The focus of the Biomedical Modeling Laboratory (BML) is the development of new paradigms in detection, diagnosis, characterization, and treatment of disease through the integration of computational models into research and clinical practice.
In the past, clinicians skilled at image-guided surgery have relied solely on preoperative scans for their navigational information. Often in the course of surgery, tissue is purposely retracted/resected or inadvertently moved resulting in a significant misregistration between the preoperative scans and the current state of the operating field. This research proposes to use a model-updating image-guided technique which employs standard image-guided techniques augmented by a predictive model which is driven by low-cost intraoperatively acquired data (e.g. surface deformation measurements, and co-reigstered ultrasound). The model-image guided approach is shown here:
One interesting aspect that has come out of this research is the manner in which we are characterizing the organ shape during surgery. Currently, we are using a laser range scanner to measure organ surfaces during surgery. Below is a movie of the laser range scanner "in action" and a collage of the data from this ball phantom. One unusual aspect to the laser range scanner is that it captures both geometric features as well as texture. In the images to the right, the spherical geometric point cloud can be seen (in white) followed by successive images whereby the point cloud is color-encoded with intensity values of the field of view. These textured point clouds are rich sources of data for the model-updating method.
It is of equal importance to demonstrate that the laser range scanner can capture similar data within the operating room (OR). Below to the left is a photo of the brain surface during surgery. The textured-point cloud is shown in the movie to the right.
We are currently using these sources of data for developing novel methods for surface registration and deformation measurement within the brain and liver. The predominant model-updated image-guided surgery efforts in the laboratory are directed at neuro- and liver surgery.
Often, standard imaging modalities cannot detect or differentiate cancerous regions due to a fundamental limitation in the relationship between the physical imaging principles and the pathology. Alternative imaging modalities attempt to image tissue properties based on known constitutive laws thought to reflect the tissue continuum. These new imaging techniques are related to an area of research known as inverse problems and their full potential has not yet been realized. The focus of this research is to use computational models in conjunction with new data acquisition techniques to generate reconstructed images of material properties and to characterize these new modalities with respect to both detection and diagnostic capabilities. Current projects are focused in the area of elastography which is concerned with characterizing tissue by its mechanical properties.
While all elastography techniques are of interest, one particular method that we have been developing is called Modality Independent Elastography (MIE). MIE combines image processing with an inverse problem framework. More specifically, image similarity metrics routinely used with image registration methods are recast to satisfy an inverse elasticity problem framework whereby mechanical properties within a biomechanical model of deforming tissue become the driving parameters for improved image similarity. In this way, MIE circumvents two potential limitations of current elastographic techniques. First, it is not inherently dependent on pre-processing steps such as homologous feature selection and tracking which drive active contour models and other traditional displacement-based iterative methods. Secondly, because it is an image processing methodology, MIE is not reliant on a particular imaging modality (such as in ultrasound and magnetic resonance elastography) as long as the acquired images provide sufficient pattern to allow for comparison. To date we have demonstrated the feasibility of MIE with CT, MR, and optical images.
The two images below represent cross-sectional images of a normal breast as taken by CT and MR, respectively. While these areas of increased tissue stiffness may be present clinically, there is no guarantee that tissue morphological changes will reflect these regions in CT or MR. MIE could be used to augment these modalities to provide an additional layer of information.
As a test to our methodology, we generated two data sets whereby some regions of increasing stiff mechanical properties are present within these images (1 and 2 regions). The MIE reconstruction is shown below:
MIE CT Reconstruction (targets were highlighted in this set)
MIE MR Reconstruction:
While the above is encouraging and images are based on real clinical data, the reconstructions in this case are from simulated deformation acquisition. We have performed some preliminary experiments using membranes and optical (i.e. photographic) data which represents a third modality for MIE. Below is an experiment where by a membrane is stretched. It is easy to notice the stiff inclusion from the surrounding deflection of the grid pattern.
This data was used as an input to the MIE algorithm and the reconstructions are shown below.
The goal of this research is to model and understand the effect of a transjugular intrahepatic portosystemic shunt (TIPS) on splanchnic blood flow. TIPS is a corrective procedure for a life-threatening splanchnic ailment known as portal hypertension. This condition is attributed to an elevated hepatic vascular resistance most commonly caused by cirrhosis of the liver and pathologically characterized by the liver tissue becoming fibrous and stiff. By implanting a stent through scarred liver tissue, TIPS creates an alternate pathway for blood to travel from the portal vein to the hepatic vein. Although its immediate effects are quite dramatic, often, over a short period of time (6 months to a year), the stent becomes occluded by thrombogenic activity and hyperplasia, of which, blood flow dynamics play an integral role. This research is focused on correlating blood flow dynamics from the TIPS procedure with the onset of stent occlusion in the hopes of understanding how to avoid blockages by varying implantation techniques.
 |
Return to Top |