Comparison of 3D Image Reconstruction Algorithms from Tracked 2D Free-Hand Ultrasound Scans

Ultrasound (US) provides an attractive means of acquiring 3D medical images. However, 3D US transducers face a clinical-use limitation due to spatial constraints in endocavital scans. Circumventing these limitations is possible through 3D image reconstruction algorithms from 2D images. Our group aims to employ these algorithms in cancer diagnosis using a 2D transrectal US (TRUS) probe. To compare the accuracy of 3D image reconstruction algorithms, we require a method of quantifying image quality. We predict that a Signal-to-Noise Ratio (SNR) quantitative analysis will improve upon existing image quality metrics. To test this hypothesis, volumetric image data of a pelvic US phantom (taken by a 3D transducer) will be used as ground truth for comparing reconstruction accuracy. We will artificially extract 2D slices from the volume to verify the reconstruction process. Then a 2D probe—spatially localized by an optical tracker—will acquire real 2D images. We will reconstruct the artificial and real 2D image sets in 3D using three different algorithms. We will then compare the quality of the reconstructed images with two known methods: quantitatively by comparing intensity values and qualitatively by visual analysis. Finally, we will test the SNR hypothesis by comparing its results to those of the previously described methods. This study will provide a basis for improving 3D image reconstruction algorithms for use in 2D TRUS. Our group plans to use these results as a guideline for acquiring 2D tracked freehand US images for cancer diagnosis applications – such as preclinical trials of contrast-enhanced molecular US imaging.