Injecting cutting-edge technology
To transform its vision into reality, Cloud Therapy is taking advantage of some of IBM’s most advanced technologies.
The company first feeds a “training set” of information from medical journals and other sources into IBM Watson™, and helps the solution learn to correctly identify relevant pieces of information within the corpus of text in response to a set of sample questions. The cognitive pathways established by this training phase are then tested by loading a second “testing set” of information from the same journals, and seeing whether Watson is able to make the same correct inferences about the test data.
By iterating this process of training and testing, it is possible for Watson to gain a very sophisticated understanding of the domain within just a few months – and provide insight into a much larger volume of medical research data than any individual doctor could possibly hope to read and absorb manually.
To make it easy for doctors to interrogate the solution, Cloud Therapy uses IBM® Watson Natural Language Classifier to parse users’ questions into topics that Watson understands. IBM Watson Retrieve and Rank then identifies the answers most likely to address the person’s query, and presents them in a list sorted by relevancy.
Using IBM Watson Speech to Text and IBM Watson Text to Speech, the solution can even understand and convert questions and answers between spoken words and text. As a result, doctors enjoy maximum flexibility in terms of how they use the solution.
“We experiment with a wide range of application programming interfaces [APIs] as soon as they are released, to see whether we can benefit from any new technology that comes onto the market,” Sandoval adds. “We consistently find that IBM is a leader in many different areas of cognitive computing and big data analytics, and we’re currently using around seven different cognitive APIs from the Watson portfolio.”
To address the wide range of rare conditions from which patients can suffer, Cloud Therapy’s solution also needs to store, search and process mountains of data. To do so, it relies on the IBM Cloudant® database.
“When doctors use our solutions, we want to give them the best possible experience,” Sandoval explains. “It is imperative that the data is always available and can be accessed and searched extremely quickly. Equally, because the data is very valuable and in some cases confidential, it needs to be stored securely. IBM Cloudant enables us to offer a robust, scalable, high-performance platform, so our users benefit from a superb in-app experience.
“We use Cloudant as a database-as-a-service solution, so we can delegate day-to-day administration to IBM, and instead focus on developing our cognitive solutions and providing a brilliant user experience. One of the best things about Cloudant and the managed service from IBM is that we could grow from 1,000 to 10,000 users overnight, and they won’t suffer from any decline in performance or reliability.
“The data we use is a mixed bag – some is structured, but a lot of it is unstructured – for example, medical studies in PDF format. In a traditional database, it is very difficult to deal with that complexity – but as a NoSQL database, Cloudant is flexible enough to handle data even when it doesn’t have a fixed schema. That means we don’t have to spend too much time worrying about normalizing the data before we ingest it – so we can focus on analytics, not data management.”
The solutions that Cloud Therapy develops all run on the IBM Bluemix® cloud application development platform, which makes it easy to coordinate a wide range of services and data sources into a coherent solution, without needing to spend time developing custom integration logic to help the data flow between them.
“Compared to the older model of software development, where you had to set up everything yourself from scratch, the agility that Bluemix gives us is absolutely amazing,” Sandoval says. “The ease with which you can swap components in and out of the solution means that you never feel it’s a waste of time to experiment. You can try out something new in a few hours – if it works, great! If it doesn’t, you can scrap it without any regrets, because you haven’t invested weeks of time and resources in it.”
He adds: “Bluemix is also a very open platform – it’s not limited to using IBM APIs, so you can plug in third-party services very easily. That’s really important when you’re working on leading-edge projects – if a new service comes onto the market and you think it might solve one of your problems, you need to be able to try it out, regardless of who the vendor is.”
Helping patients get back to glowing health
Targeting faster diagnosis of rare diseases, Cloud Therapy teamed up with Alexion Pharmaceuticals, a world leader in treating ultra-rare medical conditions, which occur in fewer than one in 20 million people.
