Faces of Experimental Pain: Transferability of Deep Learned Heat Pain Features to Electrical Pain

TL;DR

This work tackles the problem of limited pain datasets by examining whether deep features learned from heat-pain facial expressions can transfer to electrical-pain recognition. Using a BioVid heat-pain CNN as a fixed feature extractor, faces from the AI4Pain dataset are encoded into $1024$-dimensional frame features and fed into two classifiers: a Simple ANN with majority voting and an LSTM that processes 300-frame sequences. Both models outperform challenge baselines on validation, with the LSTM achieving the best validation performance and the ANN yielding strong test results, illustrating practical cross-dataset transferability of pain representations. The findings suggest that features learned from one experimental pain modality can generalize to another, offering a path to robust pain recognition under data-scarce conditions.

Abstract

The limited size of pain datasets are a challenge in developing robust deep learning models for pain recognition. Transfer learning approaches are often employed in these scenarios. In this study, we investigate whether deep learned feature representation for one type of experimentally induced pain can be transferred to another. Participating in the AI4Pain challenge, our goal is to classify three levels of pain (No-Pain, Low-Pain, High-Pain). The challenge dataset contains data collected from 65 participants undergoing varying intensities of electrical pain. We utilize the video recording from the dataset to investigate the transferability of deep learned heat pain model to electrical pain. In our proposed approach, we leverage an existing heat pain convolutional neural network (CNN) - trained on BioVid dataset - as a feature extractor. The images from the challenge dataset are inputted to the pre-trained heat pain CNN to obtain feature vectors. These feature vectors are used to train two machine learning models: a simple feed-forward neural network and a long short-term memory (LSTM) network. Our approach was tested using the dataset's predefined training, validation, and testing splits. Our models outperformed the baseline of the challenge on both the validation and tests sets, highlighting the potential of models trained on other pain datasets for reliable feature extraction.

Figures (5)

• Figure 1: An illustration of the machine learning pipeline with an overview of the various components.
• Figure 2: Few saliency maps taken from Prajod et al. prajod2022using showing the features learned by their model.
• Figure 3: Visualization of our CNN feature extractor.
• Figure 4: Visualization of our simple ANN model with majority voting scheme for video level prediction.
• Figure 5: Visualization of our LSTM model for 10-second videos (300 frames).