A doctor looks at a breast image to find a suspicious lump. If the picture is beautiful and smooth, but the lump has vanished from it, such an image is of little use. This is exactly the problem faced by researchers who developed a new way to test software for photoacoustic tomography.
Photoacoustic tomography (PACT) uses short laser pulses: tissues absorb light, warm up slightly, and begin to emit ultrasound. From these sound waves, the computer reconstructs an image of the interior. But different programs do this differently, and comparing them was difficult: there was no common standard or realistic data.
To address this, the scientists created a dataset of more than 11,000 virtual breast models with typical lesions.
Then they ran several reconstruction algorithms through these models — both traditional (physics-based) and modern (deep learning). It turned out that conventional image quality metrics, such as sharpness or absence of noise, do not always reflect what matters most: whether the tumor itself is visible in the image. Some programs produced perfect-looking images but lost important details.
Therefore, the authors proposed adding "clinical" tests: checking how accurately the program recovers the size and shape of the lesion. This approach will help create more reliable tools for breast cancer diagnosis.
In photoacoustic tomography, sound is born from light: a laser pulse heats the blood in vessels by fractions of a degree, and that is enough to make the tissues "sound."