Chinese Team Builds M³Net AI to Spot Lung Nodules More Accurately

Lung cancer is a common deadly disease, and early screening is key to saving lives. Reading CT scans by hand takes time, while regular AI models only look at one scale of image features, which limits their accuracy.

A research team from the Chinese Academy of Sciences Hefei Institutes and its partners has developed a new model called M³Net. It copies how doctors think, using three input scales to capture the whole lung, tissue around the nodule, and fine details, following a "big picture first, then local, then details" reading flow.

In tests on public and clinical datasets, M³Net beat 12 popular AI algorithms in accuracy. It is especially good at spotting tricky nodules with unclear edges, and it runs at lower cost, offering a new way to make AI diagnosis both accurate and easier to understand.