pith. sign in

arxiv: 2604.16491 · v5 · pith:DKDTDX6Wnew · submitted 2026-04-13 · 💻 cs.CV · cs.AI

A Lightweight Transformer for Pain Recognition from Brain Activity

Stefanos Gkikas , Christian Arzate Cruz , Yu Fang , Lu Cao , Muhammad Umar Khan , Thomas Kassiotis , Giorgos Giannakakis , Raul Fernandez Rojas
show 1 more author
Randy Gomez
This is my paper

Pith reviewed 2026-05-21 00:11 UTC · model grok-4.3

classification 💻 cs.CV cs.AI
keywords fNIRSpain recognitiontransformertokenizationbrain signalsmultimodal fusionlightweight modelreal-time inference
0
0 comments X

The pith

A lightweight transformer fuses raw waveform and spectral fNIRS signals through unified tokenization for competitive pain recognition.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper sets out to show that a single compact transformer can combine different views of brain activity measured by fNIRS without separate modules for each view. By projecting both the raw time series and its frequency content into one shared token space, the model keeps spatial, temporal, and time-frequency details while avoiding extra complexity. If this holds, automated pain assessment becomes feasible on modest hardware and closer to real-time clinical use. The evaluation on the AI4Pain dataset reports performance on par with heavier approaches while staying computationally light for both GPU and CPU.

Core claim

The architecture projects stacked raw waveform and power spectral density representations of fNIRS inputs onto a shared latent space using a structured segmentation scheme; this token-mixing strategy lets the transformer model complementary signal characteristics jointly without modality-specific adaptations, yielding competitive pain recognition accuracy at low computational cost.

What carries the argument

The unified tokenization mechanism that projects heterogeneous fNIRS inputs onto a shared latent representation while controlling local aggregation and global interaction through structured segmentation.

If this is right

  • The model supports real-time inference on standard CPU hardware without custom accelerators.
  • Spatial, temporal, and time-frequency signal properties are preserved in a single forward pass.
  • No separate branches or adapters are required when adding new fNIRS representations.
  • Computational footprint remains small enough for portable monitoring devices.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same token-mixing approach could apply to other paired biosignals such as EEG and ECG without redesigning the backbone.
  • If the shared latent space generalizes, fewer labeled samples might be needed for new pain-related tasks.
  • Portable headsets could run the model locally, enabling continuous home-based pain tracking.

Load-bearing premise

That stacking raw waveforms with power spectral density and projecting them into one token space is enough to capture pain patterns without extra preprocessing or testing on other datasets.

What would settle it

Retraining the same architecture on a second fNIRS pain dataset with different preprocessing shows a clear drop in accuracy relative to modality-specific baselines.

Figures

Figures reproduced from arXiv: 2604.16491 by Christian Arzate Cruz, Giorgos Giannakakis, Lu Cao, Muhammad Umar Khan, Randy Gomez, Raul Fernandez Rojas, Stefanos Gkikas, Thomas Kassiotis, Yu Fang.

Figure 1
Figure 1. Figure 1: Overview of the proposed tokenization framework, using the two fNIRS representations. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Effect of latent segmentation on fNIRS accuracy [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
read the original abstract

Pain is a multifaceted and widespread phenomenon with substantial clinical and societal burden, making reliable automated assessment a critical objective. This paper presents a lightweight transformer architecture that fuses multiple fNIRS representations through a unified tokenization mechanism, enabling joint modeling of complementary signal views without requiring modality-specific adaptations or increasing architectural complexity. The proposed token-mixing strategy preserves spatial, temporal, and time-frequency characteristics by projecting heterogeneous inputs onto a shared latent representation, using a structured segmentation scheme to control the granularity of local aggregation and global interaction. The model is evaluated on the AI4Pain dataset using stacked raw waveform and power spectral density representations of fNIRS inputs. Experimental results demonstrate competitive pain recognition performance while remaining computationally compact, making the approach suitable for real-time inference on both GPU and CPU hardware.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 1 minor

Summary. The manuscript presents a lightweight transformer architecture for pain recognition from fNIRS brain signals. It introduces a unified tokenization mechanism that fuses stacked raw waveform and power spectral density (PSD) representations by projecting them onto a shared latent space via structured segmentation, enabling joint modeling of spatial, temporal, and time-frequency characteristics without modality-specific modules or added complexity. The model is evaluated on the AI4Pain dataset and claims competitive performance with computational compactness suitable for real-time GPU/CPU inference.

