Complex vascular lesion. Vessels are shown alone at left and selected vessel trees are shown relative to the tumor surface at right. Supported by NIH-NCI R01-CA67812

One of the reasons why treating complex AVMs and base of the brain tumors is so difficult is that is often difficult to understand the vascular feeding patterns. If a vessel passing through the lesion is interrupted, the patient may suffer a stroke.

Presenting a manipulable, 3D display of the lesion's surface relative to the vasculature can be helpful during surgical planning and execution. The figure above shows a complex vascular lesion. At right, selected vessel subtrees are shown relative to the tumor surface so as to best understand the contributions of the anterior choroidal artery.

A second major problem is that of mapping preoperative images to the intraoperative patient since, during surgery, there can be major distortions produced by retraction and/or edema as well as tissue loss of unknown amount. Dr. Stephen Aylward is developing a new, very fast, model-to-image form of registration in which vessels segmented from a preoperative image can be deformably mapped to intraoperatively acquired ultrasound. Since vessels surround and permeate all organs of the human body, this approach seems well-suited to conditions in which the surgeon can see only a portion of the anatomy and in which unknown distortions and amounts of tissue loss have occurred.

The figure below is courtesy of Dr. Stephen Aylward, and illustrates registration of intraoperative ultrasound images to preoperative MR images of the liver.

Bullitt E, Aylward S, Bernard E, Gerig G (2001) Special Article. Computer-assisted visualization of arteriovenous malformations on the home pc (2001) Neurosurgery: 48: 576-583. <pdf>

Bullitt E, Aylward SR (2003) Patient-specific vascular models for endovascular and open operative procedures. International Congress Series 1247C:129-138 <pdf>.

Aylward SR, Jomier J, Weeks S, Bullitt E (2003) Registration and analysis of vascular images. International Journal of Computer Vision. 55:123-138 <pdf>.