Automated Speech Analysis Correlates With Hepatic Encephalopathy Tests

TUESDAY, Oct. 8, 2024 (HealthDay News) -- Automated speech analysis correlates with validated hepatic encephalopathy (HE) tests and may predict future overt HE, according to a study published online Sept. 12 in Hepatology.

In a two-center prospective cohort study, Patricia P. Bloom, M.D., from the University of Michigan in Ann Arbor, and colleagues examined whether analysis of speech patterns using a modern speech platform correlates with validated HE tests, correlates with minimal HE, and can predict future overt HE. Data were included for 200 outpatients with cirrhosis and 50 control patients; patients underwent baseline speech recording and validated HE diagnostic testing with a psychometric HE score (PHES). Patients were followed for six months.

An automated speech analysis platform was used to extract 752 speech variables. The researchers observed significant associations for more than 100 speech variables with PHES. For predicting minimal HE, a three-variable model (two acoustic and one speech tempo variable) was similar to the animal naming test (area under the curve [AUC], 0.76 versus 0.69). The accuracy of the speech model was improved by adding age and Model for End-Stage Liver Disease-Sodium score (AUC, 0.82). Time to overt HE was predicted by a combined clinical speech model, with a concordance of 0.74.

"Future research will be necessary to validate this model in larger studies in different clinical settings and to determine whether changes in speech during follow-up can detect early overt HE allowing for timely interventions to prevent hospital admissions," the authors write.

Several authors disclosed ties to the biopharmaceutical industry.

Abstract/Full Text (subscription or payment may be required)