Significance. If the performance claims hold under rigorous validation, the work could contribute a compact architecture for multi-view fNIRS fusion that preserves complementary signal properties through token mixing. This design choice addresses a practical need for efficient real-time pain assessment tools, and the emphasis on controlling granularity of local aggregation and global interaction is a positive aspect of the tokenization strategy.

major comments (3)
  1. [Abstract and Experimental Results] Abstract and Experimental Results section: The central claim of 'competitive pain recognition performance' is unsupported by any numeric metrics, baseline comparisons, statistical tests, data split details, or error bars. Without these, the assertion that the unified token-mixing strategy achieves the stated results cannot be verified or compared to alternatives.
  2. [Method and Evaluation] Method and Evaluation sections: No ablation studies isolate the contribution of the token-mixing strategy (e.g., versus direct concatenation of raw waveform and PSD inputs or separate modality encoders). This omission is load-bearing because the paper's core argument rests on the fusion benefit without increased complexity, yet high inter-subject variability in fNIRS signals makes it impossible to attribute gains specifically to the unified tokenization.
  3. [Dataset and Splits] Dataset and Splits subsection: Evaluation is confined to the single AI4Pain dataset without cross-dataset testing or explicit subject-independent splits. Given documented fNIRS variability across subjects and sessions, this limits the ability to substantiate generalizability of the joint-modeling approach.
minor comments (1)
  1. [Abstract] The abstract could more precisely define 'tokenization granularity and segmentation parameters' to aid reproducibility.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We sincerely thank the referee for the thorough and constructive feedback. We have revised the manuscript to strengthen the experimental validation and address concerns about metrics, ablations, and evaluation scope. Point-by-point responses follow.

read point-by-point responses

  1. Referee: [Abstract and Experimental Results] Abstract and Experimental Results section: The central claim of 'competitive pain recognition performance' is unsupported by any numeric metrics, baseline comparisons, statistical tests, data split details, or error bars. Without these, the assertion that the unified token-mixing strategy achieves the stated results cannot be verified or compared to alternatives.

    Authors: We agree that explicit quantitative support is necessary. The revised manuscript now includes specific metrics (accuracy, F1-score), baseline comparisons (CNNs, standard transformers), subject-independent split details, statistical significance tests, and error bars from repeated runs to substantiate the competitive performance claim. revision: yes

  2. Referee: [Method and Evaluation] Method and Evaluation sections: No ablation studies isolate the contribution of the token-mixing strategy (e.g., versus direct concatenation of raw waveform and PSD inputs or separate modality encoders). This omission is load-bearing because the paper's core argument rests on the fusion benefit without increased complexity, yet high inter-subject variability in fNIRS signals makes it impossible to attribute gains specifically to the unified tokenization.

    Authors: We have added ablation studies in the revision comparing unified token-mixing to direct concatenation of raw and PSD inputs and to separate modality encoders. Results confirm performance benefits from the unified approach with no added complexity, evaluated under subject-independent splits to account for inter-subject variability. revision: yes

  3. Referee: [Dataset and Splits] Dataset and Splits subsection: Evaluation is confined to the single AI4Pain dataset without cross-dataset testing or explicit subject-independent splits. Given documented fNIRS variability across subjects and sessions, this limits the ability to substantiate generalizability of the joint-modeling approach.

    Authors: We have clarified that subject-independent splits are used and added discussion of this limitation with suggestions for future multi-dataset work. Cross-dataset experiments are not feasible without additional public datasets matching the task. revision: partial

Circularity Check

0 steps flagged

No circularity: new architecture with empirical evaluation on external dataset

full rationale

The paper proposes a lightweight transformer using unified tokenization to fuse raw waveform and PSD fNIRS representations, then reports competitive results on the AI4Pain dataset. No load-bearing derivation reduces to self-definition, fitted parameters renamed as predictions, or self-citation chains; the token-mixing strategy is presented as a novel architectural choice evaluated externally rather than forced by construction from the inputs. The derivation chain is self-contained as a standard empirical ML contribution without the enumerated circular patterns.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard assumptions about transformer attention and the representativeness of the AI4Pain dataset; no new physical entities are postulated.

