Volumetric soft tissue perfusion assessment on a region basis from x-ray angiography images: Motion compensation.

Abstract

Background: Assessing the soft tissue perfusion quantitatively in interventional suites before, during, and after interventional procedures is desired. The method, if possible, has to assess the perfusion volumetrically and quantitatively, be robust against lesion overlaps and patient motion, require no additional radiation dose, be quick (possibly in real-time), and fit to the clinical workflow well. We have developed a method called IPEN (for Intra-operative PErfusion assessment with No gantry rotation) that has potential to accomplish all of the desired goals except for the patient motion. The innovation with IPEN is not to reconstruct volumetric images, but to estimate enhancement of multiple three-dimensional regions-of-interest directly from x-ray projections acquired at one angle.

Purpose: To further develop the IPEN method such that it can compensate for patient motion when the patient moves quickly during the angiography scan but stays still otherwise.

Methods: The proposed motion-compensating IPEN (MCI) consists of the following three steps: (Step 1) The time segment is broken into multiple segments, that is, a set of rapid motion segments and a set of stationary segments; (Step 2) the MCI estimates ROI enhancement within each stationary segment; and (Step 3) MCI connects segments. The performance of the proposed MCI and the original IPEN were assessed using the digital perfusion phantom, simulating 13 ischemic stroke "patients." The head moved within 0.6 s each time, and seven times during 16-s scans; motion magnitude parameter a (for ± a mm and ± a degrees) was 0.0 (no motion), 0.5, 2.0, 5.0, and 25.0 for each scan. The accuracy of time-enhancement curves (TECs) and calculated perfusion-like parameter ("max-slope" for the maximum of slope of TEC; similar to Patlak plot analysis) was assessed. In addition, the effect of the motion segments on the accuracy of the estimated TEC has been studied systematically.

Results: Head motion induced very severe inconsistency and artifact in synthesized digital subtraction angiography images. The original IPEN had disjoint TECs, and the correlation coefficients (r) against the true values decreased from 0.475 at a = 0.5 to 0.023 at a = 25.0. The proposed MCI provided smooth and accurate TECs with r = 0.995 at a = 0.5 and r = 0.989 at a = 25.0. The 𝓁₂-norm of the error vectors of the max-slope values was 5.6-64.2 (d.l.) for the original IPEN, whereas it was < 0.1 for the MCI for the motion magnitudes investigated. There was an strong linear relationship between the non-linearity of the derivative of TECs and biases in TEC: r was 0.999. MCI would have a significant bias when a lengthy motion occurs when an ROI enhancement changes non-linearly during the time.

Conclusion: The proposed MCI can compensate for the patient motion very effectively and accurately when the motion is not continuous and the ROI enhancement does not change non-linearly and significantly during the motion segment.