Continuous Pain Intensity Estimation from Autonomic Signals with Recurrent Neural Networks
Date:
I delivered an oral presentation at the Engineering in Medicine and Biology Conference (EMBC), describing our work on continuously estimating experimental heat pain intensity from autonomic physiological signals. The project sought to develop an objective pain monitoring method that provides high temporal resolution estimates using data that can be collected noninvasively from wearable sensors.
The presentation summarized the full methodological pipeline described in our paper, including skin conductance deconvolution to recover sudomotor nerve activity and point process modeling of heart rate variability. We extracted time resolved features from overlapping windows of the deconvolved skin conductance and HRV signals and used these features as input to recurrent neural network regression models. Two architectures were evaluated: a simple recurrent network and a long short term memory (LSTM) network, both trained to estimate pain intensity on a continuous scale.
I presented results on the BioVid Heat Pain Database showing that skin conductance features significantly outperformed HRV features for both binary pain detection and continuous pain estimation. The LSTM based regression model achieved the best overall performance across mean absolute error, root mean square error, and coefficient of determination metrics, and produced temporally smooth estimates that closely followed the time varying pattern of the stimuli.
The slides from the presentation are available here: Slides.