Detection and classification of connectional silences in palliative care conversations using deep learning

Presenter's Name(s)

Jeremy E. MattFollow

Conference Year

January 2020

Abstract

Improving clinical conversations in serious illness is a priority for twenty-first century healthcare. Successful conversations between patients, families, and clinicians leave patients and families feeling heard and understood. However, these conversations are complex and dynamic. One potential feature of a successful palliative care conversation is the presence of “connectional silences” where a connection is made between the clinician and the patient and family. To help evaluate these complex interactions, we investigate the use of Deep-learning learning networks (DLNNs) to help automate the measurement of these connectional silences for use in large sample epidemiological studies. Specifically, we investigate whether DLNN can distinguish images (i.e., Gammatonegrams) of Connectional Silences from those representing other types of Conversational Pauses.

Primary Faculty Mentor Name

Dr. Donna Rizzo

Faculty/Staff Collaborators

Dr. Donna Rizzo (Advisor), Dr. Bob Gramling (PI), Dr. Safwan Wshah (Advisor), Larry Clarfeld (Collaborator), Brigitte Durieux (Collaborator), Cailin Gramling (Collaborator), Laura Hirsch (Collaborator), Jack Straton (Collaborator), Viktoria Manukyan (Collaborator)

Status

Graduate

Student College

College of Engineering and Mathematical Sciences

Program/Major

Complex Systems

Primary Research Category

Engineering & Physical Sciences

Abstract only.

Share

COinS
 

Detection and classification of connectional silences in palliative care conversations using deep learning

Improving clinical conversations in serious illness is a priority for twenty-first century healthcare. Successful conversations between patients, families, and clinicians leave patients and families feeling heard and understood. However, these conversations are complex and dynamic. One potential feature of a successful palliative care conversation is the presence of “connectional silences” where a connection is made between the clinician and the patient and family. To help evaluate these complex interactions, we investigate the use of Deep-learning learning networks (DLNNs) to help automate the measurement of these connectional silences for use in large sample epidemiological studies. Specifically, we investigate whether DLNN can distinguish images (i.e., Gammatonegrams) of Connectional Silences from those representing other types of Conversational Pauses.