Quantifying emphysema extent from weakly labeled CT scans of the lungs using label proportions learning

TY - GEN

T1 - Quantifying emphysema extent from weakly labeled CT scans of the lungs using label proportions learning

AU - Ørting, Silas Nyboe

AU - Petersen, Jens

AU - Wille, Mathilde

AU - Thomsen, Laura

AU - de Bruijne, Marleen

N1 - Conference code: 6

PY - 2016

Y1 - 2016

N2 - Quantification of emphysema extent is important in diagnosing and monitoring patients with chronic obstructive pulmonary disease (COPD). Several studies have shown that emphysema quantification by supervised texture classification is more robust and accurate than traditional densitometry. Current techniques require highly time consuming manual annotations of patches or use only weak labels indicating overall disease status (e.g, COPD or healthy). We show how visual scoring of regional emphysema extent can be exploited in a learning with label proportions (LLP) framework to both predict presence of emphysema in smaller patches and estimate regional extent. We evaluate performance on 195 visually scored CT scans and achieve an intraclass correlation of 0.72 (0.65–0.78) between predicted region extent and expert raters. To our knowledge this is the first time that LLP methods have been applied to medical imaging data.

AB - Quantification of emphysema extent is important in diagnosing and monitoring patients with chronic obstructive pulmonary disease (COPD). Several studies have shown that emphysema quantification by supervised texture classification is more robust and accurate than traditional densitometry. Current techniques require highly time consuming manual annotations of patches or use only weak labels indicating overall disease status (e.g, COPD or healthy). We show how visual scoring of regional emphysema extent can be exploited in a learning with label proportions (LLP) framework to both predict presence of emphysema in smaller patches and estimate regional extent. We evaluate performance on 195 visually scored CT scans and achieve an intraclass correlation of 0.72 (0.65–0.78) between predicted region extent and expert raters. To our knowledge this is the first time that LLP methods have been applied to medical imaging data.

M3 - Article in proceedings

SN - 978-1537038582

SP - 31

EP - 42

BT - The Sixth International Workshop on Pulmonary Image Analysis

A2 - Beichel, Reinhard R.

A2 - Farahani, Keyvan

A2 - Jacobs, Colin

A2 - Kabus, Sven

A2 - Kiraly, Atilla P.

A2 - Kuhnigk, Jan-Martin

A2 - McClelland, Jamie R.

A2 - Mori, Kensaku

A2 - Petersen, Jens

A2 - null, Simon Rit

PB - CreateSpace Independent Publishing Platform

T2 - Sixth International Workshop on Pulmonary Image Analysis

Y2 - 21 October 2016 through 21 October 2016

ER -