Investigating image registration impact on preterm birth classification: An interpretable deep learning approach

TY - CHAP

T1 - Investigating image registration impact on preterm birth classification

T2 - 1st International Workshop on Smart Ultrasound Imaging, SUSI 2019, and the 4th International Workshop on Preterm, Perinatal and Paediatric Image Analysis, PIPPI 2019, held in conjunction with the 22nd International Conference on Medical Imaging and Computer-Assisted Intervention, MICCAI 2019

AU - Grigorescu, Irina

AU - Cordero-Grande, Lucilio

AU - David Edwards, A.

AU - Hajnal, Joseph V.

AU - Modat, Marc

AU - Deprez, Maria

PY - 2019

Y1 - 2019

N2 - Deep learning algorithms have recently become the dominant trend in medical image classification. However, the decision-making rationale of convolutional neural network (CNN) classifiers can be obscure. Interpretable machine learning techniques, such as layer-wise relevance propagation (LRP), can provide a visual interpretation of these decisions. In this work, we build a 3D CNN model to classify neonatal $$T:2$$ -weighted magnetic resonance (MR) scans into term or preterm. Additionally, we investigate the impact of different registration techniques applied to the image dataset on the classifier’s predictions. Finally, we compute LRP ‘relevance maps’, which indicate each voxel’s importance to the outcome of the decision. Our resulting LRP heatmaps show no visually striking differences between the different registration techniques, while also revealing anatomically plausible features for term and preterm birth.

AB - Deep learning algorithms have recently become the dominant trend in medical image classification. However, the decision-making rationale of convolutional neural network (CNN) classifiers can be obscure. Interpretable machine learning techniques, such as layer-wise relevance propagation (LRP), can provide a visual interpretation of these decisions. In this work, we build a 3D CNN model to classify neonatal $$T:2$$ -weighted magnetic resonance (MR) scans into term or preterm. Additionally, we investigate the impact of different registration techniques applied to the image dataset on the classifier’s predictions. Finally, we compute LRP ‘relevance maps’, which indicate each voxel’s importance to the outcome of the decision. Our resulting LRP heatmaps show no visually striking differences between the different registration techniques, while also revealing anatomically plausible features for term and preterm birth.

KW - Classification

KW - Layer-wise relevance propagation

KW - Preterm birth

UR - http://www.scopus.com/inward/record.url?scp=85075714674&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-32875-7_12

DO - 10.1007/978-3-030-32875-7_12

M3 - Conference paper

AN - SCOPUS:85075714674

SN - 9783030328740

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 104

EP - 112

BT - Smart Ultrasound Imaging and Perinatal, Preterm and Paediatric Image Analysis - 1st International Workshop, SUSI 2019, and 4th International Workshop, PIPPI 2019, Held in Conjunction with MICCAI 2019, Proceedings

A2 - Wang, Qian

A2 - Gomez, Alberto

A2 - Hutter, Jana

A2 - Gomez, Alberto

A2 - Zimmer, Veronika

A2 - Hutter, Jana

A2 - Robinson, Emma

A2 - Christiaens, Daan

A2 - Melbourne, Andrew

A2 - McLeod, Kristin

A2 - Zettinig, Oliver

A2 - Licandro, Roxane

A2 - Turk, Esra Abaci

PB - SPRINGER

Y2 - 17 October 2019 through 17 October 2019

ER -