An Injury-Mimicking Ultrasound Phantom as a Training Tool for Diagnosis of Internal Trauma

Rowan, Matthew Ivan

Etd

An Injury-Mimicking Ultrasound Phantom as a Training Tool for Diagnosis of Internal Trauma

公开

Ultrasound phantoms that mimic injury are training devices that can emulate pre- and post-injury conditions within specific regions of human anatomy. They have the potential to be useful tools for teaching medical personnel how to recognize trauma conditions based on ultrasound images. This is particularly important because the increased use of portable ultrasound systems allows earlier diagnosis of internal trauma at locations such as traffic accidents, earthquakes, battlefields and terrorist attacks. A physical injury mimicking ultrasound phantom of the peritoneal cavity was constructed that mimicked the ultrasonic appearance of internal bleeding. Bleeding was simulated by injecting 600 mL of fluid of varying densities into the bulk of the phantom and comparing the ultrasonic appearance to before bleeding was simulated. The physical phantom was used to investigate whether or not the density of the injected fluid had any influence on the increase of inter-organ fluid volumes. The physical phantom was imaged in 3D with a 4.5 MHz phased array transducer, and two fluid volumes were segmented using the segmentation software ITK-SNAP. The 3D image representation of the phantom showed a difference qualitatively and quantitatively between pre-injury and post-injury conditions. Qualitatively, the physical model was analyzed. These specific criteria were analyzed within each image: 1) the number of individual organs that are present, 2) the number of other organs that each individual organ touches, 3) the appearance of fluid between the organs and the scanning membrane and 4) the merging of two separate fluid pockets. Using a Wilcoxon Rank-Sum test, a statistically significant difference was shown to exist between pre-injury and post-injury ultrasound images with a 95% level of confidence. Quantitatively, a Chi-Squared test was used to show that the volume of fluid between adjacent organs, calculated by ITK-SNAP, had no dependence on the density of the injected fluid. Furthermore, using a one-tailed T-test, there was at least a 99.9% confidence that the inter-organ volume estimations for the pre-injury and post-injury configurations were statistically different. As a final means of evaluation, the experimental phantom was taken to Harvard Medical School in November 2006 and analyzed by ultrasonographers. The doctors were very excited about its potential uses and found other interesting characteristics that the phantom was not designed for. In addition to modeling the appearance of an injected fluid volume, visualization of fluid flowing into the phantom, modeling the appearance of air in the inter-peritoneal space and simulating a surgical tool or bandage being accidentally left inside the patient could be modeled as well. The injury mimicking phantom was also modeled numerically, using ADINA finite element software. Using the same external dimensions as the experimental model, the numerical model showed that for physiologically unrealistic, very high fluid injection densities, the displacement of the organs had no statistical dependence on the density of the injected fluid, using an acceptance criterion of: P-value < 0.05. This was confirmed using an F-test of the average organ phantom tip displacement tabulated at several different times during simulation. The P-value obtained for analyzing the average tip displacement was 0.0506. However, a plot of the mass ratio, an expression of how the injected fluid has dispersed into the bulk of the phantom, showed that an unrealistically high fluid injection density had a different mass ratio profile than the other fluid injection densities that were simulated. This F-test revealed a strong indication, P-value = 0.0069, that the very high density caused a different fluid dispersion pattern. The numerical phantom offered a distinct advantage over the experimental model in that the dispersion of the injected fluid could be modeled numerically but not observed experimentally. Modeling the phantom numerically had some disadvantages. The numerical model had to have a large gap between adjacent organs. This had to occur because the contact algorithm within ADINA is incapable of modeling dynamic contact when fluid-structure interactions are modeled. This led to a volume fraction representation of the solid domain that was too low compared with the experimental model and what is found anatomically. For future iterations of the injury mimicking phantom, the numerical model will be used to help design the physical phantoms.

Creator