Lung Segmentation

The outline is what turns a position on the image into a sentence a person can use: upper or lower, left or right, inside the lung fields or outside them. It is drawn as an optional overlay on your result and never alters the image beneath it. It also gives the pipeline a way to stop: about one scan in thirty-five is rejected because no lungs could be found at all, which is a better outcome than a verdict about something that is not a chest.

It is trained to find air-filled lung, and consolidated lung — the thing this whole product is looking for — resembles soft tissue to it and gets left out. So the two models carry opposite biases: the classifiers are trained to react to consolidation, the segmenter to exclude it. That asymmetry is the reason its positive claims can be trusted and its negative ones cannot. A highlighted region falling outside this outline is not evidence of an artefact, and on a flagged scan it must never be read as one.