Cédric Dumas
Research Subjects
Cédric Dumas has a total of 54 co-authors. Here are the top co-authors by number of shared publications:
- Amine Chellali
IBISC Lab, Univ Evry, Université Paris-Saclay, Evry, France
10 shared publications Isabelle Milleville-Pennel
PsyCoTech Research Group, IRCCyN-CNRS
4 shared publications- Agus Budi Raharjo
Department of Informatics Engineering, Institut Teknologi Sepuluh Nopember, Surabaya, Indonesia
3 shared publications - Aurélien Léchappé
IMT Atlantique, LS2N, UMR CNRS 6004, Nantes, France and IMT Atlantique, Lab-STICC, UMR CNRS 6285, Brest, France
3 shared publications - Bilqis Amaliah
Department of Informatics, Institut Teknologi Sepuluh Nopember, Surabaya, Indonesia
3 shared publications - Caroline G. L. Cao
Tufts University
3 shared publications
Latest Publications
- Open Access
- 15 October 2026
- Published by Springer Nature in Journal on Multimodal User Interfaces
- p. 1-22
Social pressure: stress responses to virtual humans in immersive virtual job interviews
Prior research suggests that the attitude of virtual humans, in situations such as job interviews, can influence users’ stress levels. However, the mechanisms through which designers can precisely modulate stress levels in these settings remain unclear. We propose a mechanism for modulating users’ stress levels through multimodal social interactions with virtual humans. These interactions are modulated along two dimensions: friendliness (friendly vs. hostile) and dominance (dominant vs. non-dominant). To test this mechanism, we conducted two studies in which participants interacted with virtual humans that varied in friendliness and dominance. Friendliness was expressed through cues of appreciation or dislike, such as facial expressions (e.g., smiling or frowning) and speech content (e.g., “that’s good!”), whereas dominance was expressed through cues of control or lack of control, such as social rank (i.e., supervisor vs. subordinate) and assertiveness (i.e., giving instructions vs. asking permission). The results indicate that the behavior of the virtual humans plays a role in shaping stress responses; however, the findings reveal a complex relationship and therefore need to be interpreted with caution. In particular, the findings showed that the dominant virtual human was perceived as less friendly than the non-dominant virtual human. Future research should further investigate this effect and consider alternative models that provide more clearly distinguishable and independent control parameters for modulating stress responses in social virtual environments.
- 15 October 2026
- Published by Springer Nature in Journal on Multimodal User Interfaces
- p. 1-22
Open AccessSocial pressure: stress responses to virtual humans in immersive virtual job interviews
Prior research suggests that the attitude of virtual humans, in situations such as job interviews, can influence users’ stress levels. However, the mechanisms through which designers can precisely modulate stress levels in these settings remain unclear. We propose a mechanism for modulating users’ stress levels through multimodal social interactions with virtual humans. These interactions are modulated along two dimensions: friendliness (friendly vs. hostile) and dominance (dominant vs. non-dominant). To test this mechanism, we conducted two studies in which participants interacted with virtual humans that varied in friendliness and dominance. Friendliness was expressed through cues of appreciation or dislike, such as facial expressions (e.g., smiling or frowning) and speech content (e.g., “that’s good!”), whereas dominance was expressed through cues of control or lack of control, such as social rank (i.e., supervisor vs. subordinate) and assertiveness (i.e., giving instructions vs. asking permission). The results indicate that the behavior of the virtual humans plays a role in shaping stress responses; however, the findings reveal a complex relationship and therefore need to be interpreted with caution. In particular, the findings showed that the dominant virtual human was perceived as less friendly than the non-dominant virtual human. Future research should further investigate this effect and consider alternative models that provide more clearly distinguishable and independent control parameters for modulating stress responses in social virtual environments.
