Sam Brusco, Associate Editor08.08.16
IBM’s Watson boasts an impressive list of achievements, in which the supercomputer routinely dwarfs human intelligence. So far, the computer has defeated several geniuses on Jeopardy and cooked a slew of exotic dishes.
Realizing that Watson’s computing power could work wonders for healthcare, it has been enlisted for a number of life-saving ventures. In February 2013, IBM and American healthcare provider WellPoint launched Watson’s first commercial application, providing utilization management decisions for lung cancer treatment at Memorial Sloan-Kettering Cancer Center.
IBM Watson Health launched in April 2015, in order to leverage its formidable cognitive computing power as an open platform for physicians, researchers, insurers, and medical companies. This endeavor has thus far drawn in a number of organizations partnering with IBM to optimize personal health, through data collection from consumer and medical devices. The list is pretty star-studded, including Apple, Johnson & Johnson, and Medtronic.
Now, Watson has taunted human intelligence once again—this time by diagnosing a rare disease that had long stumped human doctors.
According to Japanese NHK News, Watson’s artificial intelligence sifted through 20 million oncological documents, resulting in a diagnosis of 66-year-old Ayako Yamashita’s extremely rare form of leukemia. In January 2015, she was diagnosed with acute myeloid leukemia, but wasn’t responding to treatment. So, the doctors decided to enter Yamashita’s genetic profile into Watson’s program, and it generated her correct diagnosis in about 10 minutes. According to Silicon Angle, Watson also suggested a different treatment, which was reported to be more effective.
This achievement marks a significant advance in the diagnostic capabilities of data analysis and artificial intelligence. No longer may doctors have to waste time combing through hundred of pages of research to diagnose a cryptic disease. But that’s not all: with adequate genetic data and a handy set of algorithms, Watson could greatly advance the personalized medicine initiative by prescribing customized dosages of medicine tailored to each patient’s genetic makeup.
However, collecting that much genetic information from patients comes with a set of problems of its own. Watson’s database of intimate health information is going to need extremely high-level security protocols, especially if prominent figures submit their information to be archived. Some may be uneasy about their personal health data out there in the cloud—Watson would likely also contain information on patients’ physical traits and ethnic background. A “hack” of Watson’s genetic database could release a staggering amount of information (provided Watson isn’t clever enough to divert hackers by then).
Another roadblock to Watson’s efficacy may be that it can only search through existing literature on a given disease. Especially rare diseases are not likely to have many clinical studies available. These would be more difficult to diagnose because there’s just not enough data for Watson to work its magic—unfortunately, (or perhaps, fortunately?) its power is limited by humanity’s research capabilities.
Further, who exactly would be considered liable should Watson be wrong? Say it generates an incorrect diagnosis or selects a treatment for a correct diagnosis, which worsens the disease or ultimately results in the patient’s death. Most would determine it’s ultimately the physician “operating” Watson that’s responsible—in the same manner a robotic surgery gone awry would place blame on the surgeon operating the device.
But would it be fair to penalize the physician? Watson’s analytics purportedly churn out the most likely diagnosis and optimal treatment, so why should a clinician have any reason to doubt the results? Presumably, clinicians would do their due diligence and verify Watson’s decision, but even the world’s smartest computer might make an error. And you can’t exactly sue a computer, so would IBM shoulder the blame for Watson’s screw-ups?
That’s not likely, but before Dr. Watson enters the clinic on a regular basis, regulations like these need to be set in place—because with great (computing) power comes great responsibility.
Realizing that Watson’s computing power could work wonders for healthcare, it has been enlisted for a number of life-saving ventures. In February 2013, IBM and American healthcare provider WellPoint launched Watson’s first commercial application, providing utilization management decisions for lung cancer treatment at Memorial Sloan-Kettering Cancer Center.
IBM Watson Health launched in April 2015, in order to leverage its formidable cognitive computing power as an open platform for physicians, researchers, insurers, and medical companies. This endeavor has thus far drawn in a number of organizations partnering with IBM to optimize personal health, through data collection from consumer and medical devices. The list is pretty star-studded, including Apple, Johnson & Johnson, and Medtronic.
Now, Watson has taunted human intelligence once again—this time by diagnosing a rare disease that had long stumped human doctors.
According to Japanese NHK News, Watson’s artificial intelligence sifted through 20 million oncological documents, resulting in a diagnosis of 66-year-old Ayako Yamashita’s extremely rare form of leukemia. In January 2015, she was diagnosed with acute myeloid leukemia, but wasn’t responding to treatment. So, the doctors decided to enter Yamashita’s genetic profile into Watson’s program, and it generated her correct diagnosis in about 10 minutes. According to Silicon Angle, Watson also suggested a different treatment, which was reported to be more effective.
This achievement marks a significant advance in the diagnostic capabilities of data analysis and artificial intelligence. No longer may doctors have to waste time combing through hundred of pages of research to diagnose a cryptic disease. But that’s not all: with adequate genetic data and a handy set of algorithms, Watson could greatly advance the personalized medicine initiative by prescribing customized dosages of medicine tailored to each patient’s genetic makeup.
However, collecting that much genetic information from patients comes with a set of problems of its own. Watson’s database of intimate health information is going to need extremely high-level security protocols, especially if prominent figures submit their information to be archived. Some may be uneasy about their personal health data out there in the cloud—Watson would likely also contain information on patients’ physical traits and ethnic background. A “hack” of Watson’s genetic database could release a staggering amount of information (provided Watson isn’t clever enough to divert hackers by then).
Another roadblock to Watson’s efficacy may be that it can only search through existing literature on a given disease. Especially rare diseases are not likely to have many clinical studies available. These would be more difficult to diagnose because there’s just not enough data for Watson to work its magic—unfortunately, (or perhaps, fortunately?) its power is limited by humanity’s research capabilities.
Further, who exactly would be considered liable should Watson be wrong? Say it generates an incorrect diagnosis or selects a treatment for a correct diagnosis, which worsens the disease or ultimately results in the patient’s death. Most would determine it’s ultimately the physician “operating” Watson that’s responsible—in the same manner a robotic surgery gone awry would place blame on the surgeon operating the device.
But would it be fair to penalize the physician? Watson’s analytics purportedly churn out the most likely diagnosis and optimal treatment, so why should a clinician have any reason to doubt the results? Presumably, clinicians would do their due diligence and verify Watson’s decision, but even the world’s smartest computer might make an error. And you can’t exactly sue a computer, so would IBM shoulder the blame for Watson’s screw-ups?
That’s not likely, but before Dr. Watson enters the clinic on a regular basis, regulations like these need to be set in place—because with great (computing) power comes great responsibility.