Appel Choose France
L’appel Choose France est ouvert jusqu’au 31 mars. C’est une belle opportunité pour attirer en France des talents...
7 December 2023
Catégorie : Stagiaire
Title :Assessment of patient pain based on Deep Learning approaches using multimodal data
Keywords :Chronic pain, Deep Learning, Multimodal data, Machine Learning, Body Movement Analysis, Facial Expression Analysis
Description :
Automated assessment of pain experiences aims to enhance personalized care, empower patients, and improve self-management of chronic conditions through digital technology [1, 2, 3]. Datasets play a role in achieving these objectives. They not only form the foundation for developing machine learning models for automated assessment but also contribute significantly to gaining a deeper understanding of the support needs. This understanding informs the integration of assessment technology into the overall care framework.
EmoPain@Home is a dataset that consists of body movements data captured during functional activities at home, from both people with and without chronic pain. Chronic pain is of particular significance due to its effect on the sense of self, engagement in valued activities, and interaction with others [4, 5]. The EmoPain@Home dataset is densely labelled with self-reported levels of pain and related worry and confidence for the participants with chronic pain. It additionally includes labels of the activities performed, for both participants with and without chronic pain.
Missions :
This internship focuses on developing a multimodal deep learning approach for the automatic recognition of activity contexts, leveraging facial and body movement data. Deep learning methods, including CNNs, RNNs, and GCNs, are gaining attention in the assessment of patient body movement. Inspired by Graph Convolutional Networks (GCN) [6], our approach aims to capture spatial and temporal dependencies within body movements, while facial expressions will be analyzed using architectures inspired by ConvNeXts [7]. By leveraging the advantages of GCNs and ConvNeXts, we can achieve a high degree of precision in evaluating patient pain in real, complex settings of everyday life. An essential criterion for our approach is real-time performance, as it needs to seamlessly integrate with a real-time demonstrator that we intend to develop and assess pain using Kinect technology. This internship opportunity offers a compelling exploration of the dynamic field surrounding the validation and verification of chronic pain in people’s homes, driven by advancements in Deep Learning. The research internship will have the following objectives :
1. Improve the performances of an existing approach for pain detection based on a new multi-modal deep learning architecture.
2. Evaluate the model’s performance using identified public datasets,
3. Implement the tool in a GUI that can be integrated into the complete workflow of the patient (and used by healthcare professionals).
4. Test the results with real data using a Kinect camera.
5. Write an article for an international conference
Required Skills :
This internship is open to students in the second year of computer science or those pursuing an equivalent degree. The selection committee will pay particular attention to the following profiles :
— Good knowledge of machine/deep learning methods
— Excellent proficiency in Python programming
— Mastery of PyTorch or TensorFlow libraries is indispensable
— Autonomy
Duration & Contact :
The internship, lasting 5/6 months, will take place at the CESI École d’Ingénieurs - Campus Dijon. The application package should include a comprehensive CV, post-baccalaureate transcripts with rankings, and one or more letters of recommendation. It should be submitted before 31/12/2023 to Amine BOHI (abohi@cesi.fr) and Youssef MOURCHID (ymourchid@cesi.fr), both professors and researchers at CESI LINEACT.
Références :
[1]Francisco RAvilaet al. « Wearable electronic devices for chronic pain intensity assessment : A systematic review ». In :Pain Practice21.8 (2021), p. 955-965.
[2]JerryChen, MaysamAbbodet Jiann-ShingShieh. « Pain and stress detection using wearable sensors and devices—A review ». In :Sensors21.4 (2021), p. 1030.
[3]SteffenWalteret al. « “What about automated pain recognition for routine clinical use ?” A survey of physicians and nursing staff on expectations, requirements, and acceptance ». In : Frontiers in medicine7 (2020), p. 566278.
[4]KaiKaroset al. « Pain as a threat to the social self : a motivational account ». In :Pain 159.9 (2018), p. 1690-1695.
[5]Caitlin BMurrayet al. « Long-term impact of adolescent chronic pain on young adult educational, vocational, and social outcomes ». In :Pain161.2 (2020), p. 439.
[6]ZonghanWuet al. « A comprehensive survey on graph neural networks ». In :IEEE transactions on neural networks and learning systems32.1 (2020), p. 4-24.
[7]ZhuangLiuet al. « A convnet for the 2020s ». In :Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2022, p. 11976-11986.