Sandoval remarks: “Over the last ten years, Alexion has accumulated massive quantities of information concerning around 9,000 ultra-rare diseases. This information has been collected through research at its partner hospitals, and from patient records – for example, showing which genetic mutations correlate with certain illnesses.
“Up until now, this highly valuable information has been trapped in formats that are difficult to search and process, so doctors have not been able to access it quickly. Our solution will be able to read and understand all of this data automatically at great speed, and then present relevant findings to physicians at Alexion’s partner hospitals.”
He continues: “To take a personal example, I suffer from a symphalangism, a rare genetic condition of the hand that occurs when some of the finger joints are fused together. When I was originally diagnosed, it took doctors six months to identify my condition.
“With our solution, to diagnose my condition, a physician would first take a DNA sample and examine my genotype, phenotype and chromosomes. Then, the doctor would start using Cloud Therapy to interrogate the huge Alexion rare disease dataset to narrow down the list of likely conditions.
“From the 9,000 diseases on Alexion’s database, a search by ‘hands’ might filter the results down to 500. Then the doctor tells the solution that the patient shows signs of fusion in the joints and digits, and the results shrink to 200 conditions. From the DNA test, the doctor knows that I have a specific genetic mutation, so when they enter that information the solution condenses the results down to 10 likely illnesses – a manageable number for the doctor to study before making a final diagnosis.
“At the moment, most physicians use online medical databases to search for conditions and help them reach a diagnosis. However, traditional search engines don’t understand anything about the content that they are searching or the context of the question, so they often miss important answers or come back with irrelevant information.
“By contrast, our solution understands the domain and can see patterns in the information and the kinds of questions that users are asking. The IBM Watson machine-learning capabilities mean that as physicians use Cloud Therapy, it learns from its mistakes and delivers even more relevant results. You can even engage in a two-way dialog with the solution, rather than just running through an exhaustive list of search-terms until you find the right answer.
“We predict that Cloud Therapy will help doctors halve the time taken to diagnose rare diseases. As a result, patients will be able to receive the right treatment earlier, limiting the impact of the disease on their quality of life, and giving them a better chance of making a full recovery.”
While building the rare disease diagnosis solution, Cloud Therapy has also been able to use its new knowledge of cognitive computing to branch out into a range of other use cases. For example, it is also now working with a Danish pharmaceutical company called Leo Pharma to gain insights into how nutrition affects psoriasis.
“Leo Pharma gave us a whole body of data regarding nutrition and psoriasis,” Sandoval recalls. “We fed that information into Watson to help it learn about the relationship between patients’ diets and the severity of their condition. Finally, we combined all of this functionality in a mobile app, which helps people with psoriasis work out which foods are most likely trigger flare-ups. That insight will help them to manage their condition.”
Eventually, Cloud Therapy also plans to launch a B2C application to bring Watson to the masses. Sandoval comments: “The main diseases threatening people across South America are Chikungunya, Dengue Fever, the Zika virus and Malaria. We are working on an app that will allow people to input their symptoms, work out whether they are likely to have contracted one of those illnesses, then find out where to seek treatment.
“Many of these diseases are common in rural areas, where internet connectivity can be patchy at best. Because Cloudant offers offline synchronization capabilities, our app will be able to retain basic functionality even without an internet connection.”
Cloud Therapy is already looking to broader horizons, expanding beyond the healthcare sector into other industries.
“The technology we rely on isn’t just limited to healthcare use cases; we can teach it to understand just about any data,” Sandoval adds. “For instance, in conjunction with IBM Ecuador we are processing mainframe installation guides and manuals using Watson. Essentially, we are building a cognitive engine that technicians can use to describe the issue they are having with their mainframe, work out what is causing it, and see how to fix it. That will act as a hugely useful resource during mainframe implementations.”
Sandoval concludes: “Using cutting-edge IBM technology, we are building solutions that stand to spark real change in the healthcare industry and beyond. We’re excited to see what the future holds.”