Studies demonstrate that the use of 3D printed models can improve preoperative surgical planning and enhance patient-provider communication. We present a pilot program of a 3D printed breast anatomy using transmission ultrasound (TU). Our 3D printed models allow for differentiation of normal breast tissue, including the skin, fat, and fibroglandular tissues, as well as delineation of underlying masses in selected cases.

We have performed non-invasive 3D TU imaging of whole in vivo breasts immersed in a water bath, with patients imaged in the prone position. The image reconstruction of the projection data results in co-registered 3D reflection, speed-of-sound, and attenuation images. We use machine learning at a voxel level to quantitively differentiate and segment breast tissues types. These tissue types are then 3D printed into respective anatomy models.

We have developed a fully-automated breast segmentation algorithm using reflection, speed-of-sound, and attenuation tissue characteristics that operate on transmission ultrasound images. The machine learning algorithm employs image voxel values from the co-registered images to classify breast tissue types: skin, fat, and fibroglandular tissues. We have validated this classifier on whole-breast TU images to provide a color-coded classification of the breast tissue volume and outline the presence of breast masses in selected cases, followed by 3D printing of individual tissue types.

Transmission ultrasound can generate whole-breast image volumes that can be objectively segmented and color-coded by tissue type. This unique ability to individually image 3D segmented breast tissue volumes provides a potentially clinically useful technology for 3D printed models.

3D printing of individual breast tissue types based on TU imaging has a wide range of potential clinical benefits, including preoperative surgical planning and improved patient-provider communication.