Study: Artificial intelligence can diagnose skin cancer as well as doctors do

Computers could be a new partner for dermatologists when it comes to identifying skin cancers.

Through machine learning, computers now rival trained physicians when it comes to identifying cancerous skin lesions.

By feeding nearly 130,000 images of 2,000 different skin lesions into a computerized algorithm, researchers at Stanford University found that artificial intelligence was just as good as 21 board-certified dermatologists at identifying instances of skin cancer. Their findings, which were published in Nature, prompted researchers to imagine a scenario where patients could use their smartphone to get a diagnosis of a suspicious lesion.

Roberto Novoa, one of the authors of the study, told Wired that although the computer was slow to accurately identify cancerous lesions at first, the algorithm got stronger as it processed more images.


13th Partnering with ACOS & IDNS Summit

This two-day summit taking place on June 10–11, 2019, offers a unique opportunity to have invaluable face-to-face time with key executives from various ACOs and IDNs from the entire nation – totaling over 3.5 million patients served in 2018. Exclusively at this summit, attendees are provided with inside information and data from case studies on how to structure an ACO/IDN pitch, allowing them to gain the tools to position their organization as a “strategic partner” to ACOs and IDNs, rather than a merely a “vendor.”

“It was definitely an incremental process, but it was exciting to see it slowly be able to actually do better than us at classifying these lesions,” Novoa said.

RELATED: HIMSS 2017--Shahid Shah predicts Trump effect, machine learning, blockchain, genomics will be hot topics

In an accompanying commentary, cancer researchers at Oregon Health and Science University and the National Cancer Institute noted that machine learning may outperform dermatologists as more data is fed into the algorithm, but also raised concerns about the implications of a smartphone application that reduces physicians to the role of technicians and potentially removes the incentive for patients to undergo regular skin screenings.

Eric Topol, director of the Scripps Translational Science Institute told Wired that medical professions involving images—including radiology, dermatology, and pathology—are ripe for AI's influence, and that integrating machine learning would alter a clinician’s responsibilities but not overtake them completely.

Topol has previously said that radiologists and pathologists should consider merging into a single specialty to make room for AI advancements that are poised to infiltrate both professions. Mount Sinai’s Center for Computational Systems Pathology has launched new initiatives to use AI and algorithms to precisely treat and classify diseases.

Suggested Articles

The FTC is suing Surescripts, accusing the health IT company of employing illegal restraints to maintain its monopolies over the e-prescribing market.

Boston-based Athenahealth is laying off a portion of its workforce to “decrease bureaucracy and consolidate capabilities" as part of a reorganization.

The Trump administration wants to allow state Medicaid programs to test new models of integrated care to treat dual-eligible beneficiaries.