free parameters (1)
  • tokenization granularity and segmentation parameters
    Chosen to control local aggregation and global interaction; values are not stated in the abstract.
axioms (1)
  • domain assumption Heterogeneous fNIRS representations can be projected onto a shared latent space without loss of complementary information
    Invoked when describing the unified tokenization mechanism.

pith-pipeline@v0.9.0 · 5681 in / 1258 out tokens · 54432 ms · 2026-05-21T00:11:32.326357+00:00 · methodology

Review history (3 revisions) →

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

67 extracted references · 67 canonical work pages

  1. [1]

    Painful truth: The need to re-center chronic pain on the functional role of pain,

    V . Santiago, “Painful truth: The need to re-center chronic pain on the functional role of pain,”Journal of Pain Research, vol. 15, pp. 497–512, 2022

    work page 2022

  2. [2]

    Marchand,The pain phenomenon

    S. Marchand,The pain phenomenon. Springer Nature, 2024, vol. 13

    work page 2024

  3. [3]

    Pain: The silent public health epidemic,

    J. G. Katzman and R. M. Gallagher, “Pain: The silent public health epidemic,”Journal of Primary Care & Community Health, vol. 15, p. 21501319241253547, 2024

    work page 2024

  4. [4]

    Prescription opioid abuse in chronic pain: an updated review of opioid abuse predictors and strategies to curb opioid abuse: part 1,

    A. D. Kaye, M. R. Jones, A. M. Kaye, J. G. Ripoll, V . Galan, B. D. Beakley, F. Calixto, J. L. Bolden, R. D. Urman, and L. Manchikanti, “Prescription opioid abuse in chronic pain: an updated review of opioid abuse predictors and strategies to curb opioid abuse: part 1,”Pain physician, vol. 20, no. 2, p. S93, 2017

    work page 2017

  5. [5]

    The first 24 h: opioid administration in people with spinal cord injury and neurologic recovery,

    A. Stampas, C. Pedroza, J. N. Bush, A. R. Ferguson, J. L. K. Kramer, and M. Hook, “The first 24 h: opioid administration in people with spinal cord injury and neurologic recovery,”Spinal Cord, vol. 58, no. 10, pp. 1080–1089, 2020

    work page 2020

  6. [6]

    Opioid complications and side effects,

    R. Benyamin, A. M. Trescot, S. Datta, R. M. Buenaventura, R. Adlaka, N. Sehgal, S. E. Glaser, and R. Vallejo, “Opioid complications and side effects,”Pain physician, vol. 11, no. 2S, p. S105, 2008

    work page 2008

  7. [7]

    Analgesic administration, pain intensity, and patient satisfaction in cardiac surgical patients,

    D. Meehan, M. McRae, D. Rourke, C. Eisenring, and F. Imperial, “Analgesic administration, pain intensity, and patient satisfaction in cardiac surgical patients,”American Journal of Critical Care, vol. 4, no. 6, pp. 435–442, 1995

    work page 1995

  8. [8]

    Use of a pain assessment and intervention notation (p.a.i.n.) tool in critical care nursing practice: Nurses’ evaluations,

    K. A. Puntillo, D. Stannard, C. Miaskowski, K. Kehrle, and S. Gleeson, “Use of a pain assessment and intervention notation (p.a.i.n.) tool in critical care nursing practice: Nurses’ evaluations,”Heart & Lung, vol. 31, no. 4, pp. 303–314, 2002

    work page 2002

  9. [9]

    Update on prevalence of pain in patients with cancer 2022: A systematic literature review and meta-analysis,

    R. A. H. Snijders, L. Brom, M. Theunissen, and M. H. J. van den Beuken-van Everdingen, “Update on prevalence of pain in patients with cancer 2022: A systematic literature review and meta-analysis,”Cancers, vol. 15, no. 3, 2023

    work page 2022

  10. [10]

    Cognition and pain: A review,

    T. Khera and V . Rangasamy, “Cognition and pain: A review,”Frontiers in Psychology, vol. V olume 12 - 2021, 2021

    work page 2021

  11. [11]

    Pain in focus: How persistent pain disrupts the attentional bias towards pain- related information,

    J. Li, X. Lyu, X. Li, X. Yang, L. Weng, Y . Wang, and W. Peng, “Pain in focus: How persistent pain disrupts the attentional bias towards pain- related information,”NeuroImage, vol. 321, p. 121539, 2025

    work page 2025

  12. [12]

