Carnegie Mellon University
Advanced Search   
  Look in
       Name    Email
       Former RI Members 
Devin Amin
PhD Student, BioMed
No longer a member of RI.
Research Interests

Image-guided surgery combines current surgical practice with quantitative anatomical data derived from preoperative medical images. A surgical plan is developed based on three-dimensional preoperative images such as computed tomography (CT) or magnetic resonance (MR) scans. This plan, for example, may include the desired trajectory for a bone screw, the removal of a small region of tissue while sparing nearby critical structures, or the accurate alignment of a total hip joint replacement procedure. Image-guided surgery techniques rely on the ability to register the preoperative images to the patient's anatomy on the operating table and track any movement of the patient's anatomy throughout the procedure. Currently, the surgical plan is typically registered to the patient's anatomy using surface based registration techniques. The locations of points along the bone surface are collected with a optically tracked probe that accurately measures three-dimensional positions. One of the problems with the current technique is that the bone surfaces must be surgically exposed for the probe to collect points for registration. Thus surface based registration must be accomplished using only the regions of the bone surface which would normally be exposed by the surgical procedure. Our goal is to develop a non-invasive ultrasound based method to register and track the patient's bony anatomical structures within the operating room. By tracking the ultrasound probe the images can be used to provide detailed information about the 3D locations of anatomical structures within the operating room. Registration of the ultrasound images with the preoperative images allows precise determination of the 3D loaction and orientation of the patient's anatomy on the operating table. Currently available medical image registration techniques rely on the mapping of separate tissue classes to distinct image intensities. However, the image intensity with ultrasound is produced by transitions in the acoustic impedance rather than by the specific impedance property of the tissue. Thus, in the ultrasound image the transitions between tissue classes are greatly enhanced. For the transition between soft tissue and bone, only the surface of the bone is visible in the ultrasound image. The ultrasound image of an anatomical area is also dependent on depth since the ultrasound energy is being absorbed and reflected as it penetrates the tissue. To compensate for these image modality differences a model based on the physics of ultrasound imaging was constructed. The model can produce simulated ultrasound images based on the CT data and the ultrasound probe's location. These simulated ultrasound images are compared to the actual ultrasound images. The similarity of the images is optimized over the entire set of images collected in the operating room, producing the registration estimate.

Additional Interests

I have graduated from Carnegie Mellon University with my PhD and am now finishing the medical school portion of the MD/PhD program at the University of Pittsburgh.

Research Interest Keywords
bioengineeringmedical imagingmedical robotics