Learning the shape of female breasts: an open-access 3D statistical shape model of the female breast built from 110 breast scans

• DOI: 10.1007/s00371-022-02431-3
• Publication Date: 2022-03-05T13:02:43Z
• Authors: Maximilian Weiherer; Andreas Eigenberger; Bernhard Egger; Vanessa Brébant; Lukas Prantl; Christoph Palm
• Keywords: Breast tissue; Breast imaging; Thorax (insect anatomy)

Abstract We present the Regensburg Breast Shape Model (RBSM)—a 3D statistical shape model of female breast built from 110 scans acquired in a standing position, and first publicly available. Together with model, fully automated, pairwise surface registration pipeline used to establish dense correspondence among is introduced. Our method computationally efficient requires only four landmarks guide process. A major challenge when modeling breasts surface-only non-separability thorax. In order weaken strong coupling between surrounding areas, we propose minimize variance outside region as much possible. To achieve this goal, novel concept called probability masks (BPMs) BPM assigns probabilities each point scan, telling how likely it that particular belongs area. During registration, use BPMs align template target accurately possible inside roughly outside. This simple yet effective strategy significantly reduces unwanted region, leading better models which shapes are quite well decoupled The RBSM thus able produce variety different independently systematic experimental evaluation reveals generalization ability 0.17 mm specificity 2.8 mm. underline expressiveness proposed finally demonstrate two showcase applications can be for surgical outcome simulation prediction missing remaining one. available at https://www.rbsm.re-mic.de/ .