Fusing deep learned and hand-crafted features of appearance, shape, and dynamics for automatic pain estimation

Automatic continuous time, continuous value assessment of a patient's pain from face video is highly sought after by the medical profession. Despite the recent advances in deep learning that attain impressive results in many domains, pain estimation risks not being able to benefit from this due...

Full description

Bibliographic Details
Main Authors: Egede, Joy Onyekachukwu, Valstar, Michel F., Martinez, Brais
Format: Conference or Workshop Item
Published: 2017
Subjects:
Online Access:https://eprints.nottingham.ac.uk/40801/
_version_ 1848796136825946112
author Egede, Joy Onyekachukwu
Valstar, Michel F.
Martinez, Brais
author_facet Egede, Joy Onyekachukwu
Valstar, Michel F.
Martinez, Brais
author_sort Egede, Joy Onyekachukwu
building Nottingham Research Data Repository
collection Online Access
description Automatic continuous time, continuous value assessment of a patient's pain from face video is highly sought after by the medical profession. Despite the recent advances in deep learning that attain impressive results in many domains, pain estimation risks not being able to benefit from this due to the difficulty in obtaining data sets of considerable size. In this work we propose a combination of hand-crafted and deep-learned features that makes the most of deep learning techniques in small sample settings. Encoding shape, appearance, and dynamics, our method significantly outperforms the current state of the art, attaining a RMSE error of less than 1 point on a 16-level pain scale, whilst simultaneously scoring a 67.3% Pearson correlation coefficient between our predicted pain level time series and the ground truth.
first_indexed 2025-11-14T19:43:11Z
format Conference or Workshop Item
id nottingham-40801
institution University of Nottingham Malaysia Campus
institution_category Local University
last_indexed 2025-11-14T19:43:11Z
publishDate 2017
recordtype eprints
repository_type Digital Repository
spelling nottingham-408012020-05-04T18:47:22Z https://eprints.nottingham.ac.uk/40801/ Fusing deep learned and hand-crafted features of appearance, shape, and dynamics for automatic pain estimation Egede, Joy Onyekachukwu Valstar, Michel F. Martinez, Brais Automatic continuous time, continuous value assessment of a patient's pain from face video is highly sought after by the medical profession. Despite the recent advances in deep learning that attain impressive results in many domains, pain estimation risks not being able to benefit from this due to the difficulty in obtaining data sets of considerable size. In this work we propose a combination of hand-crafted and deep-learned features that makes the most of deep learning techniques in small sample settings. Encoding shape, appearance, and dynamics, our method significantly outperforms the current state of the art, attaining a RMSE error of less than 1 point on a 16-level pain scale, whilst simultaneously scoring a 67.3% Pearson correlation coefficient between our predicted pain level time series and the ground truth. 2017-05-30 Conference or Workshop Item PeerReviewed Egede, Joy Onyekachukwu, Valstar, Michel F. and Martinez, Brais (2017) Fusing deep learned and hand-crafted features of appearance, shape, and dynamics for automatic pain estimation. In: 12th IEEE Conference on Face and Gesture Recognition (FG 2017), 30 May-3 June 2017, Washington, D.C., U.S.A.. Pain Estimation Feature extraction Face Shape Physiology Machine learning http://ieeexplore.ieee.org/abstract/document/7961808/
spellingShingle Pain
Estimation
Feature extraction
Face
Shape
Physiology
Machine learning
Egede, Joy Onyekachukwu
Valstar, Michel F.
Martinez, Brais
Fusing deep learned and hand-crafted features of appearance, shape, and dynamics for automatic pain estimation
title Fusing deep learned and hand-crafted features of appearance, shape, and dynamics for automatic pain estimation
title_full Fusing deep learned and hand-crafted features of appearance, shape, and dynamics for automatic pain estimation
title_fullStr Fusing deep learned and hand-crafted features of appearance, shape, and dynamics for automatic pain estimation
title_full_unstemmed Fusing deep learned and hand-crafted features of appearance, shape, and dynamics for automatic pain estimation
title_short Fusing deep learned and hand-crafted features of appearance, shape, and dynamics for automatic pain estimation
title_sort fusing deep learned and hand-crafted features of appearance, shape, and dynamics for automatic pain estimation
topic Pain
Estimation
Feature extraction
Face
Shape
Physiology
Machine learning
url https://eprints.nottingham.ac.uk/40801/
https://eprints.nottingham.ac.uk/40801/