Open Navigation

IEEE.org IEEE Xplore Digital Library IEEE Standards IEEE Spectrum More Sites

IEEE Transmitter Facebook Twitter Instagram LinkedIn

Open Search
Open Navigation
Open Search

July 30, 2018

Free of supervision, deep learning algorithms can be trained to work faster and smarter than humans. Perhaps more importantly, they can also be used to diagnose diseases, including cancer. Just how effective is deep learning in a medical setting? Last year, researchers at the Houston Methodist Research Institute developed AI software that can accurately diagnose breast cancer 30 times faster than doctors can. In other words: very effective.

Before we dive deeper, here’s a quick refresher on deep learning from Karen Panetta, IEEE Fellow and Dean of Graduate Engineering at Tufts University: “Deep Learning is a field of AI that looks at massive amounts of real data to extract behaviors and trends. Then, when the AI is presented with something it has not seen before, it can tell you what it is with excellent accuracy.”

By checking its responses against an answer key to see how well it did, AI can then rearrange the patterns and run the data set again to see if it achieves a better result. You can read more about the learning process on IEEE Spectrum.

So, back to health. This iterative learning process means AI can be trained to look for and diagnose cancer based on sample medical images prepared by researchers.

To understand why that’s such an exciting proposition, consider the current setup. IEEE member Sukanya Mandal: “Currently, cancer is being diagnosed by visual examination of tissue samples under a microscope by considering factors such as cell shape, mass, number and appearance.

As the diversity among cancer cells is high, it becomes very difficult to assess every variation with the naked eye, especially things like the morphology of cells, the arrangement of cells within a tumor and the genetic diversity of cancer,” says Mandal. Therefore, AI could well be the difference maker in judging whether tumor cells are malignant or benign. That accuracy could then help prevent unnecessary surgeries.

But the potential power of AI doesn’t stop there. We’ve written before about a machine learning system that studies lesions, demographics, family histories, biopsies and pathology reports, and uses all of that data to better predict if a certain lesion will develop into breast cancer. Software also already exists that diagnoses breast cancer with 99 percent accuracy when looking at mammogram results and medical histories.

The excitement regarding this technology is profound. In our global survey of 2,000 parents aged 20–36 years-old, a majority in China (85 percent), India (83 percent) and Brazil (70 percent) all think AI medical breakthroughs will eradicate cancer in their children’s lifetime. Read our 2018 study on Generation AI for more findings.

Karen Panetta headshot

Karen Panetta

IEEE Fellow, IEEE Robotics and Automation Society, IEEE Education Society, IEEE Computer Society, IEEE Oceanic Engineering Society, IEEE Product Safety Engineering Society, IEEE Signal Processing Society, IEEE Systems, Man and Cybernetics Society

SEE MORE FROM THIS IMPACT CREATOR

Related Stories

Engineer-examining-robotic-arm-in-office

Artificial Intelligence (AI) | 5 days ago

Hands-on Collaboration: Building Robot Muscle Memory

Blue Colored Computer Chip on Tip of Index Finger Against Dark Black Background Extreme Close-up View, Macro Photo.

Artificial Intelligence (AI) | 3 weeks ago

AI is Big. Some Want to Make It Small

touching optical fiber

Artificial Intelligence (AI) | 3 weeks ago

What’s Next for Generative AI?

Flowers petals and leaves forming brain shape inside male's head silhouette made out of grass on yellow background.

Artificial Intelligence (AI) | 3 months ago

AI, Energy and Sustainability – Your Questions Answered

INTERACTIVE EXPERIENCES

Close Navigation