Predicting Acute Pain Levels Implicitly from Vocal Features

Evaluating pain in speech represents a critical challenge in high-<br/>stakes clinical scenarios, from analgesia delivery to emergency<br/>triage. Clinicians have predominantly relied on direct verbal<br/>communication of pain which is difficult for patients with com-<br/>munication barriers, such as those affected by stroke, autism,<br/>and learning difficulties. Many previous efforts have focused<br/>on multimodal data which does not suit all clinical applications.<br/>Our work is the first to collect a new English speech dataset<br/>wherein we have induced acute pain in adults using a cold pres-<br/>sor task protocol and recorded subjects reading sentences out<br/>loud. We report pain discrimination performance as F1 scores<br/>from binary (pain vs. no pain) and three-class (mild, moder-<br/>ate, severe) prediction tasks, and support our results with ex-<br/>plainable feature analysis. Our work is a step towards provid-<br/>ing medical decision support for pain evaluation from speech to<br/>improve care across diverse and remote healthcare settings