Automatic Segmentation of Vertebrae in Ultrasound Images

被引：1

作者：

Berton, Florian ^{[1
]}

Azzabi, Wassim ^{[2
]}

Cheriet, Farida ^{[1
]}

Laporte, Catherine ^{[2
]}

机构：

[1] Ecole Polytech, Montreal, PQ H3T 1J4, Canada

[2] Ecole Technol Super, Montreal, PQ H3C 1K3, Canada

来源：

IMAGE ANALYSIS AND RECOGNITION (ICIAR 2015) | 2015年 / 9164卷

关键词：

Segmentation; Vertebrae; Ultrasound; Acoustic shadow; Random forests; IDENTIFICATION;

D O I：

10.1007/978-3-319-20801-5_37

中图分类号：

TP18 [人工智能理论];

学科分类号：

140502 [人工智能];

摘要：

This paper presents an automatic method for the segmentation of vertebrae in ultrasound images. Its goal is to determine whether each pixel belongs to the bone surface, its acoustic shadow or other tissues. The method is based on the extraction of several image features described in the literature and which we adapted to our problem, and on a random forest classifier. Morphological operations and vertebra-specific constraints are then used in a regularisation step in order to obtain homogeneous regions of both the surface and the acoustic shadow of the vertebra. Experiments on a test database of 9 images show promising results, with average recognition rates for the bone surface and acoustic shadow of 81.87 %, and 91.01 %, respectively.

引用

页码：344 / 351

页数：8

共 14 条

[1]

An Augmented Reality System for Epidural Anesthesia (AREA): Prepuncture Identification of Vertebrae [J].