    The effect of induced and chronic pain on attention,

    D. J. Moore, S. M. Meints, A. Lazaridou, D. Johnson, O. Franceschelli, M. Cornelius, K. Schreiber, and R. R. Edwards, “The effect of induced and chronic pain on attention,”The Journal of Pain, vol. 20, no. 11, pp. 1353–1361, 2019

    work page 2019

  13. [13]

    A systematic review of neurophysiological sensing for the assessment of acute pain,

    R. Fernandez Rojas, N. Brown, G. Waddington, and R. Goecke, “A systematic review of neurophysiological sensing for the assessment of acute pain,”NPJ Digital Medicine, vol. 6, no. 1, p. 76, 2023

    work page 2023

  14. [14]

    Automatic assessment of pain based on deep learning methods: A systematic review,

    S. Gkikas and M. Tsiknakis, “Automatic assessment of pain based on deep learning methods: A systematic review,”Computer Methods and Programs in Biomedicine, vol. 231, p. 107365, 2023

    work page 2023

  15. [15]

    Twifly: A data analysis framework for twitter,

    P. Chatziadam, A. Dimitriadis, S. Gikas, I. Logothetis, M. Michalodim- itrakis, M. Neratzoulakis, A. Papadakis, V . Kontoulis, N. Siganos, D. Theodoropoulos, G. V ougioukalos, I. Hatzakis, G. Gerakis, N. Pa- padakis, and H. Kondylakis, “Twifly: A data analysis framework for twitter,”Information, vol. 11, no. 5, 2020

    work page 2020

  16. [16]

    Region of interest detection and evaluation in functional near infrared spectroscopy,

    R. F. Rojas, X. Huang, and K.-L. Ou, “Region of interest detection and evaluation in functional near infrared spectroscopy,”Journal of Near Infrared Spectroscopy, vol. 24, no. 4, pp. 317–326, 2016

    work page 2016

  17. [17]

    1bt: One-block transformer for eeg-based cognitive workload assessment,

    S. Gkikas, C. A. Cruz, T. Kassiotis, G. Giannakakis, R. F. Rojas, and R. Gomez, “1bt: One-block transformer for eeg-based cognitive workload assessment,” 2026

    work page 2026

  18. [18]

    Cortical network response to acupuncture and the effect of the hegu point: An fnirs study,

    R. Fernandez Rojas, M. Liao, J. Romero, X. Huang, and K.-L. Ou, “Cortical network response to acupuncture and the effect of the hegu point: An fnirs study,”Sensors, vol. 19, no. 2, 2019. TABLE III: performance and computational/inference cost of fNIRS representations across latent segmentation settings. Input #Segm. Computational Cost Inference Cost Perfo...

    work page arXiv 2019

  19. [19]

    —✓ Handcrafted ENS 53.66 [55]✓— Deep Transformer 55.00 —✓52.60 ✓— 53.67 [56] ✓ ✓ Deep Transformer 55.69 Our — ✓ Deep Transformer 50.00† ✓indicates the modality is used – indicates not used ENS: Ensemble Classifier †: Wave & PSD stack fusion

    work page

  20. [20]

    A pain assessment framework based on multimodal data and deep machine learning methods,

    S. Gkikas, “A pain assessment framework based on multimodal data and deep machine learning methods,” 2025, arXiv preprint arXiv:2505.05396. [Online]. Available: https://arxiv.org/abs/2505.05396

    work page arXiv 2025

  21. [21]

    Spatio-temporal pain estimation network with measuring pseudo heart rate gain,

    D. Huang, X. Feng, H. Zhang, Z. Yu, J. Peng, G. Zhao, and Z. Xia, “Spatio-temporal pain estimation network with measuring pseudo heart rate gain,”IEEE Transactions on Multimedia, vol. 24, pp. 3300–3313, 2022

    work page 2022

  22. [22]

    Acute pain recognition from facial expression videos using vision transformers,

    G. Bargshady, C. Joseph, N. Hirachan, R. Goecke, and R. F. Rojas, “Acute pain recognition from facial expression videos using vision transformers,” in2024 46th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2024, pp. 1–4

    work page 2024

  23. [23]

    A full transformer-based framework for automatic pain estimation using videos,

    S. Gkikas and M. Tsiknakis, “A full transformer-based framework for automatic pain estimation using videos,” in2023 45th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), 2023, pp. 1–6

    work page 2023

  24. [24]

