Improving Catheter Segmentation & Localization in 3D Cardiac Ultrasound Using Direction-Fused Fcn

Abstract

Fast and accurate catheter detection in cardiac catheterization using harmless 3D ultrasound (US) can improve the efficiency and outcome of the intervention. However, the low image quality of US requires extra training for sonographers to localize the catheter. In this paper, we propose a catheter detection method based on a pre-trained VGG network, which exploits 3D information through re-organized cross-sections to segment the catheter by a shared fully convolutional network (FCN), which is called a Direction- Fused FCN (DF-FCN). Based on the segmented image of DF-FCN, the catheter can be localized by model fitting. Our experiments show that the proposed method can successfully detect an ablation catheter in a challenging ex-vivo 3D US dataset, which was collected on the porcine heart. Extensive analysis shows that the proposed method achieves a Dice score of 57.7%, which offers at least an 11.8% improvement when compared to state-ofthe-art instrument detection methods. Due to the improved segmentation performance by the DF-FCN, the catheter can be localized with an error of only 1.4 mm.

Medical Instrument Detection in Ultrasound-Guided Interventions: A Review

“…However, in terms of 3D volumetric data, the decomposition approach limits the semantic information usage due to the compromised 3D information after slicing. To address this limitation, patch-based 2.5D or 3D UNet were proposed to segment the cardiac catheter [83], [84], [85] or prostate needles [86] in 3D volumetric data by dividing the image into smaller patches, which preserves the 3D contextual information and clearly reduces the GPU memory requirements for 3D deep learning. Nevertheless, this patch-based strategy limits the whole image contextual information usage.…”

“…Needle biopsy/anesthesia/therapy ex-vivo Kaya et al [41] 2015 2D+t Needle biopsy/drug delivery in-vitro Pourtaherian et al [37] 2016 3D Needle anesthesia/ablation ex-vivo Beigi et al [45] 2016 2D+t Needle biopsy/nerve block/anesthesias in-vitro/in-vivo Mwikirize et al [43] 2016 2D Needle biopsy/ablation/anesthesia ex-vivo Beigi et al [48] 2016 2D+t Needle biopsy/nerve block/anesthesia in-vivo Kaya et al [46] 2016 2D+t Needle biopsy/drug delivery in-vitro Daoud et al [49] 2018 3D Needle intervention ex-vivo Daoud et al [24] 2018 2D Needle intervention ex-vivo Agarwal et al [35] 2019 2D+t Anesthesia/biopsy/brachytherapy in-vitro Needle biopsy/ablation/anesthesia ex-vivo Yang et al [61] 2019 3D Cardiac catheterization in-vitro/ex-vivo/in-vivo Yang et al [70] 2019 3D Cardiac catheterization ex-vivo Yang et al [82] 2019 3D Cardiac catheterization ex-vivo Yang et al [83] 2019 3D Cardiac catheterization ex-vivo Mwikirize et al [89] 2019 2D+t Needle biopsy/anesthesia in-vitro/ex-vivo Mwikirize et al [88] 2019 2D+t Needle biopsy/anesthesia ex-vivo Arif et al [87] 2019 3D Needle biopsy in-vitro/in-vivo Yang et al [85] 2019 3D Cardiac catheterization ex-vivo/in-vivo Min et al [76] 2020 3D Cardiac catheterization ex-vivo Rodgers et al [80] 2020 2D/3D Interstitial gynecologic brachytherapy in-vitro/in-vivo Zhang et al [68], [67] 2020 3D prostate brachytherapy in-vivo Zhang et al [86] 2020 3D Prostate brachytherapy in-vivo Zhang et al [90] 2020 3D Prostate brachytherapy in-vivo Lee et al [79] 2020 2D Needle biopsy in-vivo With the above summaries, there are remaining some challenges and limitations for this area, despite the current methods obtain satisfactory results. We discuss these challenges below.…”

End-to-End Real-time Catheter Segmentation with Optical Flow-Guided Warping during Endovascular Intervention

“…Machine learning techniques are also widely used for catheter segmentation and tracking [35]- [38]. With the rise of deep learning, methods based on Convolutional Neural Networks (CNN) are adapted for catheter segmentation [39] [40]. Early work in [41] used a simple neural network to detect chest tubes then post-processed the results using a curve fitting technique to connect discontinued segments.…”

Masked convolutional neural network for supervised learning problems

“…In the past decade, convolutional neural network (CNN) models have received much attention due to their competitive performance in various tasks, including object classification (He et al, ; Arvidsson et al, ; Xie et al, ) and semantic segmentation (Badrinarayanan, Kendall, & Cipolla, ; Chen, Papandreou, Kokkinos, Murphy, & Yuille, ; Yang, Shan, Kolen, & De With Peter, ; Falk et al, ). For instance, in object classification problems, we may be interested in identifying objects in the images that are associated with class labels.…”