Roupen Odabashian, Hematology/Oncology Fellow at Karmanos Cancer Institute, and Podcast Host at OncoDaily, shared a post on LinkedIn:
“UK researchers just proved accuracy isn’t the point!
New AI software reads stroke brain scans twice as accurately as doctors. Trained on 800 scans, tested on 2,000 patients.
But here’s what matters: It pinpoints when the stroke happened and whether it can still be treated—critical because treatments only work within 4.5 to 6 hours.
Researchers estimate this could enable 50% more stroke patients to be treated appropriately.
The breakthrough isn’t that it’s twice as accurate. It’s that it answers the question doctors actually need: ‘Can I save this patient?’
Most stroke patients can’t tell you when their stroke started. Some happen during sleep. Others affect speech.
Doctors eyeball the scan, guess how dark the lesion looks, estimate the time window. Every brain is different. Blood flow varies. The stroke could be progressing fast or slow.
There are types of MRI that enable you to have a better idea but these tests are not available everywhere!
This AI doesn’t just read darkness => it analyzes texture, accounts for variations, estimates biological age of the lesion. Then tells you: treat now or too late.
Speed over precision. Actionability over accuracy. Impact over metrics.
That’s how AI should work in healthcare.”
Read the full article: New AI stroke brain scan readings are twice as accurate as current method
More posts featuring Roupen Odabashian.