditions in patients (a video of the study is available here). By bringing together healthcare datatsets and programmers to create neural networks, the system was tested on diagnosing melanoma, heart disease and cancer in patients. They claim that the human doctors rate of error in diagnosis was between 18-32%, whereas the Skychain system was only incorrect four to 14 percent of the time.
In addition, the team claims these diagnosis were made in an average of 0.1 seconds by the AI system, and 20-30 seconds by the real doctors. If these results can be proven and verified, there is certainly scope for improving both accuracy and time efficiency in healthcare diagnosis, potentially saving millions of lives worldwide. Beyond diagnostic capabilities, Skychain hopes these supposedly positive results can be repeated with the system being given the ability to discover new diseases and develop new treatments.
How does the system actually work?
Skychain is a project which aims to use Blockchain to train and use AI systems in medical care. Their white paper states that they aim to control 70 percent of the projected $200 bln medical AI market (estimated by IBM), through their ‘distributed open network’ system of artificial neural networks (ANNs), which can diagnose patients and prescribe the relevant treatments. They aim to “provide an opportunity to engineer, teach and host neural networks and provide paid access for independent specialists and organizations.” By using smart contracts, they hope to unite many individual parties (healthcare big data providers, independent AI developers, crypto miners and the consumers – doctors and patients) to create one effective solution.
Developers can submit ready-made ANN templates for doctors to choose from when diagnosing a patient. Once the networks have analyzed the data and returned the diagnosis information to the doctor, the developers and miners who provided the computing power will receive financial remuneration.
Setting up these symbiotic relationships will involve medical institution laboratories with large datasets, looking to set up and train their own neural networks (which can then also be used by others). They can offer these datasets for ANN training by developers. These developers can use a ‘SkyConstructor’ interface, and using a ready-made ANN template can edit it to meet the requirements of the institution once uploaded into Skychain. Once the ANN has completed the learning process, it can be published. Skychain uses the analogy of the wildly popular Uber taxiservice: ANN developers are the drivers, doctors and patients the passengers, and the computer and servers of miners the cars.