AUTOMATED speech analysis accurately distinguished clinical, anatomical, and pathological variants of primary progressive aphasia (PPA), according to a cross-sectional study of 214 participants.
Researchers evaluated speech recordings collected during a brief picture description task and automatically extracted 40 linguistic and acoustic features.
The study included 43 cognitively healthy controls, 50 individuals with non-fluent primary progressive aphasia, 56 with logopenic PPA, and 65 with semantic PPA.
Analysis identified 25 speech features that differed between at least two PPA variants.
Researchers generated three variant specific speech profiles based on four to eight selected features per subtype. Classification performance was strong and achieved external validation in an independent cohort.
The resulting profiles reflected recognised clinical characteristics. Non-fluent PPA was associated with features linked to reduced elaboration and altered loudness dynamics.
Logopenic PPA showed patterns consistent with lexical and phonological retrieval difficulties, while semantic PPA demonstrated features suggestive of preserved fluency but reduced semantic specificity.
The speech profiles also aligned with established neuro-anatomical patterns in 195 participants who underwent magnetic resonance imaging.
Higher non-fluent PPA profile scores were associated with lower grey matter volume in the left superior and middle frontal gyri and premotor regions.
Logopenic PPA profile scores corresponded to atrophy in the left posterior temporal lobe and angular gyrus, while semantic PPA scores were linked to reduced brain volume in the bilateral anterior temporal lobes, particularly on the left.
These findings suggest that the automatically derived speech profiles capture biologically meaningful disease characteristics.
Among 56 participants with autopsy confirmed pathology representing the most common underlying disease processes, speech profile scores discriminated neuropathology with overall classification accuracy reaching 79%.
The researchers concluded that automated speech analysis can generate concise and interpretable speech profiles capable of distinguishing PPA subtypes across clinical, anatomical, and pathological domains.
They suggested that these tools could support differential diagnosis and longitudinal monitoring, particularly in healthcare settings where specialist speech and language assessment is not readily available.
Reference
Vonk JMJ et al. Automated Speech Analysis to Identify Clinical, Anatomical, and Pathological Variants of Primary Progressive Aphasia. JAMA Neurol. 2026;DOI:10.1001/jamaneurol.2026.2520
Featured image: Alina on Adobe Stock
Each article is made available under the terms of the Creative Commons Attribution-Non Commercial 4.0 License.