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 8408-8425
Peritoneal Carcinomatosis Detection Based on Improved YOLOv8 With Attention Mechanism and Optimized Backbone Integration
Peritoneal Carcinomatosis (PC) represents a severe form of metastatic cancer, with lesions spreading across the peritoneum and often leading to a poor prognosis. Early detection of PC lesions is essential for timely treatment, yet identifying smaller lesions in medical images remains challenging due to their subtle appearance and varied sizes. Traditional detection techniques often fall short in this area, underscoring the need for more advanced solutions. This study introduces improvements to the YOLOv8 architecture aimed at enhancing the detection of small PC lesions in laparoscopic images. The proposed modifications include adding a Convolutional Block Attention Module (CBAM) to improve feature extraction, incorporating a P2 layer for better detection of small objects, and using a CIoU loss function for more effective training. CIoU surpassed DIoU, WIoU, GIoU, EIoU, and SIoU loss functions in performance. Experimental results across Nano, Small, and Medium model scales demonstrated significant gains in Precision, Recall, mAP50, and F1 Score over the original YOLOv8. The modified model achieved a Precision of 0.887, Recall of 0.769, mAP50 of 0.861, and an F1 Score of 0.824. Ablation studies validated the benefits of each modification, with the enhanced model outperforming other YOLO versions (v5–v10) and transformer-based detectors such as RT-DETR, as well as alternative attention mechanisms including ECA and SA. These findings highlight the effectiveness of the proposed YOLOv8 modifications for detecting small lesions in medical imaging.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 8408-8425
Open AccessPeritoneal Carcinomatosis Detection Based on Improved YOLOv8 With Attention Mechanism and Optimized Backbone Integration
Peritoneal Carcinomatosis (PC) represents a severe form of metastatic cancer, with lesions spreading across the peritoneum and often leading to a poor prognosis. Early detection of PC lesions is essential for timely treatment, yet identifying smaller lesions in medical images remains challenging due to their subtle appearance and varied sizes. Traditional detection techniques often fall short in this area, underscoring the need for more advanced solutions. This study introduces improvements to the YOLOv8 architecture aimed at enhancing the detection of small PC lesions in laparoscopic images. The proposed modifications include adding a Convolutional Block Attention Module (CBAM) to improve feature extraction, incorporating a P2 layer for better detection of small objects, and using a CIoU loss function for more effective training. CIoU surpassed DIoU, WIoU, GIoU, EIoU, and SIoU loss functions in performance. Experimental results across Nano, Small, and Medium model scales demonstrated significant gains in Precision, Recall, mAP50, and F1 Score over the original YOLOv8. The modified model achieved a Precision of 0.887, Recall of 0.769, mAP50 of 0.861, and an F1 Score of 0.824. Ablation studies validated the benefits of each modification, with the enhanced model outperforming other YOLO versions (v5–v10) and transformer-based detectors such as RT-DETR, as well as alternative attention mechanisms including ECA and SA. These findings highlight the effectiveness of the proposed YOLOv8 modifications for detecting small lesions in medical imaging.
- Open Access
- 2 May 2025
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Human-Computer Interaction
- Vol. 9 (2) , 1-46
How to Categorize Collaboration during a Collaborative Puzzle-solving Task? Validation of Collaboration Profiles using Multimodal Data in Virtual Reality Context
In high-stakes collaborative situations, a decline in collaboration quality can lead to adverse events with significant consequences. Analyses performed by Human factor (HF) specialists, while effective in identifying and addressing collaboration issues, are case-specific and most of the time performed a posteriori. To address these limitations, our research focuses on a real-time assessment of collaboration processes using multimodal signals collected and analyzed during the activity. Existing collaboration profiles taxonomies face limitations such as a posteriori profiles detection and the absence of quantitative behavioral indicators that can be measured during the activity. Leveraging Virtual Reality (VR), we have developed a framework for evaluating collaboration in controlled setting, testing the effectiveness of a subset of multimodal signals to detect collaboration profiles. We test our approach in a study including 11 stereotyped collaborative scenarios applied to a VR puzzle-solving task. This study reveals the effectiveness of our approach in distinguishing between non-collaborative and highly collaborative profiles. However, challenges arise in discriminating between closely related collaborative profiles. This paper also proposes some guidelines on how to improve the collaboration profile detection framework and address other collaborative situations.
