PATHFINDER

PATHFINDER is an AI system that stitches millions of brain cross-section photos together into a single three-dimensional map of neural circuits. Electron-microscope images of a fly's brain are stacked in thin layers, like slices of bread, and to reassemble them into the shape of a whole brain, you have to figure out which pixels belong to the same neuron across every slice. PATHFINDER is the reconstruction system that does this automatically.

The system grew out of technology Google Research originally built to recognize objects in ordinary photos, later adapted and developed over a long period for brain cross-section imagery. The latest version mixes computer-generated synthetic neuron images into the training data, boosting both the speed and accuracy of reconstruction. Even so, the automatically assembled result still has to be checked case by case by human experts before it becomes a brain map that can actually be trusted.

The article notes that "recently, synthetic neurons have been mixed into the training data, boosting both the speed and accuracy of PATHFINDER, the latest reconstruction system." It's worth keeping straight that PATHFINDER itself isn't a finished brain map — it's the reconstruction tool used to stitch the images together into one.