Computer Science > Computer Vision and Pattern Recognition
[Submitted on 25 Dec 2016 (this version), latest version 21 Jun 2017 (v4)]
Title:Pancreas Segmentation in Abdominal CT Scan: A Coarse-to-Fine Approach
View PDFAbstract:Deep neural networks have been widely adopted for automatic organ segmentation from CT-scanned images. However, the segmentation accuracy on some small organs (e.g., the pancreas) is sometimes below satisfaction, arguably because deep networks are easily distracted by the complex and variable background region which occupies a large fraction of the input volume.
In this paper, we propose a coarse-to-fine approach to deal with this problem. We train two deep neural networks using different regions of the input volume. The first one, the coarse-scaled model, takes the entire volume as its input. It is used for roughly locating the spatial position of the pancreas. The second one, the fine-scaled model, only sees a small input region covering the pancreas, thus eliminating the background noise and providing more accurate segmentation especially around the boundary areas. At the testing stage, we first use the coarse-scaled model to roughly locate the pancreas, then adopt the fine-scaled model to refine the initial segmentation in an iterative manner to obtain increasingly better segmentation. We evaluate our algorithm on the NIH pancreas segmentation dataset with 82 volumes, and outperform the state-of-the-art [18] by more than 4% measured by the Dice-Sorensen Coefficient (DSC). In addition, we report 62.43% DSC for our worst case, significantly better than 34.11% reported in [18].
Submission history
From: Lingxi Xie [view email][v1] Sun, 25 Dec 2016 02:15:50 UTC (974 KB)
[v2] Mon, 29 May 2017 07:41:05 UTC (947 KB)
[v3] Sun, 18 Jun 2017 02:52:24 UTC (947 KB)
[v4] Wed, 21 Jun 2017 04:00:59 UTC (947 KB)
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
Connected Papers (What is Connected Papers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.