- 2 May 2025
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Human-Computer Interaction
- Vol. 9 (2) , 1-46
Open AccessHow to Categorize Collaboration during a Collaborative Puzzle-solving Task? Validation of Collaboration Profiles using Multimodal Data in Virtual Reality Context
In high-stakes collaborative situations, a decline in collaboration quality can lead to adverse events with significant consequences. Analyses performed by Human factor (HF) specialists, while effective in identifying and addressing collaboration issues, are case-specific and most of the time performed a posteriori. To address these limitations, our research focuses on a real-time assessment of collaboration processes using multimodal signals collected and analyzed during the activity. Existing collaboration profiles taxonomies face limitations such as a posteriori profiles detection and the absence of quantitative behavioral indicators that can be measured during the activity. Leveraging Virtual Reality (VR), we have developed a framework for evaluating collaboration in controlled setting, testing the effectiveness of a subset of multimodal signals to detect collaboration profiles. We test our approach in a study including 11 stereotyped collaborative scenarios applied to a VR puzzle-solving task. This study reveals the effectiveness of our approach in distinguishing between non-collaborative and highly collaborative profiles. However, challenges arise in discriminating between closely related collaborative profiles. This paper also proposes some guidelines on how to improve the collaboration profile detection framework and address other collaborative situations.
- Open Access
- 5 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 43532-43556
Deep Learning-Based Lesion Detection in Endoscopy: A Systematic Literature Review
A lesion is an abnormal change in body tissue or organ caused by various factors such as inflammation, infection, or abnormal cell growth. The presence of lesions often indicates a serious condition requiring medical attention. Early detection of lesions is crucial for effective treatment and preventing disease progression. Endoscopy is a key tool for lesion identification and evaluation, and the integration of deep learning techniques has further enhanced its potential. This systematic literature review (SLR) examines 35 studies to provide a comprehensive overview of lesion detection using deep learning on endoscopic images. The review explores deep learning techniques, modifications to baseline models, datasets, preprocessing and augmentation methods, and evaluation metrics. A meta-analysis of average performance metrics—accuracy, precision, recall, and specificity—was also conducted to identify trends across datasets and lesion types. Additionally, emerging trends such as federated learning, lightweight models, and ethical considerations are briefly discussed, addressing critical aspects for clinical translation. While significant advancements have been made, challenges remain, including dataset biases, scalability in resource-limited settings, and ensuring generalizability. Insights gained from this review are expected to guide future developments in lesion detection models, contributing to improved diagnostic accuracy, scalability, and patient outcomes.
- 5 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 43532-43556
Open AccessDeep Learning-Based Lesion Detection in Endoscopy: A Systematic Literature Review
A lesion is an abnormal change in body tissue or organ caused by various factors such as inflammation, infection, or abnormal cell growth. The presence of lesions often indicates a serious condition requiring medical attention. Early detection of lesions is crucial for effective treatment and preventing disease progression. Endoscopy is a key tool for lesion identification and evaluation, and the integration of deep learning techniques has further enhanced its potential. This systematic literature review (SLR) examines 35 studies to provide a comprehensive overview of lesion detection using deep learning on endoscopic images. The review explores deep learning techniques, modifications to baseline models, datasets, preprocessing and augmentation methods, and evaluation metrics. A meta-analysis of average performance metrics—accuracy, precision, recall, and specificity—was also conducted to identify trends across datasets and lesion types. Additionally, emerging trends such as federated learning, lightweight models, and ethical considerations are briefly discussed, addressing critical aspects for clinical translation. While significant advancements have been made, challenges remain, including dataset biases, scalability in resource-limited settings, and ensuring generalizability. Insights gained from this review are expected to guide future developments in lesion detection models, contributing to improved diagnostic accuracy, scalability, and patient outcomes.
- 18 October 2024
- Published by Springer Nature
- p. 93-104
Comparative Analysis of YOLO-Based Object Detection Models for Peritoneal Carcinomatosis
- 18 October 2024
- Published by Springer Nature
- p. 93-104
Comparative Analysis of YOLO-Based Object Detection Models for Peritoneal Carcinomatosis
- Open Access
- 8 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-4
Real-Time Feedback for Enhancing Team Collaboration: A Comparative Analysis of Collaboration Metrics with a Production Line Scenario using Virtual Reality
Effective collaboration is essential in high-stakes environments where poor teamwork can lead to critical errors and adverse outcomes. This Work in Progress aims to contribute to research by providing real-time feedback to prevent critical situations arising from inadequate collaboration. We are developing an experiment to compare seven indicators of collaboration for their effectiveness in real-time context. Using a collaborative virtual environment, we can control the situation and environmental effects, allowing for precise and reliable assessment of each indicator. The goal is to identify the most efficient indicators for real-time assessment of collaboration, thereby enhancing team performance and preventing critical failures. This research will contribute to optimizing teamwork and operational success in critical fields, such as industrial applications, where collaboration is crucial.
