Feature Learning Based on Visual Similarity Triplets in Medical Image Analysis: A Case Study of Emphysema in Chest CT Scans

TY - GEN

T1 - Feature Learning Based on Visual Similarity Triplets in Medical Image Analysis

T2 - 7th Joint International Workshop, CVII-STENT 2018<br/>and Third International Workshop, LABELS 2018

AU - Nyboe Ørting, Silas

AU - Petersen, Jens

AU - Cheplygina, Veronika

AU - Thomsen, Laura H.

AU - Wille, Mathilde M. W.

AU - De Bruijne, Marleen

PY - 2018

Y1 - 2018

N2 - Supervised feature learning using convolutional neural networks (CNNs) can provide concise and disease relevant representations of medical images. However, training CNNs requires annotated image data. Annotating medical images can be a time-consuming task and even expert annotations are subject to substantial inter- and intra-rater variability. Assessing visual similarity of images instead of indicating specific pathologies or estimating disease severity could allow non-experts to participate, help uncover new patterns, and possibly reduce rater variability. We consider the task of assessing emphysema extent in chest CT scans. We derive visual similarity triplets from visually assessed emphysema extent and learn a low dimensional embedding using CNNs. We evaluate the networks on 973 images, and show that the CNNs can learn disease relevant feature representations from derived similarity triplets. To our knowledge this is the first medical image application where similarity triplets has been used to learn a feature representation that can be used for embedding unseen test images.

AB - Supervised feature learning using convolutional neural networks (CNNs) can provide concise and disease relevant representations of medical images. However, training CNNs requires annotated image data. Annotating medical images can be a time-consuming task and even expert annotations are subject to substantial inter- and intra-rater variability. Assessing visual similarity of images instead of indicating specific pathologies or estimating disease severity could allow non-experts to participate, help uncover new patterns, and possibly reduce rater variability. We consider the task of assessing emphysema extent in chest CT scans. We derive visual similarity triplets from visually assessed emphysema extent and learn a low dimensional embedding using CNNs. We evaluate the networks on 973 images, and show that the CNNs can learn disease relevant feature representations from derived similarity triplets. To our knowledge this is the first medical image application where similarity triplets has been used to learn a feature representation that can be used for embedding unseen test images.

U2 - 10.1007/978-3-030-01364-6_16

DO - 10.1007/978-3-030-01364-6_16

M3 - Article in proceedings

SN - 978-3-030-01363-9

T3 - Lecture Notes in Computer Science

SP - 140

EP - 149

BT - Intravascular Imaging and Computer Assisted Stenting and Large-Scale Annotation of Biomedical Data and Expert Label Synthesis

A2 - Stoyanov, Danail

A2 - Taylor, Zeike

A2 - Balocco, Simone

A2 - Sznitman, Raphael

PB - Springer

Y2 - 18 September 2018

ER -