Analgesic Sounds and Artificial Intelligence in Patient Care

By Adaora M. Chima, MBBS, MPH, from the IARS, AUA and SOCCA 2019 Annual Meetings*

Isabelle Peretz, PhD, a professor of psychology at University of Montreal, and expert in music cognition, and Phillipe Jouvet, MD, PhD, MBA, a well-published physician scientist with a focus on artificial intelligence in clinical care, kicked off the first panel of the AUA 2019 Annual Meeting.

The Analgesic Sounds of Music

Dr. Peretz shared her work on music as a potential analgesic tool. Music, a universal phenomenon that cuts across geographic, cultural and time boundaries, has been in existence for >35,000 years. Surveys show that ~78% of the general population listen to music at least once a day and it is well known to enhance emotional regulation. What role does it play in pain regulation though?

Music has been found to modulate defensive reflexes such as the startle reflex, a defensive reaction to sudden and brief stimuli, which can be measured by recording eye blinks to brief bursts of white noise.

Participants in the studies reviewed showed an increase in amplitude and decrease in latency of the startle reflex when exposed to unpleasant music, compared to pleasant music. This encouraging result has led to further studies with thermal stimulation that showed reduced pain sensation associated with exposure to pleasant music; temperature sensation was unaffected. A study of the association of music with nociceptive flexion reflex showed amplification of pain by unpleasant music, pleasant music did not appear to mitigate pain though. Neuroimaging has revealed that pleasurable sounds activate the mesolimbic system while unpleasant sounds activate the aversive learning circuitry in the amygdala. Music activates the ventral striatum, causing dopamine release like other euphoria inducing stimuli such as food, sex, etc.

As the perception of pleasurable music is subjective, similar studies have been performed with music chosen and/or paid for by the subjects, with similar results. A recently published double-blind study attempted to pharmacologically demonstrate this dopaminergic effect by showing opposite effects of levodopa and risperidone in participants exposed to music. In essence, music is a nonpharmacological pain modulating tool that can be used to modulate pain. It provides a safe, non-invasive, low-cost intervention that can be included in the anesthesiologist’s armamentarium.

Artificial Intelligence (AI) and its Application in Clinical Care

Phillipe Jouvet, MD, PhD, MBA, a well published physician scientist, provided an overview of his application of artificial intelligence in creating clinical decision support systems for critically ill pediatric patients.

In an era of big data, evolving clinical guidelines, a plethora of publications and monitoring systems, clinical decision making can be challenging. The burgeoning cost of healthcare introduces more complexity to decision making which sometimes has to occur in the span of seconds. In this environment, human brain limitations, time constraints and fatigue can lead to errors, missed information and poor decisions.

Artificial intelligence can be used to create computerized clinical decision support systems that can address these limitations by optimizing available data resources and providing support 24/7. Computerized clinical decision systems (CCDS) involve collecting patient data, and developing an algorithm that can provide recommendations based on the information provided. Creating a robust CCDS requires digital data (EMR), evidence-based data that has been validated, a robust cyber infrastructure and a data valorization team.

Perpetual data consists of data collected over the course of a patient’s care, incorporating the patient’s clinical data, clinical interventions that have been implemented and the subsequent clinical response, telling a story that can be used in a predictive algorithm model for decision making.

Challenges in the application of this tool include problems with data integration/harmony, data corruption such as erroneous/missing data, and potential complexity of the tool, making it difficult to interpret for use.

Dr. Jouvet shared multiple examples of trail models of CCDS at his institution, CHU Sainte-Justine, including diagnostic models for chest ray analysis, cardio respiratory models for predicting the impact of ventilatory changes on blood gas, critical event evaluation and prediction such as recognition of life-threatening conditions during medical transport, and human resource programs for staff allocation in hospital units. Validation of these models are ongoing for the application of AI in clinical scenarios.

Artificial Intelligence in the form of decision-making programs can be applied as an educational tool with the use of simulation models based on real patient clinical data. It assists the clinical provider in optimizing available data resources into useful information that be utilized in practice. It requires dedicated infrastructure and a cohesive multidisciplinary team to function effectively and efficiently.

*Coverage of the AUA Host Program Panel I from the AUA 2019 Annual Meeting