Spectral analysis of the tremor motion for needle detection in curvilinear ultrasound via spatiotemporal linear sampling
Autor: | Gary C. Ng, Septimiu E. Salcudean, Robert Rohling, Parmida Beigi |
---|---|
Rok vydání: | 2016 |
Předmět: |
Motion analysis
Channel (digital image) Swine Physics::Medical Physics Biomedical Engineering Optical flow Health Informatics Standard deviation Displacement (vector) 030218 nuclear medicine & medical imaging Motion 03 medical and health sciences Spatio-Temporal Analysis 0302 clinical medicine Tremor Animals Humans Coherence (signal processing) Radiology Nuclear Medicine and imaging Computer vision Least-Squares Analysis Root-mean-square deviation Ultrasonography Physics business.industry Spectrum Analysis General Medicine Computer Graphics and Computer-Aided Design Computer Science Applications Surgery Computer-Assisted Needles Trajectory Surgery Computer Vision and Pattern Recognition Artificial intelligence business 030217 neurology & neurosurgery |
Zdroj: | International Journal of Computer Assisted Radiology and Surgery. 11:1183-1192 |
ISSN: | 1861-6429 1861-6410 |
DOI: | 10.1007/s11548-016-1402-7 |
Popis: | This paper presents a new approach to detect a standard handheld needle in ultrasound-guided interventions. Our proposal is to use natural hand tremor, which causes minute displacement of the needle, to detect the needle in ultrasound B-mode images. Subtle displacements arising from tremor motion have a periodic pattern which is usually imperceptible to the naked eye in the B-mode image. We use these displacement measurements in a spatiotemporal framework to detect linear structures with periodic pattern among a sequence of frames. The needle trajectory is estimated as a linear path in the image having maximum spectral correlation with the time trace of displacement due to tremor. A coarse estimation process is followed by a fine estimation step, where the motion pattern is analyzed along spatiotemporal linear paths with various angles originating from the estimated puncture site, within the trajectory channel. Spectral coherency is derived for each sample path versus the reference path, and the needle trajectory is identified as the mean of the sample paths with the maximum coherence within the tremor frequency range. To evaluate the detection accuracy, we tested the method in vivo on porcine tissue, where the needle was inserted into the biceps femoris muscle. To understand whether tremor itself affects needle position, the maximum angular change due to tremor was calculated: mean, standard deviation (SD) and root-mean-square (RMS) measurement of $$0.43^\circ , 0.23^\circ $$ and $$0.48^\circ $$ . The accuracy of the needle trajectory was calculated by comparing to an expert manual segmentation, averaged over the captured data and presented in mean, SD and RMS error of $$2.83^\circ , 1.64^\circ $$ and $$3.23^\circ $$ , respectively. Results demonstrate that natural tremor motion creates minute coherent motion along the needle, which could be used to localize the needle trajectory within the acceptable accuracy. This method is suitable for standard needles used clinically. |
Databáze: | OpenAIRE |
Externí odkaz: |