- 8 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-4
Open AccessReal-Time Feedback for Enhancing Team Collaboration: A Comparative Analysis of Collaboration Metrics with a Production Line Scenario using Virtual Reality
Effective collaboration is essential in high-stakes environments where poor teamwork can lead to critical errors and adverse outcomes. This Work in Progress aims to contribute to research by providing real-time feedback to prevent critical situations arising from inadequate collaboration. We are developing an experiment to compare seven indicators of collaboration for their effectiveness in real-time context. Using a collaborative virtual environment, we can control the situation and environmental effects, allowing for precise and reliable assessment of each indicator. The goal is to identify the most efficient indicators for real-time assessment of collaboration, thereby enhancing team performance and preventing critical failures. This research will contribute to optimizing teamwork and operational success in critical fields, such as industrial applications, where collaboration is crucial.
- Open Access
- 9 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 21-25
Characterization of collaboration in a virtual environment with gaze and speech signals
Collaboration during the completion of complex tasks requires synchronization and effective communication within teams. In this paper, we study multimodal collaboration metrics with the intention of designing collaboration support systems that enable the evaluation of collaboration quality. The goal is to provide real-time feedback and prevent the emergence of critical situations due to poor collaboration. We use a collaborative virtual environment to control the situation and the effects of the environment. As a first step to measure collaboration, we present a work in progress that aims to test the effectiveness of verbal and gaze measurements as an indicator of collaboration quality. To do so, an assembly task was performed by twelve dyads in our virtual environment. A tool was designed to process activity inputs in real time. Our findings show that verbal and gaze metrics can provide feedback on collaboration in virtual environment. Using these measures together with others would provide more accurate feedback on collaboration. Lastly, suggestions for improving the tool and an interest for other collaboration indicators are raised.
- 9 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 21-25
Open AccessCharacterization of collaboration in a virtual environment with gaze and speech signals
Collaboration during the completion of complex tasks requires synchronization and effective communication within teams. In this paper, we study multimodal collaboration metrics with the intention of designing collaboration support systems that enable the evaluation of collaboration quality. The goal is to provide real-time feedback and prevent the emergence of critical situations due to poor collaboration. We use a collaborative virtual environment to control the situation and the effects of the environment. As a first step to measure collaboration, we present a work in progress that aims to test the effectiveness of verbal and gaze measurements as an indicator of collaboration quality. To do so, an assembly task was performed by twelve dyads in our virtual environment. A tool was designed to process activity inputs in real time. Our findings show that verbal and gaze metrics can provide feedback on collaboration in virtual environment. Using these measures together with others would provide more accurate feedback on collaboration. Lastly, suggestions for improving the tool and an interest for other collaboration indicators are raised.
- 14 December 2022
- Published by Springer Nature
- p. 369-380
Epidural Motor Skills Measurements for Haptic Training
- 14 December 2022
- Published by Springer Nature
- p. 369-380
Epidural Motor Skills Measurements for Haptic Training
- Open Access
- 22 November 2022
- Published by Wiley in Annals of Clinical and Translational Neurology
- Vol. 9 (12) , 1863-1873
Multiple sclerosis clinical decision support system based on projection to reference datasets
Objective: Multiple sclerosis (MS) is a multifactorial disease with increasingly complicated management. Our objective is to use on‐demand computational power to address the challenges of dynamically managing MS.Methods: A phase 3 clinical trial data (NCT00906399) were used to contextualize the medication efficacy of peg‐interferon beta‐1a vs placebo on patients with relapsing–remitting MS (RRMS). Using a set of reference patients (PORs), selected based on adequate features similar to those of an individual patient, we visualize disease activity by measuring the percentage of relapses, accumulation of new T2 lesions on MRI, and worsening EDSS during the clinical trial.Results: We developed MS Vista, a functional prototype of clinical decision support system (CDSS), with a user‐centered design and distributed infrastructure. MS Vista shows the medication efficacy of peginterferon beta‐1a versus placebo for each individual patient with RRMS. In addition, MS Vista initiated the integration of a longitudinal magnetic resonance imaging (MRI) viewer and interactive dual physician‐patient data display to facilitate communication.Interpretation: The pioneer use of PORs for each individual patient enables personalized analytics sustaining the dialog between neurologists, patients and caregivers with quantified evidence.
