Global and local interpretability for cardiac MRI classification

TY - CHAP

T1 - Global and local interpretability for cardiac MRI classification

AU - Clough, James R.

AU - Oksuz, Ilkay

AU - Puyol-Antón, Esther

AU - Ruijsink, Bram

AU - King, Andrew P.

AU - Schnabel, Julia A.

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Deep learning methods for classifying medical images have demonstrated impressive accuracy in a wide range of tasks but often these models are hard to interpret, limiting their applicability in clinical practice. In this work we introduce a convolutional neural network model for identifying disease in temporal sequences of cardiac MR segmentations which is interpretable in terms of clinically familiar measurements. The model is based around a variational autoencoder, reducing the input into a low-dimensional latent space in which classification occurs. We then use the recently developed ‘concept activation vector’ technique to associate concepts which are diagnostically meaningful (eg. clinical biomarkers such as ‘low left-ventricular ejection fraction’) to certain vectors in the latent space. These concepts are then qualitatively inspected by observing the change in the image domain resulting from interpolations in the latent space in the direction of these vectors. As a result, when the model classifies images it is also capable of providing naturally interpretable concepts relevant to that classification and demonstrating the meaning of those concepts in the image domain. Our approach is demonstrated on the UK Biobank cardiac MRI dataset where we detect the presence of coronary artery disease.

AB - Deep learning methods for classifying medical images have demonstrated impressive accuracy in a wide range of tasks but often these models are hard to interpret, limiting their applicability in clinical practice. In this work we introduce a convolutional neural network model for identifying disease in temporal sequences of cardiac MR segmentations which is interpretable in terms of clinically familiar measurements. The model is based around a variational autoencoder, reducing the input into a low-dimensional latent space in which classification occurs. We then use the recently developed ‘concept activation vector’ technique to associate concepts which are diagnostically meaningful (eg. clinical biomarkers such as ‘low left-ventricular ejection fraction’) to certain vectors in the latent space. These concepts are then qualitatively inspected by observing the change in the image domain resulting from interpolations in the latent space in the direction of these vectors. As a result, when the model classifies images it is also capable of providing naturally interpretable concepts relevant to that classification and demonstrating the meaning of those concepts in the image domain. Our approach is demonstrated on the UK Biobank cardiac MRI dataset where we detect the presence of coronary artery disease.

KW - Cardiac MRI

KW - Coronary artery disease

KW - Interpretable ML

UR - http://www.scopus.com/inward/record.url?scp=85075650743&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-32251-9_72

DO - 10.1007/978-3-030-32251-9_72

M3 - Conference paper

AN - SCOPUS:85075650743

SN - 9783030322502

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

SP - 656

EP - 664

BT - Medical Image Computing and Computer Assisted Intervention – MICCAI 2019 - 22nd International Conference, Proceedings

A2 - Shen, Dinggang

A2 - Yap, Pew-Thian

A2 - Liu, Tianming

A2 - Peters, Terry M.

A2 - Khan, Ali

A2 - Staib, Lawrence H.

A2 - Essert, Caroline

A2 - Zhou, Sean

PB - SPRINGER

T2 - 22nd International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2019

Y2 - 13 October 2019 through 17 October 2019

ER -