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

cs.CVApr 13, 2026

A Lightweight Transformer for Pain Recognition from Brain Activity

Authors: Stefanos Gkikas, Christian Arzate Cruz, Yu Fang, Lu Cao, Muhammad Umar Khan, Thomas Kassiotis, Giorgos Giannakakis, Raul Fernandez Rojas,

Organizations: Honda Research Institute Japan, Wako City, Japan · BioSIS (Biosensing & Intelligent Systems) Lab, Centre for Intelligent Computing and Systems, University of Canberra, Canberra, Australia · Department of Electronic Engineering, Hellenic Mediterranean University, Chania, Greece

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.

Figures & tables

Fig. 1 : Overview of the proposed tokenization framework, using the two fNIRS representations.

Hyperparameter	Value
Depth ( $L$ )	4
Number of latents ( $M$ )	32
Latent dimension ( $d$ )	128
Cross-attention heads	1
Latent self-attention heads	8
Cross-attention head dimension	64

TABLE I : Architectural details.

Optimizer	LR	LR decay	Weight decay	Total epochs	Warmup epochs	Cooldown epochs	Batch size
AdamW	2e-5	cosine	0.1	200	20	10	32

TABLE II : Training configuration.

Input	#Segm.	Computational Cost		Inference Cost				Performance
		Params(M)	GFLOPs	Latency (ms) GPU $\downarrow$	Samples/s GPU $\uparrow$	Latency (ms) CPU $\downarrow$	Samples/s CPU $\uparrow$	Accuracy	Precision	F1
fNIRS-Wave	–	7.81	2.58	14.26	70.14	22.44	44.56	43.94	42.94	39.34
fNIRS-PSD	–	15.05	15.07	14.36	69.64	56.14	17.81	49.26	47.91	39.90
fNIRS-Wave	2	7.81	0.25	13.33	74.99	11.33	88.27	46.06	42.63	43.20
fNIRS-PSD	2	15.05	12.74	13.40	74.62	51.03	19.59	52.31	46.66	47.76
fNIRS-Wave	4	7.81	0.41	14.31	69.90	11.80	84.75	46.99	44.32	43.59

TABLE III : performance and computational/inference cost of fNIRS representations across latent segmentation settings.

Fig. 2 : Effect of latent segmentation on fNIRS accuracy (averaged across waveform and spectrogram representations) versus computational and inference cost.

Modality	Fusion	Computational Cost		Inference Cost				Performance
		Params(M)	GFLOPs	Latency (ms) GPU $\downarrow$	Samples/s GPU $\uparrow$	Latency (ms) CPU $\downarrow$	Samples/s CPU $\uparrow$	Accuracy	Precision	F1
fNIRS-Wave	stack	15.15	15.20	13.79	72.54	49.00	20.41	48.52	45.62	44.62
fNIRS-PSD

TABLE IV : performance and computational/inference cost for the representation fusion.

Study	Modality		Method		Accuracy
	Video	fNIRS	Features	Model
	—	✓			43.30
	✓	—			40.10
[ 49 ]	✓	✓	Handcrafted	SVM	41.70
[ 51 ]	✓	✓	Deep	Transformer	46.67
[ 52 ]	✓	—	Deep	2D CNN	49.00

TABLE V : Comparison of studies on the AI4Pain testing set.

Transformer

Transformers are neural architectures that use attention to model relationships among sequence elements in language, vision, and scientific data.

See all 540 papers

A Unified Tokenization Framework for Pain Recognition using Heterogeneous 3D Modalities
Jul 22, 2026Stefanos Gkikas, Christian Arzate Cruz, Valentina Becchetti +3Multimodal Learning
Honda Research Institute Japan Wako City, Japan · Sapienza University of Rome Rome, Italy · University of Canberra Canberra, Australia
An Exploratory Analysis of Pain Localization via Explainable Computational Modeling
Jul 22, 2026Ioannis Kyprakis, Stefanos Gkikas, Eric Nichols +2Multimodal Classification
Hellenic Mediterranean University Heraklion, Greece · Honda Research Institute Japan Wako City, Japan
ReFace: Reorganizing Facial Spatiotemporal Representations for Improved Pain Assessment
Jul 22, 2026Stefanos Gkikas, Yu Fang, Christian Arzate Cruz +2Visual Tokenization
Honda Research Institute Japan Wako City, Japan · BioSIS (Biosensing & Intelligent Systems) Lab Centre for Intelligent Computing and Systems University of Canberra Canberra, Australia

Jul 22, 2026J/K move · Enter open · S save

A Unified Tokenization Framework for Pain Recognition using Heterogeneous 3D Modalities

Stefanos Gkikas, Christian Arzate Cruz, Valentina Becchetti, Muhammad Umar Khan+2

Honda Research Institute Japan Wako City, Japan · Sapienza University of Rome Rome, Italy · University of Canberra Canberra, Australia

Multimodal Learning

Open paper AskPDF arXiv

Pain is a complex and pervasive phenomenon affecting a large percentage of the population, and accurate assessment is essential for effective clinical management and intervention. Computational pain recognition systems enable continuous monitoring, support clinical decision-making, and help mitigate pain-related distress and functional decline. This study introduces a unified tokenization framework for heterogeneous 3D modalities in pain recognition that provides a single processing pipeline across behavioral and brain-activity 3D data, without requiring separate architectures for each modality or handcrafted inductive biases. The framework preserves spatial, temporal, and time--frequency structure while mapping diverse inputs into a shared token space. Extensive experiments show that the proposed approach effectively processes facial videos and fNIRS data in both raw-signal and spectrogram-based representations. On the AI4Pain benchmark dataset, the proposed framework achieves state-of-the-art performance while maintaining high computational efficiency and enabling real-time assessment on both GPU and CPU hardware.

Abstract

Transformer

Explore similar work

A Unified Tokenization Framework for Pain Recognition using Heterogeneous 3D Modalities

An Exploratory Analysis of Pain Localization via Explainable Computational Modeling

ReFace: Reorganizing Facial Spatiotemporal Representations for Improved Pain Assessment