Temporal convolution networks for real-time abdominal fetal aorta analysis with ultrasound

TY - CHAP

T1 - Temporal convolution networks for real-time abdominal fetal aorta analysis with ultrasound

AU - Savioli, Nicoló

AU - Visentin, Silvia

AU - Cosmi, Erich

AU - Grisan, Enrico

AU - Lamata, Pablo

AU - Montana, Giovanni

PY - 2018/1/1

Y1 - 2018/1/1

N2 - The automatic analysis of ultrasound sequences can substantially improve the efficiency of clinical diagnosis. In this work we present our attempt to automate the challenging task of measuring the vascular diameter of the fetal abdominal aorta from ultrasound images. We propose a neural network architecture consisting of three blocks: a convolutional layer for the extraction of imaging features, a Convolution Gated Recurrent Unit (C-GRU) for enforcing the temporal coherence across video frames and exploiting the temporal redundancy of a signal, and a regularized loss function, called CyclicLoss, to impose our prior knowledge about the periodicity of the observed signal. We present experimental evidence suggesting that the proposed architecture can reach an accuracy substantially superior to previously proposed methods, providing an average reduction of the mean squared error from 0.31mm2 (state-of-art) to 0.09mm2, and a relative error reduction from 8.1% to 5.3%. The mean execution speed of the proposed approach of 289 frames per second makes it suitable for real time clinical use.

AB - The automatic analysis of ultrasound sequences can substantially improve the efficiency of clinical diagnosis. In this work we present our attempt to automate the challenging task of measuring the vascular diameter of the fetal abdominal aorta from ultrasound images. We propose a neural network architecture consisting of three blocks: a convolutional layer for the extraction of imaging features, a Convolution Gated Recurrent Unit (C-GRU) for enforcing the temporal coherence across video frames and exploiting the temporal redundancy of a signal, and a regularized loss function, called CyclicLoss, to impose our prior knowledge about the periodicity of the observed signal. We present experimental evidence suggesting that the proposed architecture can reach an accuracy substantially superior to previously proposed methods, providing an average reduction of the mean squared error from 0.31mm2 (state-of-art) to 0.09mm2, and a relative error reduction from 8.1% to 5.3%. The mean execution speed of the proposed approach of 289 frames per second makes it suitable for real time clinical use.

KW - Cardiac imaging

KW - Convolutional networks

KW - CyclicLoss

KW - Diameter

KW - Fetal imaging

KW - GRU

KW - Ultrasound

UR - http://www.scopus.com/inward/record.url?scp=85054883392&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-01421-6_15

DO - 10.1007/978-3-030-01421-6_15

M3 - Conference paper

AN - SCOPUS:85054883392

SN - 9783030014209

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

SP - 148

EP - 157

BT - Artificial Neural Networks and Machine Learning – ICANN 2018 - 27th International Conference on Artificial Neural Networks, 2018, Proceedings

PB - Springer Verlag

T2 - 27th International Conference on Artificial Neural Networks, ICANN 2018

Y2 - 4 October 2018 through 7 October 2018

ER -