- 22 November 2022
- Published by Wiley in Annals of Clinical and Translational Neurology
- Vol. 9 (12) , 1863-1873
Open AccessMultiple sclerosis clinical decision support system based on projection to reference datasets
Objective: Multiple sclerosis (MS) is a multifactorial disease with increasingly complicated management. Our objective is to use on‐demand computational power to address the challenges of dynamically managing MS.Methods: A phase 3 clinical trial data (NCT00906399) were used to contextualize the medication efficacy of peg‐interferon beta‐1a vs placebo on patients with relapsing–remitting MS (RRMS). Using a set of reference patients (PORs), selected based on adequate features similar to those of an individual patient, we visualize disease activity by measuring the percentage of relapses, accumulation of new T2 lesions on MRI, and worsening EDSS during the clinical trial.Results: We developed MS Vista, a functional prototype of clinical decision support system (CDSS), with a user‐centered design and distributed infrastructure. MS Vista shows the medication efficacy of peginterferon beta‐1a versus placebo for each individual patient with RRMS. In addition, MS Vista initiated the integration of a longitudinal magnetic resonance imaging (MRI) viewer and interactive dual physician‐patient data display to facilitate communication.Interpretation: The pioneer use of PORs for each individual patient enables personalized analytics sustaining the dialog between neurologists, patients and caregivers with quantified evidence.
- Open Access
Addressing the Challenges and Barriers to the Integration of Machine Learning into Clinical Practice: An Innovative Method to Hybrid Human–Machine Intelligence
Machine learning (ML) models have proven their potential in acquiring and analyzing large amounts of data to help solve real-world, complex problems. Their use in healthcare is expected to help physicians make diagnoses, prognoses, treatment decisions, and disease outcome predictions. However, ML solutions are not currently deployed in most healthcare systems. One of the main reasons for this is the provenance, transparency, and clinical utility of the training data. Physicians reject ML solutions if they are not at least based on accurate data and do not clearly include the decision-making process used in clinical practice. In this paper, we present a hybrid human–machine intelligence method to create predictive models driven by clinical practice. We promote the use of quality-approved data and the inclusion of physician reasoning in the ML process. Instead of training the ML algorithms on the given data to create predictive models (conventional method), we propose to pre-categorize the data according to the expert physicians’ knowledge and experience. Comparing the results of the conventional method of ML learning versus the hybrid physician–algorithm method showed that the models based on the latter can perform better. Physicians’ engagement is the most promising condition for the safe and innovative use of ML in healthcare.
Open AccessAddressing the Challenges and Barriers to the Integration of Machine Learning into Clinical Practice: An Innovative Method to Hybrid Human–Machine Intelligence
Machine learning (ML) models have proven their potential in acquiring and analyzing large amounts of data to help solve real-world, complex problems. Their use in healthcare is expected to help physicians make diagnoses, prognoses, treatment decisions, and disease outcome predictions. However, ML solutions are not currently deployed in most healthcare systems. One of the main reasons for this is the provenance, transparency, and clinical utility of the training data. Physicians reject ML solutions if they are not at least based on accurate data and do not clearly include the decision-making process used in clinical practice. In this paper, we present a hybrid human–machine intelligence method to create predictive models driven by clinical practice. We promote the use of quality-approved data and the inclusion of physician reasoning in the ML process. Instead of training the ML algorithms on the given data to create predictive models (conventional method), we propose to pre-categorize the data according to the expert physicians’ knowledge and experience. Comparing the results of the conventional method of ML learning versus the hybrid physician–algorithm method showed that the models based on the latter can perform better. Physicians’ engagement is the most promising condition for the safe and innovative use of ML in healthcare.