    Synthetic thermal and rgb videos for automatic pain assessment utilizing a vision-mlp architecture,

    ——, “Synthetic thermal and rgb videos for automatic pain assessment utilizing a vision-mlp architecture,” in2024 12th International Conference on Affective Computing and Intelligent Interaction Workshops and Demos (ACIIW), 2024, pp. 4–12

    work page 2024

  25. [25]

    Automatic pain intensity estimation based on electrocardiogram and demographic factors

    S. Gkikas., C. Chatzaki., E. Pavlidou., F. Verigou., K. Kalkanis., and M. Tsiknakis., “Automatic pain intensity estimation based on electrocardiogram and demographic factors.” SciTePress, 2022, pp. 155–162

    work page 2022

  26. [26]

    Multi-task neural networks for pain intensity estimation using electrocardiogram and demographic factors,

    S. Gkikas, C. Chatzaki, and M. Tsiknakis, “Multi-task neural networks for pain intensity estimation using electrocardiogram and demographic factors,” inInformation and Communication Technologies for Ageing Well and e-Health. Springer Nature Switzerland, 2023, pp. 324–337

    work page 2023

  27. [27]

    Exploring deep physiological models for nociceptive pain recognition,

    P. Thiam, P. Bellmann, H. A. Kestler, and F. Schwenker, “Exploring deep physiological models for nociceptive pain recognition,”Sensors, vol. 19, p. 4503, 10 2019

    work page 2019

  28. [28]

    Ensemble neural networks for multimodal acute pain intensity evaluation using video and physiological signals,

    M. S. Patil and H. D. Patil, “Ensemble neural networks for multimodal acute pain intensity evaluation using video and physiological signals,” Journal of Computational Analysis and Applications (JoCAAA), vol. 33, no. 05, p. 779–791, Sep. 2024

    work page 2024

  29. [29]

    Multimodal pain assessment based on physiological biosignals: The impact of demographic factors on perception and sensitivity,

    E. Pavlidou and M. Tsiknakis, “Multimodal pain assessment based on physiological biosignals: The impact of demographic factors on perception and sensitivity,” inProceedings of the 11th International Conference on Information and Communication Technologies for Ageing Well and e-Health - ICT4AWE, INSTICC. SciTePress, 2025, pp. 320– 329

    work page 2025

  30. [30]

    Automatic pain assessment with ultra-short electrodermal activity signal,

    X. Ji, T. Zhao, W. Li, and A. Zomaya, “Automatic pain assessment with ultra-short electrodermal activity signal,” inProceedings of the 38th ACM/SIGAPP Symposium on Applied Computing, ser. SAC ’23. New York, NY , USA: Association for Computing Machinery, 2023, p. 618–625

    work page 2023

  31. [31]

    Pain recognition with physiological signals using multi-level context information,

    K. N. Phan, N. K. Iyortsuun, S. Pant, H.-J. Yang, and S.-H. Kim, “Pain recognition with physiological signals using multi-level context information,”IEEE Access, vol. 11, pp. 20 114–20 127, 2023

    work page 2023

  32. [32]

    Transformer encoder with multiscale deep learning for pain classification using physiological signals,

    Z. Lu, B. Ozek, and S. Kamarthi, “Transformer encoder with multiscale deep learning for pain classification using physiological signals,”Frontiers in Physiology, vol. 14, 2023

    work page 2023

  33. [33]

    Automatic pain assessment based on physiological signals: Application of multi-scale networks and cross-attention cross-attention,

    J. Li, J. Luo, Y . Wang, Y . Jiang, X. Chen, and Y . Quan, “Automatic pain assessment based on physiological signals: Application of multi-scale networks and cross-attention cross-attention,” inProceedings of the 2024 13th International Conference on Bioinformatics and Biomedical Science, ser. ICBBS ’24. New York, NY , USA: Association for Computing Machi...

    work page 2024

  34. [34]

    A two-stage architecture for identifying and locating the source of pain using novel multi-domain binary patterns of eda,

    S. Aziz, C. Joseph, N. Hirachan, L. Murtagh, G. Chetty, R. Goecke, and R. Fernandez-Rojas, “A two-stage architecture for identifying and locating the source of pain using novel multi-domain binary patterns of eda,”Biomedical Signal Processing and Control, vol. 104, p. 107454, 2025

    work page 2025

  35. [35]

