New Results From a Big 2018 Skin Photo Challenge — What It Means for Melanoma
Doctors and tech teams around the world worked together on a huge project to teach computers to spot melanoma in everyday skin pictures.
The group built a public library of more than skin photos and invited many teams to test their tools, it was held at a conference in Spain and drew lots of attention.
Some programs did very well, others still missed tricky cases — in fact top tools still miss more than ten percent of images, so problems remain.
The big surprise was that two tools that score the same on a test can behave very different when shown new photos, it means a model might look great in one setting but stumble in real life.
That matters for patients, and for regulators who decide what tools are safe to use.
The challenge set a new bar for how to test these systems.
People will keep improving them, but we need better checks first, so these helpers are reliable when lives depend on them.
Read article comprehensive review in Paperium.net:
Skin Lesion Analysis Toward Melanoma Detection 2018: A Challenge Hosted by theInternational Skin Imaging Collaboration (ISIC)
🤖 This analysis and review was primarily generated and structured by an AI . The content is provided for informational and quick-review purposes.
Top comments (0)