Helping the blind to see
Aug 25th 2017
Who would have thought ‘sight’ would be declared one of Scientific American’s Top 10 Emerging Technologies of 2017?
Of course, it’s not sight in the conventional sense, but rather the ability for computers to see or recognise things.
In Deep-Learning Networks Rival Human Vision, Scientific American technology expert Apurv Mishra explains computers’ increasingly powerful sight is due to a new deep-learning approach.
“Convolutional Neural Networks (CNN) do not need to be programmed to recognize specific features in images—for example, the shape and size of an animal’s ears,” writes Mishra.
“Instead they learn to spot features such as these on their own, through training. To train a CNN to separate an English springer spaniel from a Welsh one you start with thousands of images of animals, including examples of either breed…Once trained, a CNN can easily decide whether a new image of an animal shows a breed of interest.”
“It is projected that there will be 6.3 billion smartphone subscriptions by the year 2021, according to Ericsson Mobility Report (2016), which could potentially provide low-cost universal access to vital diagnostic care.”
A move into medicine
Being able to tell pedigree pooches apart doesn’t quite merit a place on the Top Emerging Technologies list. However, diagnosing disease is a different story. And CNN can do that too.
Mishra explains, “Deep learning for visual tasks is making some of its broadest inroads in medicine, where it can speed experts’ interpretation of scans and pathology slides and provide critical information in places that lack professionals trained to read the images—be it for screening, diagnosis, or monitoring of disease progression or response to therapy.”
The dermatologist smartphone?
As reported by Kurzweil AI, computer-vision has actually performed better at diagnosing skin cancer than 21 dermatologists.
In A deep learning algorithm outperforms some board-certified dermatologists in diagnosis of skin cancer the writer even proposes a new use for smartphones in the battle against cancer: “It is projected that there will be 6.3 billion smartphone subscriptions by the year 2021, according to Ericsson Mobility Report (2016), which could potentially provide low-cost universal access to vital diagnostic care.”
Ever more powerful
Naturally startups are closing in on the potential of CNN. San Francisco-based Arterys’ vision is, ‘A world where clinical care is data-driven, intelligent and patient focused.’ They are using CNN to visualise blood flow in the heart, which will help diagnose heart disease. In an article titled Artificial Intelligence takes on medical imaging, Modern Healthcare predicts the technology is set to get more powerful, opening up even more possibilities.
“To ‘train’ an algorithm to recognize, for instance, a stroke, developers feed the algorithm tons of imaging studies of a brain suffering from an attack,” writes journalist Rachel Z Arndt, “teaching the machine the nuances that make pattern recognition possible.
“Then, as the algorithm goes into action in the real world, acting on what it's already been trained to do, it can gain new information from new images, learning even more in a perpetual feedback loop.”
While the medical applications for CNN are most impressive, it has numerous other uses too. For example, the motor insurance industry is using the technology to assess car damage. Self-driving vehicles are being kitted out with powerful vision to help them recognise pedestrians. Even security and agriculture can benefit from CNN thanks to its ability to assess not only crowd behaviour in public places, but crop yields and diseases.
Humans still required
Unsurprisingly, of all CNN’s applications, its use in medicine has received most attention and arguably secured its place on the Emerging Technologies list. Does this mean doctors face mass job losses in future? In Artificial Intelligence is Transforming Radiography, Fierce Healthcare reports that AI expert Geoffrey Hinton suggested machines’ ability to read X-rays and scans means medical schools should stop training radiologists.
The situation isn’t quite so dire according to Medical Futurist, which sums it up beautifully in an article titled Can An Algorithm Diagnose Better Than A Doctor?:
“For the first time in history, making a decision about a patient’s case will not be based on the lucky uncovering of crucial information from the haystack of medical databases. With cognitive computing, physicians will be able to focus on helping their patients, instead of keyboards and monitors.
Clearly in technology, as in art, the truth is rarely black and white.