    Multi-representation diagrams for pain recognition: Integrating various electrodermal activity signals into a single image,

    S. Gkikas, I. Kyprakis, and M. Tsiknakis, “Multi-representation diagrams for pain recognition: Integrating various electrodermal activity signals into a single image,” inCompanion Proceedings of the 27th International Conference on Multimodal Interaction, ser. ICMI Companion ’25. New York, NY , USA: Association for Computing Machinery, 2025, p. 162–171

    work page 2025

  36. [36]

    Pain assessment based on fnirs using bi-lstm rnns,

    R. F. Rojas, J. Romero, J. Lopez-Aparicio, and K.-L. Ou, “Pain assessment based on fnirs using bi-lstm rnns,” in2021 10th International IEEE/EMBS Conference on Neural Engineering (NER), 2021, pp. 399– 402

    work page 2021

  37. [37]

    Multilevel pain assessment with functional near-infrared spectroscopy: Evaluating δhbo2 and δhhb measures for comprehensive analysis,

    M. U. Khan, M. Sousani, N. Hirachan, C. Joseph, M. Ghahramani, G. Chetty, R. Goecke, and R. Fernandez-Rojas, “Multilevel pain assessment with functional near-infrared spectroscopy: Evaluating δhbo2 and δhhb measures for comprehensive analysis,”Sensors, vol. 24, no. 2, 2024

    work page 2024

  38. [38]

    Empirical comparison of deep learning models for fnirs pain decoding,

    R. Fernandez Rojas, C. Joseph, G. Bargshady, and K.-L. Ou, “Empirical comparison of deep learning models for fnirs pain decoding,”Frontiers in Neuroinformatics, 2024

    work page 2024

  39. [39]

    Pain assessment using multi-kernel-fcn-lstm and haemoglobin difference in fnirs,

    G. Bargshady, S. Aziz, S. Gkikas, M. Tsiknakis, R. Goecke, and R. Fernandez Rojas, “Pain assessment using multi-kernel-fcn-lstm and haemoglobin difference in fnirs,”ACM Trans. Comput. Healthcare, 2025

    work page 2025

  40. [40]

    fnirs approach to pain assessment for non-verbal patients,

    R. F. Rojas, X. Huang, J. Romero, and K.-L. Ou, “fnirs approach to pain assessment for non-verbal patients,” inNeural Information Processing. Cham: Springer International Publishing, 2017, pp. 778–787

    work page 2017

  41. [41]

    A machine learning approach for the identification of a biomarker of human pain using fnirs,

    R. Fernandez Rojas, X. Huang, and K.-L. Ou, “A machine learning approach for the identification of a biomarker of human pain using fnirs,” Scientific reports, vol. 9, no. 1, p. 5645, 2019

    work page 2019

  42. [42]

    Empirical comparison of deep learning models for fnirs pain decoding,

    R. Fernandez Rojas, C. Joseph, G. Bargshady, and K.-L. Ou, “Empirical comparison of deep learning models for fnirs pain decoding,”Frontiers in Neuroinformatics, vol. 18, 2024

    work page 2024

  43. [43]

    A crossmod-transformer deep learning framework for multi-modal pain detection through eda and ecg fusion,

    J. Farmani, G. Bargshady, S. Gkikas, M. Tsiknakis, and R. Fernandez Ro- jas, “A crossmod-transformer deep learning framework for multi-modal pain detection through eda and ecg fusion,”Scientific Reports, vol. 15, no. 1, p. 29467, 2025

    work page 2025

  44. [44]

    Mul- timodal automatic assessment of acute pain through facial videos and heart rate signals utilizing transformer-based architectures,

    S. Gkikas, N. S. Tachos, S. Andreadis, V . C. Pezoulas, D. Zaridis, G. Gkois, A. Matonaki, T. G. Stavropoulos, and D. I. Fotiadis, “Mul- timodal automatic assessment of acute pain through facial videos and heart rate signals utilizing transformer-based architectures,”Frontiers in Pain Research, vol. 5, 2024

    work page 2024

  45. [45]

    Giaformer: A gradient-infused attention and transformer for pain assessment with eda-fnirs fusion,

    M. U. Khan, G. Chetty, S. Gkikas, M. Tsiknakis, R. Goecke, and R. Fernandez-Rojas, “Giaformer: A gradient-infused attention and transformer for pain assessment with eda-fnirs fusion,”Information Fusion, vol. 131, p. 104173, 2026

    work page 2026

  46. [46]

