Model-Based Needle Identification Using Image Analysis and Needle Library Matching for Ultrasound-Guided Kidney Biopsy: A Feasibility Study.

Abstract

The aim of the work described here was to determine the feasibility of using a novel biopsy needle detection technique that achieves high sensitivity and specificity in a trade-off of resolution, detectability and depth of imaging. The proposed needle detection method consists of a model-based image analysis, temporal needle projection and needle library matching: (i) Image analysis was formulated under the signal decomposition framework; (ii) temporal projection converted the time-resolved needle dynamics into a single image of the desired needle; and (iii) the enhanced needle structure was spatially refined by matching a long, straight linear object in the needle library. The efficacy was examined with respect to different needle visibility. Our method effectively eliminated confounding effects of the background tissue artifacts more robustly than conventional methods, thus improving needle visibility even with the low contrast between the needle and tissue. The improvement in needle structure further resulted in an improvement in estimation performance for the trajectory angle and tip position. Our three-step needle detection method can reliably detect needle position without the need for external devices, increasing the needle conspicuity and reducing motion sensitivity.