Download PDFOpen PDF in browser

Spine Surface Segmentation from Ultrasound Using Multi-feature Guided CNN

5 pagesPublished: October 26, 2019

Abstract

Accurate, robust, and real-time segmentation of bone surfaces is an essential objective for ultrasound (US) guided computer assisted orthopedic surgery (CAOS) procedures. In this work, we present a convolutional neural network (CNN)-based technique for segmenting spine surfaces from in vivo US scans. Proposed design utilizes fusion of feature maps extracted from multimodal images to abate sensitivity to variations caused by imaging artifacts and low intensity bone boundaries. In particular, our multimodal inputs consist of B-mode US images and their corresponding local phase filtered counterparts. Validation studies performed on 261 in vivo US scans obtained from 10 subjects achieved a mean localization accuracy of 0.1 mm with an F-score of 97%. Comparison against state-of-the-art CNN networks show an improvement of 89% in bone surface localization accuracy.

Keyphrases: bone segmentation, convolutional neural network, deep learning, fusion, spine, ultrasound

In: Patrick Meere and Ferdinando Rodriguez Y Baena (editors). CAOS 2019. The 19th Annual Meeting of the International Society for Computer Assisted Orthopaedic Surgery, vol 3, pages 6-10.

BibTeX entry
@inproceedings{CAOS2019:Spine_Surface_Segmentation_from,
  author    = {Ahmed Alsinan and Michael Vives and Vishal Patel and Ilker Hacihaliloglu},
  title     = {Spine Surface Segmentation from Ultrasound Using Multi-feature Guided CNN},
  booktitle = {CAOS 2019. The 19th Annual Meeting of the International Society for Computer Assisted Orthopaedic Surgery},
  editor    = {Patrick Meere and Ferdinando Rodriguez Y Baena},
  series    = {EPiC Series in Health Sciences},
  volume    = {3},
  publisher = {EasyChair},
  bibsource = {EasyChair, https://easychair.org},
  issn      = {2398-5305},
  url       = {/publications/paper/jnVv},
  doi       = {10.29007/f5fs},
  pages     = {6-10},
  year      = {2019}}
Download PDFOpen PDF in browser