    Tree-based models for pain detection from biomedical signals,

    H. Shi, B. Chikhaoui, and S. Wang, “Tree-based models for pain detection from biomedical signals,” inParticipative Urban Health and Healthy Aging in the Age of AI. Springer International Publishing, 2022, pp. 183–195

    work page 2022

  47. [47]

    Discrete wavelet transform based pain assessment using multiple physiological signals,

    S. Jerritta, M. Murugappan, D. Manasa, and R. Gayathri, “Discrete wavelet transform based pain assessment using multiple physiological signals,” in2022 First International Conference on Electrical, Electronics, Information and Communication Technologies (ICEEICT), 2022, pp. 1–7

    work page 2022

  48. [48]

    A multi-modal multi-expert framework for pain assessment in postoperative children,

    Z. Liang, H. Luo, X. Chen, Z. Zhong, C. Fan, X. Song, B. Li, and J. Lv, “A multi-modal multi-expert framework for pain assessment in postoperative children,”IEEE Transactions on Affective Computing, vol. 16, no. 4, pp. 2828–2841, 2025

    work page 2025

  49. [49]

    Efficient pain recognition via respiration signals: A single cross-attention transformer multi-window fusion pipeline,

    S. Gkikas, I. Kyprakis, and M. Tsiknakis, “Efficient pain recognition via respiration signals: A single cross-attention transformer multi-window fusion pipeline,” inCompanion Proceedings of the 27th International Conference on Multimodal Interaction, ser. ICMI Companion ’25. New York, NY , USA: Association for Computing Machinery, 2025, p. 70–79

    work page 2025

  50. [50]

    The ai4pain grand challenge 2024: Advancing pain assessment with multimodal fnirs and facial video analysis,

    R. Fernandez–Rojas, C. Joseph, N. Hirachan, B. Seymour, and R. Goecke, “The ai4pain grand challenge 2024: Advancing pain assessment with multimodal fnirs and facial video analysis,” in2024 12th International Conference on Affective Computing and Intelligent Interaction Workshops and Demos (ACIIW), 2024, pp. 55–60

    work page 2024

  51. [51]

    Multimodal physiological sensing for the assessment of acute pain,

    R. Fernandez Rojas, N. Hirachan, N. Brown, G. Waddington, L. Murtagh, B. Seymour, and R. Goecke, “Multimodal physiological sensing for the assessment of acute pain,”Frontiers in Pain Research, vol. 4, 2023

    work page 2023

  52. [52]

    Twins-painvit: Towards a modality-agnostic vision transformer framework for multimodal automatic pain assessment using facial videos and fnirs,

    S. Gkikas and M. Tsiknakis, “Twins-painvit: Towards a modality-agnostic vision transformer framework for multimodal automatic pain assessment using facial videos and fnirs,” in2024 12th International Conference on Affective Computing and Intelligent Interaction Workshops and Demos (ACIIW), 2024, pp. 13–21

    work page 2024

  53. [53]

    Faces of experimental pain: Transferability of deep-learned heat pain features to electrical pain*,

    P. Prajod, D. Schiller, D. W. Don, and E. Andr ´e, “Faces of experimental pain: Transferability of deep-learned heat pain features to electrical pain*,” in2024 12th International Conference on Affective Computing and Intelligent Interaction Workshops and Demos (ACIIW), 2024, pp. 31–38

    work page 2024

  54. [54]

    Multimodal model for automated pain assessment: Leveraging video and fnirs,

    J. Vianto, A. Divakaran, H. Yang, S. Yeom, S. Kim, S. Kim, and J. Shin, “Multimodal model for automated pain assessment: Leveraging video and fnirs,”Applied Sciences, vol. 15, no. 9, 2025

    work page 2025

  55. [55]

    Empirically transformed energy patterns: A novel approach for capturing fnirs signal dynamics in pain assessment,

    M. U. Khan, S. Aziz, L. Murtagh, G. Chetty, R. Goecke, and R. Fernandez Rojas, “Empirically transformed energy patterns: A novel approach for capturing fnirs signal dynamics in pain assessment,”Computers in Biology and Medicine, vol. 192, p. 110300, 2025

    work page 2025

  56. [56]

