Research
A selection of my papers from graduate school and my postdoc. The work below spans about a decade and rests on a single methodological backbone: extracting interpretable acoustic features (timing, pitch, voice quality, articulation) and lexical features (word choice, syntactic complexity, content of what the person said) from short speech recordings, and modeling those features against clinical outcomes, structural neuroimaging measures, or, in some studies, post-mortem neuropathology. The section headings below group the papers by which patient population was the focus.
Where there's an open access PDF I've linked to it; otherwise the link goes to the journal page. The full publication list lives on Google Scholar, and my ResearchGate profile is complete through summer 2024, when I left academia. My PhD thesis is here if reading a hundred-something-page dissertation for fun happens to be your thing. I don't know who that person is, but I'm always pleased to hear from them.
Motor speech and ALS
My doctoral work was on bulbar amyotrophic lateral sclerosis, the variant of ALS in which the motor neurons controlling speech and swallowing degenerate first. Bulbar onset has historically received less attention than limb onset, partly because the early clinical signs are subtler and harder to instrument. My dissertation was an attempt to address that. The papers combine detailed acoustic and kinematic measurements of speech with structural MRI and, eventually, post-mortem neuropathology of the brains of patients who had passed away, with the goal of characterizing how the disease shows up in the speech motor system and where in the brain that breakdown actually lives.
Pause patterns by themselves can separate motor-driven speech breakdown from cognitive-driven speech breakdown. Open access.
Validated tongue and jaw movement metrics as early markers of bulbar involvement, before the disease is obvious in someone's voice.
The neuroimaging half of my thesis, mapping which regions of the speech network show structural change in bulbar ALS.
The neuropathology half of my thesis, confirming in post-mortem tissue what the imaging had suggested in living patients.
A review of which acoustic features actually hold up across studies, and which keep getting reported because they're easy to compute.
Primary progressive aphasia
Primary progressive aphasia is a neurodegenerative syndrome in which language is affected first and most. PPA comes in several variants, distinguished by which aspect of language goes first, and clinicians have a hard time differentiating them in early disease. The motor speech component is part of what makes that diagnostic call hard, because apraxia of speech can mimic or co-occur with the language deficits. These papers were on quantifying that motor speech component in PPA and tying it to specific anatomical substrates.
Quantified the motor speech component of primary progressive aphasia and mapped it to specific brain regions.
Looked at how apraxia of speech shows up in everyday connected speech, not just in the contrived tasks clinicians usually use to elicit it.
Extended the digital speech approach to two related movement disorders that also affect speech.
Frontotemporal dementia and Alzheimer's
These papers are from my postdoc at the Penn Frontotemporal Degeneration Center. The recurring question across them was whether short, naturalistic speech samples, recorded in a clinical visit with no specialized equipment, could carry useful information about underlying neurodegenerative pathology that imaging and CSF biomarkers either miss or only pick up late in the disease course. The papers below take that question into different patient populations and look for partial answers.
Speech changes are detectable in people who carry a pathogenic variant for FTD years before any clinical symptoms appear. Covered by the National Institute on Aging.
Word choice and voice quality both carry signal about Alzheimer's pathology, and they carry different signal.
Used machine learning on natural speech to classify which FTD patients also have Alzheimer's pathology underneath, a clinically hard call that imaging often gets wrong.
Brought the motor speech work from the ALS line of research and the cognitive speech work from the FTD side into one analysis, since the two diseases overlap clinically.
A cognitive and biomarker profile of patients with Lewy body dementia who also have Alzheimer's pathology, a common but underdiagnosed combination.