How Data-Driven Technology Improves Patient Diagnostics
Credit: Maggie Bartlett, NHGRI
How Data-Driven Technology Improves Patient Diagnostics
One of the most critical advancements in medical technology is the rise of data-driven technology to help with patient diagnosis. Recent technology has enabled new types of data collection with important implications for both patients and medical professionals. Here is a round-up of how this type of technology is helping doctors formulate more effective treatment plans today and where this tech is poised to go in the future.
When a Japanese woman’s illness stumped her doctors, they turned to IBM’s artificial intelligence machine, Watson. The woman had been diagnosed with acute myeloid leukemia, but was failing to respond to treatments. Watson analyzed the patient’s genetic information and compared it to 20 million clinical oncology studies, ultimately determining she had a different and rare form of leukemia. The AI system was able to diagnose the patient in just ten minutes. This allowed the doctors to develop a new treatment for the woman in question — saving her life.
The advantage of an AI like Watson is its ability to analyze massive quantities of data rapidly. This ability is already being used to improve the care of other cancer patients in the US as well, with MIT researchers using machine learning techniques to modify treatment plans for patients with glioblastoma. In a paper presented in 2018, the researchers presented a model that might reduce the aggressive amounts of drugs and chemotherapy that patients with glioblastoma currently must undergo. The model, driven by machine-learning, examines current treatment regimens to find an optimal treatment plan that administers the lowest frequency of doses that would still effectively shrink a tumor. Researchers have high hopes that the U.S. Food and Drug Administration will soon create guidelines to vet these data-based treatment technologies and allow them to be more widely used.
Wearables and sensors
Most people are now familiar with wearables, whether as a fitness tracker or as an extension of a user’s smartphone. But wearables have a huge medical diagnostic potential as well. A device that constantly monitors and collects health data on a patient would, of course, have immense possibility to revolutionize diagnostics and treatment of patients. For example, Princeton researchers reported in a recent paper that they used biomedical data to detect five diseases in simulations created from various patient data. They fed publicly available biomedical data into machine learning algorithms that were trained to identify these diseases.
The new system compares data points to publicly available data about disease symptoms, allowing the algorithm to detect symptoms that even patients didn’t know about. The system diagnosed type 2 diabetes with 78 percent accuracy, hypothyroid with 95 percent accuracy, and urinary bladder disorder with 99 percent accuracy. While the system has a ways to go before it’s accurate enough to hit the market, it shows promise. Users could someday use wearables to monitor whether they’ve developed or might be at risk for developing a disease like diabetes. The earlier the patient detects symptoms and alerts their medical provider, the earlier they are able to get treatment — and the better off they are.
The Human Genome Project is at the root of much data-driven medicine today, and it continues to be a relevant and important source of medical insights. Combined with the latest and most powerful computing abilities, the Human Genome Project and other DNA sequencing projects that stem from it are now able to generate more data than ever before. The quantity and quality of this information might someday allow patients to quickly and cheaply sequence their entire genome. This would provide incredible information about drug sensitivities, family history, and even a person’s risk of certain diseases. Even though this possibility is still in the future for now, there is plenty of research and funding devoted to this field today.
In 2012, the Mayo Clinic, recognizing the importance of data-driven diagnoses, opened the Center for Individualized Medicine. The Center focuses on using genomics to help identify and treat rare diseases. One of its programs, called the Disease Odyssey, sequences a subset of the human genome that includes instructions for building proteins (called an exome). It took several years and a team of 369 data scientists to build the technology. However, it has succeeded in its mission. According to the Center, about 37 percent of patients get a diagnosis within three months. That includes a six year old boy from New Mexico who had suffered from seizures since he was two. With the help of DNA sequencing and big data, he was eventually diagnosed with a mutation so rare that it had only been documented in ten other children.
With the technology available today, data-driven diagnoses are more accessible than ever. Artificial intelligence can be trained to recognize symptoms and is able to analyze a staggering amount of data to match symptoms to diseases. Wearables open a new frontier of data collection, and as the technology becomes more widespread, the quality and quantity of data will become more robust. And as attention and funding goes to genome testing, the amount of data will grow and improve the chance of diagnosing rare diseases. Still, a good diagnosis is still just one step toward improved health care — technology that will help provide treatments is a must too.