Dave's PCF WIP: Terms

Term #1,448

Edit Term

segmentation or image segmentation

View List Terms Delete

Claim Term

Application

App Docket

Created Date

Full Desc

"Segmentation" or "Image segmentation" refers to the process of partitioning 2D images, 3D volumes, or temporal image sequences into meaningful regions, classes, or instances. These segments may correspond to different tissue classes, organs, pathologies, bones, landmarks, background, or other biologically relevant structures. Medical image segmentation addresses challenges such as low contrast, noise, artifacts, occlusion, and acquisition variability.

Non-limiting examples include registration/atlas-based methods, shape/appearance models, level-set/active-contour methods, graph-cut/energy-minimization methods, and neural-network models such as convolutional or transformer-based encoder–decoders (e.g., U-Net) and hybrids.

Segmentation may be implemented using classical computer-vision, rules-based, statistical, or machine-learning/deep-learning techniques, alone or in combination, and may operate at native or resampled resolution. Outputs may include pixel-wise representations (e.g., binary masks, label maps, probability maps, heatmaps), instance masks, and/or contours, which may be used to derive anatomic data, constructed references, measurements, and annotated images.