Click here to


Are you sure ?

Yes, do it No, cancel

A Robust Tracking Algorithm for Implanted Marker of Fluoroscopic Images in a Real-Time Tumor-Tracking System

K Miyazaki1*, Y Fujii1 , T Umekawa1 , T Yamada1 , T FUJII2 , N Miyamoto2 , (1) Hitachi, Ltd., Hitachi-shi, Ibaraki,(2) Hokkaido University, Sapporo, Hokkaido


(Tuesday, 7/31/2018) 9:30 AM - 10:00 AM

Room: Exhibit Hall | Forum 6

Purpose: In a real-time tumor-tracking system, an algorithm for tracking an implanted marker may misidentify statistical noises of fluoroscopic images as the marker when the intensity of an X-ray is extremely low. The aim is to develop an algorithm that is robust to the statistical noise.

Methods: The proposed method (1) enhances the marker, (2) extracts marker candidates, and (3) separates noise. For a pair of fluoroscopic images taken from two different orthogonal angles, the window level and width are adjusted to enhance the marker. Then one or more marker candidates are extracted by template matching. As these candidates include not only the marker but also statistical noises, a length of common perpendicular (LCP) is calculated for all combinations of candidates from each fluoroscopic image. Because LCP is nearly equal to 0 mm when the candidates are the projection of the actual marker, the algorithm can distinguish the marker and noises. This method was validated by using chest phantom images taken with various X-ray tube currents (20, 40, and 60 mA). The tube voltage and exposure duration were 100 kV and 3 ms. A 2 mm gold marker and 200-mm-thick plastic plate were placed in front of the chest phantom, and the marker was moved in the 1D-direction. 2D projected marker position errors by the proposed and conventional (only template matching) methods were calculated for 450 frames.

Results: Maximum 2D position errors were 0.49, 0.55, and 0.99 mm for the proposed method and 0.46, 0.47, and 121.4 mm for conventional method (20, 40 and 60 mA). At 20 mA, the conventional method often misidentified statistical noise as the marker.

Conclusion: The proposed method can prevent the misidentification of statistical noise as an implanted marker and improve the tracking accuracy compared with the conventional method.

Funding Support, Disclosures, and Conflict of Interest: This research was partially supported by Development of Medical Devices and Systems for Advanced Medical Services from Japan Agency for Medical Research and development, AMED.


Image-guided Therapy, Fluoroscopy, Noise


IM- X-ray: Image processing

Contact Email