Projection-Based Needle Segmentation in 3D Ultrasound Images

Ding, Mingyue; Fenster, Aaron

doi:10.1007/978-3-540-39903-2_40

Cited by 13 publications

(12 citation statements)

References 12 publications

(25 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…From several studies and clinical practices, it is known that off-line medical imaging does not provide enough accuracy for precise procedures. Therefore, real-time imaging techniques have been developed [23][24][25]. Image registration, image segmentation and augmented reality are the major topics in image-guided surgery.…”

Section: Solutionmentioning

confidence: 99%

Needle insertion into soft tissue: A survey

Abolhassani

Patel

Moallem

2007

Medical Engineering & Physics

626

431

View full text Add to dashboard Cite

Section: Solutionmentioning

confidence: 99%

Needle insertion into soft tissue: A survey

Abolhassani

Patel

Moallem

2007

Medical Engineering & Physics

626

431

View full text Add to dashboard Cite

“…Image-based approaches for tool guidance, avoiding the requirement for additional equipment, signal processing, and set-up, have been proposed for general tool segmentation in US images, leveraging techniques based on image properties, including projections, [8][9][10][11] random sample consensus (RANSAC), [12][13][14] filtering, 13,[15][16][17] and Hough or Radon transforms, [18][19][20][21][22] or physical properties, such as analyses of motion, [22][23][24][25][26] beam steering, 27 and circular wave generation. 28 Many of these algorithms were developed for 3D US images; [8][9][10][11][12][13][14][15]20,21,23 however, real-time 2D US is the clinical standard for guiding tools during image-guided minimally invasive interventions at most institutions 1,2 and therefore is the focus of our work presented in this study. Many of the published approaches were tested only on phantom or ex vivo tissue images, [8][9][10][11][12][13][14]…”

Section: Introductionmentioning

confidence: 99%

Deep learning segmentation of general interventional tools in two‐dimensional ultrasound images

et al. 2020

Self Cite

View full text Add to dashboard Cite

Purpose: Many interventional procedures require the precise placement of needles or therapy applicators (tools) to correctly achieve planned targets for optimal diagnosis or treatment of cancer, typically leveraging the temporal resolution of ultrasound (US) to provide real-time feedback. Identifying tools in two-dimensional (2D) images can often be time-consuming with the precise position difficult to distinguish. We have developed and implemented a deep learning method to segment tools in 2D US images in near real-time for multiple anatomical sites, despite the widely varying appearances across interventional applications. Methods: A U-Net architecture with a Dice similarity coefficient (DSC) loss function was used to perform segmentation on input images resized to 256 × 256 pixels. The U-Net was modified by adding 50% dropouts and the use of transpose convolutions in the decoder section of the network. The proposed approach was trained with 917 images and manual segmentations from prostate/gynecologic brachytherapy, liver ablation, and kidney biopsy/ablation procedures, as well as phantom experiments. Real-time data augmentation was applied to improve generalizability and doubled the dataset for each epoch. Postprocessing to identify the tool tip and trajectory was performed using two different approaches, comparing the largest island with a linear fit to random sample consensus (RAN-SAC) fitting. Results: Comparing predictions from 315 unseen test images to manual segmentations, the overall median [first quartile, third quartile] tip error, angular error, and DSC were 3.5 [1.3, 13.5] mm, 0.8 [0.3, 1.7]°, and 73.3 [56.2, 82.3]%, respectively, following RANSAC postprocessing. The predictions with the lowest median tip and angular errors were observed in the gynecologic images (median tip error: 0.3 mm; median angular error: 0.4°) with the highest errors in the kidney images (median tip error: 10.1 mm; median angular error: 2.9°). The performance on the kidney images was likely due to a reduction in acoustic signal associated with oblique insertions relative to the US probe and the increased number of anatomical interfaces with similar echogenicity. Unprocessed segmentations were performed with a mean time of approximately 50 ms per image. Conclusions: We have demonstrated that our proposed approach can accurately segment tools in 2D US images from multiple anatomical locations and a variety of clinical interventional procedures in near real-time, providing the potential to improve image guidance during a broad range of diagnostic and therapeutic cancer interventions.

show abstract

“…Software-based approaches include: localization using Radon/Hough transform to detect linear (curved) structures, 9, 10 line filtering to increase micro-tools contrast with respect to tissue structures, 11 projection-based algorithms to detect highest intensity line-like features 12,13 and model fitting approaches using RANSAC algorithm. 11,14 Techniques in this group rely on the needle to be at least partially visible in the image with line-like features.…”

Section: Introductionmentioning

confidence: 99%

Automatic detection of a hand-held needle in ultrasound via phased-based analysis of the tremor motion

Beigi

Salcudean

Rohling

et al. 2016

Medical Imaging 2016: Image-Guided Procedures, Robotic Interventions, and Modeling

View full text Add to dashboard Cite

This paper presents an automatic localization method for a standard hand-held needle in ultrasound based on temporal motion analysis of spatially decomposed data. Subtle displacement arising from tremor motion has a periodic pattern which is usually imperceptible in the intensity image but may convey information in the phase image. Our method aims to detect such periodic motion of a hand-held needle and distinguish it from intrinsic tissue motion, using a technique inspired by video magnification. Complex steerable pyramids allow specific design of the wavelets' orientations according to the insertion angle as well as the measurement of the local phase. We therefore use steerable pairs of even and odd Gabor wavelets to decompose the ultrasound Bmode sequence into various spatial frequency bands. Variations of the local phase measurements in the spatially decomposed input data is then temporally analyzed using a finite impulse response bandpass filter to detect regions with a tremor motion pattern. Results obtained from different pyramid levels are then combined and thresholded to generate the binary mask input for the Hough transform, which determines an estimate of the direction angle and discards some of the outliers. Polynomial fitting is used at the final stage to remove any remaining outliers and improve the trajectory detection. The detected needle is finally added back to the input sequence as an overlay of a cloud of points. We demonstrate the efficiency of our approach to detect the needle using subtle tremor motion in an agar phantom and in-vivo porcine cases where intrinsic motion is also present. The localization accuracy was calculated by comparing to expert manual segmentation, and presented in (mean, standard deviation and root-mean-square error) of (0.93 • , 1.26 • and 0.87 • ) and (1.53 mm, 1.02 mm and 1.82 mm) for the trajectory and the tip, respectively.

show abstract

Projection-Based Needle Segmentation in 3D Ultrasound Images

Cited by 13 publications

References 12 publications

Needle insertion into soft tissue: A survey

Needle insertion into soft tissue: A survey

Deep learning segmentation of general interventional tools in two‐dimensional ultrasound images

Automatic detection of a hand-held needle in ultrasound via phased-based analysis of the tremor motion

Contact Info

Product

Resources

About