Multi-Scale Convolutional-Stack Aggregation for Robust White Matter Hyperintensities Segmentation

TY - GEN

T1 - Multi-Scale Convolutional-Stack Aggregation for Robust White Matter Hyperintensities Segmentation

AU - Li, Hongwei

AU - Zhang, Jianguo

AU - Muehlau, Mark

AU - Kirschke, Jan

AU - Menze, Bjoern

N1 - Conference code: 21

PY - 2019/1/26

Y1 - 2019/1/26

N2 - Segmentation of both large and small white matter hyperintensities/lesions in brain MR images is a challenging task which has drawn much attention in recent years. We propose a multi-scale aggregation model framework to deal with volume-varied lesions. Firstly, we present a specifically-designed network for small lesion segmentation called Stack-Net, in which multiple convolutional layers are ‘one-by-one’ connected, aiming to preserve rich local spatial information of small lesions before the sub-sampling layer. Secondly, we aggregate multi-scale Stack-Nets with different receptive fields to learn multi-scale contextual information of both large and small lesions. Our model is evaluated on recent MICCAI WMH Challenge Dataset and outperforms the state-of-the-art on lesion recall and lesion F1-score under 5-fold cross validation. It claimed the first place on the hidden test set after independent evaluation by the challenge organizer. In addition, we further test our pre-trained models on a Multiple Sclerosis lesion dataset with 30 subjects under cross-center evaluation. Results show that the aggregation model is effective in learning multi-scale spatial information.

AB - Segmentation of both large and small white matter hyperintensities/lesions in brain MR images is a challenging task which has drawn much attention in recent years. We propose a multi-scale aggregation model framework to deal with volume-varied lesions. Firstly, we present a specifically-designed network for small lesion segmentation called Stack-Net, in which multiple convolutional layers are ‘one-by-one’ connected, aiming to preserve rich local spatial information of small lesions before the sub-sampling layer. Secondly, we aggregate multi-scale Stack-Nets with different receptive fields to learn multi-scale contextual information of both large and small lesions. Our model is evaluated on recent MICCAI WMH Challenge Dataset and outperforms the state-of-the-art on lesion recall and lesion F1-score under 5-fold cross validation. It claimed the first place on the hidden test set after independent evaluation by the challenge organizer. In addition, we further test our pre-trained models on a Multiple Sclerosis lesion dataset with 30 subjects under cross-center evaluation. Results show that the aggregation model is effective in learning multi-scale spatial information.

KW - Deep learning

KW - White matter hyperintensities

UR - https://www.scopus.com/pages/publications/85063546070

U2 - 10.1007/978-3-030-11723-8_20

DO - 10.1007/978-3-030-11723-8_20

M3 - Conference contribution

SN - 9783030117221

VL - 11393

T3 - Lecture Notes in Computer Science

SP - 199

EP - 207

BT - Brainlesion

A2 - Crimi, Alessandro

A2 - Bakas, Spyridon

A2 - Kuijf, Hugo

A2 - Keyvan, Farahani

A2 - Reyes, Mauricio

A2 - van Walsum, Theo

PB - Springer Verlag

CY - Switzerland

T2 - MICCAI 2018

Y2 - 16 September 2018 through 20 September 2018

ER -