    Transformer with leveraged masked autoencoder for video-based pain assessment,

    M.-D. Nguyen, H.-J. Yang, S.-H. Kim, J.-E. Shin, and S.-W. Kim, “Transformer with leveraged masked autoencoder for video-based pain assessment,” 2024

    work page 2024

  57. [57]

    Painformer: A vision foundation model for automatic pain assessment,

    S. Gkikas, R. F. Rojas, and M. Tsiknakis, “Painformer: A vision foundation model for automatic pain assessment,”IEEE Transactions on Affective Computing, vol. 16, no. 4, pp. 3369–3386, 2025

    work page 2025

  58. [58]

    Tiny-biomoe: a lightweight embedding model for biosignal analysis,

    S. Gkikas, I. Kyprakis, and M. Tsiknakis, “Tiny-biomoe: a lightweight embedding model for biosignal analysis,” inCompanion Proceedings of the 27th International Conference on Multimodal Interaction, ser. ICMI Companion ’25. New York, NY , USA: Association for Computing Machinery, 2025, p. 117–126

    work page 2025

  59. [59]

    A view on edge caching applications,

    D. Antonogiorgakis, A. Britzolakis, P. Chatziadam, A. Dimitriadis, S. Gikas, E. Michalodimitrakis, M. Oikonomakis, N. Siganos, E. Tza- gkarakis, Y . Nikoloudakis, S. Panagiotakis, E. Pallis, and E. K. Markakis, “A view on edge caching applications,” 2019

    work page 2019

  60. [60]

    Empatheticexchanges: Toward understanding the cues for empathy in dyadic conversations,

    E. C. Montiel-Vazquez, C. Arzate Cruz, J. A. R. Uresti, and R. Gomez, “Empatheticexchanges: Toward understanding the cues for empathy in dyadic conversations,”IEEE Access, vol. 12, pp. 195 097–195 110, 2024

    work page 2024

  61. [61]

    When and how to express empathy in human-robot interaction scenarios,

    C. Arzate Cruz, E. C. Montiel-Vazquez, C. Maeda, and R. Gomez, “When and how to express empathy in human-robot interaction scenarios,” in2025 34th IEEE International Conference on Robot and Human Interactive Communication (RO-MAN), 2025, pp. 1070–1077

    work page 2025

  62. [62]

    Empathetic robots using empathy classifiers in hri settings,

    C. Arzate Cruz, E. C. Montiel-Vazquez, C. Maeda, D. Lam, and R. Gomez, “Empathetic robots using empathy classifiers in hri settings,” in2025 20th ACM/IEEE International Conference on Human-Robot Interaction (HRI), 2025, pp. 1211–1215

    work page 2025

  63. [63]

    Data augmentation for 3dmm-based arousal-valence prediction for hri,

    C. A. Cruz, Y . Sechayk, T. Igarashi, and R. Gomez, “Data augmentation for 3dmm-based arousal-valence prediction for hri,” in2024 33rd IEEE International Conference on Robot and Human Interactive Communica- tion (ROMAN), 2024, pp. 2015–2022

    work page 2024

  64. [64]

    Enhancing social robot’s direct gaze expression through vestibulo-ocular movements,

    Y . Fang, J. M. P ´erez-Moler´on, L. Merino, S.-L. Yeh, S. Nishina, and R. Gomez, “Enhancing social robot’s direct gaze expression through vestibulo-ocular movements,”Advanced Robotics, vol. 38, no. 19-20, pp. 1457–1469, 2024

    work page 2024

  65. [65]

    Virtual reflections on a dynamic 2-d eye model improve spatial reference identification,

    M. Kruger, Y . Oshima, and Y . Fang, “Virtual reflections on a dynamic 2-d eye model improve spatial reference identification,”IEEE Transactions on Human-Machine Systems, vol. 56, no. 2, pp. 203–212, 2026

    work page 2026

  66. [66]

    A visual perceptual perspective on gaze in social robotics,

    R. S. Hessels and Y . Fang, “A visual perceptual perspective on gaze in social robotics,”Psychonomic Bulletin & Review, vol. 33, no. 4, p. 131, 2026

    work page 2026

  67. [67]

    Efficient emotion-aware iconic gesture prediction for robot co-speech,

    E. C. Montiel-Vazquez, C. A. Cruz, S. Gkikas, T. Kassiotis, G. Gi- annakakis, and R. Gomez, “Efficient emotion-aware iconic gesture prediction for robot co-speech,” 2026

    work page 2026