Imperial College London
Portrait Picture

ProfessorAndreaRockall

Faculty of Medicine, Department of Surgery & Cancer

Clinical Chair in Radiology
 
 
 
//

Contact

 

a.rockall

 
 
//

Location

 

ICTEM buildingHammersmith Campus

//

Summary

 

Publications

Citation

BibTex format

@unpublished{Kanavati:2020,
author = {Kanavati, F and Islam, S and Arain, Z and Aboagye, EO and Rockall, A},
publisher = {arXiv},
title = {Fully-automated deep learning slice-based muscle estimation from CT  images for sarcopenia assessment},
url = {http://arxiv.org/abs/2006.06432v1},
year = {2020}
}
Download

RIS format (EndNote, RefMan)

TY  - UNPB
AB  - Objective: To demonstrate the effectiveness of using a deep learning-basedapproach for a fully automated slice-based measurement of muscle mass forassessing sarcopenia on CT scans of the abdomen without any case exclusioncriteria.  Materials and Methods: This retrospective study was conducted using acollection of public and privately available CT images (n = 1070). The methodconsisted of two stages: slice detection from a CT volume and single-slice CTsegmentation. Both stages used Fully Convolutional Neural Networks (FCNN) andwere based on a UNet-like architecture. Input data consisted of CT volumes witha variety of fields of view. The output consisted of a segmented muscle mass ona CT slice at the level of L3 vertebra. The muscle mass is segmented intoerector spinae, psoas, and rectus abdominus muscle groups. The output wastested against manual ground-truth segmentation by an expert annotator.  Results: 3-fold cross validation was used to evaluate the proposed method.The slice detection cross validation error was 1.41+-5.02 (in slices). Thesegmentation cross validation Dice overlaps were 0.97+-0.02, 0.95+-0.04,0.94+-0.04 for erector spinae, psoas, and rectus abdominus, respectively, and0.96+-0.02 for the combined muscle mass.  Conclusion: A deep learning approach to detect CT slices and segment musclemass to perform slice-based analysis of sarcopenia is an effective andpromising approach. The use of FCNN to accurately and efficiently detect aslice in CT volumes with a variety of fields of view, occlusions, and slicethicknesses was demonstrated.
AU  - Kanavati,F
AU  - Islam,S
AU  - Arain,Z
AU  - Aboagye,EO
AU  - Rockall,A
PB  - arXiv
PY  - 2020///
TI  - Fully-automated deep learning slice-based muscle estimation from CT  images for sarcopenia assessment
UR  - http://arxiv.org/abs/2006.06432v1
UR  - http://hdl.handle.net/10044/1/80196
ER  -
Download