Ultrasound waves bounce off organs to draw their outlines, but on an ovarian scan it's hard to tell a harmless fluid-filled cyst from a solid cancerous mass. Even specialists draw the boundaries differently, and this directly affects treatment choices.
The OvAi Focus program tackles this by learning from thousands of examples. Its approach resembles spectroscopy: just as astrophysicists split a galaxy's light into a spectrum to reveal its chemical makeup, the algorithm breaks down the ultrasound image into layers — here's fluid in the cavity, here's dense tissue rich in carbon, and here's healthy stroma.
After testing on 1,081 women from six clinics, the program showed high accuracy. For whole tumors, the overlap with a doctor's contours (DICE score) reached 0.87, and for the fluid part, 0.85. Surprisingly, the program recognized healthy ovaries less well — only 0.62. The machine finds it easier to 'see' pathology. Thanks to this automated labeling, doctors can assess risks faster and won't miss a subtle danger.
Ovarian cysts are so common that almost every woman gets a microscopic cyst at least once in her life: it's part of the body's normal function.
The same 'deconstruct and observe' principle powers the sci-fi medical scanners in Star Trek, which instantly show any organ in microscopic detail.