ZMZachary A. Miller

Zachary A. Miller

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Zachary A. Miller has a total of 862 co-authors. Here are the top co-authors by number of shared publications:

  • Bruce L. Miller

    Fein Memory and Aging Center, Department of Neurology, UCSF Weill Institute for Neurosciences University of California San Francisco California USA

    60 shared publications
  • Maria Luisa Gorno-Tempini

    Edward and Pearl Fein Memory and Aging Center, Department of Neurology, University of California San Francisco , San Francisco , CA 94158

    40 shared publications
  • Gil D. Rabinovici

    Fein Memory and Aging Center, Department of Neurology, UCSF Weill Institute for Neurosciences University of California San Francisco California USA

    34 shared publications
  • Howard J. Rosen

    Memory and Aging Center, Department of Neurology, Weill Institute for Neurosciences University of California San Francisco California USA

    30 shared publications
  • William W. Seeley

    Memory and Aging Center, Department of Neurology, Weill Institute for Neurosciences University of California San Francisco California USA

    25 shared publications
  • Joel H. Kramer

    Memory and Aging Center, Department of Neurology, Weill Institute for Neurosciences University of California San Francisco California USA

    18 shared publications

Latest Publications

  • Open Access

    Artificial intelligence-based speech error detection to differentiate primary progressive aphasia variants

    by Jet M J Vonk, Jiachen Lian, G Lynn Kurteff, Cheol J Cho, Giada Antonicelli, Zoe Ezzes, Lisa D Wauters, Willa Keegan-Rodewald, Diana Alejandra Rodriguez, Nina Dronkers, et al.

    Artificial intelligence-based approaches to speech analysis have the potential to assist with objective speech error analysis in aphasia but off-the shelf tools often fail to detect speech errors due to prioritizing ‘fluent transcription’. Speech production errors (dysfluencies) are hallmark diagnostic features of the nonfluent and logopenic variants of primary progressive aphasia, yet they can be challenging to detect and characterize even by expert clinicians. This study aimed to evaluate whether the novel automated lightweight Scalable Speech Dysfluency Modeling system, specifically designed to detect dysfluencies, could accurately distinguish primary progressive aphasia variants using voice recordings of individuals reading a brief passage. Participants included a total of 104 individuals, 40 with non-fluent primary progressive aphasia and 40 with logopenic primary progressive aphasia (matched on disease severity), and 24 healthy controls who read aloud the ‘Grandfather Passage’ as part of a widely used motor speech evaluation. We automatically extracted ten speech error (dysfluency) variables, including insertions, replacements, and deletions at both phoneme- and word-levels, and phoneme-level prolongations and repetitions. Group differences were assessed via ANOVAs controlling for age, education, and disease severity (Mini-Mental State Examination and Clinical Dementia Rating sum-of-boxes). To test clinical relevance, we performed correlation analyses with motor speech evaluation ratings provided by experienced speech-language pathologists (i.e. gold standard) within the non-fluent primary progressive aphasia group. Classification performance was assessed by training random forest and XGBoost machine-learning models including 5-fold cross-validation. All individuals read the entire passage in less than five minutes. The lightweight Scalable Speech Dysfluency Modeling system detected 8 of the 10 predefined dysfluency features at sufficient frequency to include them in subsequent analyses. All eight features distinguished primary progressive aphasia from controls (P < 0.006). Individuals with non-fluent primary progressive aphasia made more errors than logopenic primary progressive aphasia on every feature (all P < 0.023). Each feature showed a moderate positive correlation with a global motor speech evaluation apraxia/dysarthria score (r = 0.31–0.56; P < 0.001–0.053). Together, the eight features were able to classify non-fluent versus logopenic at area under the receiver operating characteristic curve = 0.798 (held-out random forest), 0.699 (held-out XGBoost), 0.714 (cross-validated random forest), and 0.704 (cross-validated XGBoost). In sum, automated speech error analysis accurately distinguished non-fluent and logopenic variants using a brief reading task. This quick error-sensitive scalable artificial intelligence system has the potential of providing a practical tool to aid diagnosis in aphasia and motor speech disorders.


  • Open Access

    Response to rethinking Alzheimer's disease susceptibility and heterogeneity. Comment on Miller et al

    by Zachary A. Miller, Bruce L. Miller, Gil D. Rabinovici, Maria Luisa Gorno‐Tempini

  • Open Access

    Automated transcription in primary progressive aphasia: Accuracy and effects on classification

    by Natasha Clarke, Brittany Morin, Christophe Bedetti, Rian Bogley, Sophie Pellerin, Bérengère Houzé, Siddarth Ramkrishnan, Zoe Ezzes, Zachary A. Miller, Maria Luisa Gorno-Tempini, et al.

    INTRODUCTION: Connected speech analyses can help characterize linguistic impairments in primary progressive aphasia (PPA) and classify variants, however, manual transcription of speech samples is time-consuming and expensive. Automated speech recognition (ASR) may be efficacious for transcribing PPA speech. METHODS: Transcripts of picture descriptions (109 PPA, 32 healthy controls (HC)) were generated using a manual, automated (Whisper) or semi-automated approach including a quality control (QC) step. We evaluated transcript accuracy, the reliability of ASR-derived linguistic features, and classification performance. RESULTS: Whisper demonstrated lowest error rates for HC, followed by semantic, logopenic and non-fluent PPA variants. Errors correlated with overall disease severity for semantic and logopenic variants. QC of Whisper outputs reduced errors and improved the reliability of linguistic features. Overall, ASR-derived features achieved better classification performance than manual transcription features. DISCUSSION: Results support the use of off-the-shelf ASR for scalable, cost-efficient transcription of PPA speech and classification.


  • Open Access

    Data-driven Cognitive Clusters in Persistent Developmental Dyslexia

    by Pedro Pinheiro-Chagas, Margo Kersey, Nilgoun Bahar, Sarah Inkelis, Julian Siebert, Jet M. J. Vonk, Rian Bogley, Ellie Carpenter, Dolce Vita Martin-Moreno, Jessica De Leon, et al.

    Developmental dyslexia is a neurodevelopmental condition characterized by persistent reading difficulties despite adequate intelligence and educational opportunities. The cognitive underpinnings of dyslexia remain incompletely understood. Although multiple causal pathways have been proposed, there is a gap in studies that can reliably identify dyslexia subtypes based on comprehensive evaluations that go beyond measures of reading and phonology. We employed one of the largest detailed samples of children with developmental dyslexia resistant to intervention, consisting of 147 extensively phenotyped school-age children aged 7-14 years who were recruited from specialized schools for students with learning differences. We implemented data-driven analyses leveraging a comprehensive neuropsychological assessment to delineate cognitive subdomains and characterize the heterogeneity in the sample. Hierarchical clustering of 24 cognitive measures aided in constructing a framework to subsequently interpret emerging clusters. Four clusters emerged, predominantly associated with: Processing Speed & Executive Function, Visuospatial & Mathematical Reasoning, Phonological Manipulation, and Verbal Short-term Memory & Word Retrieval. Subsequent latent profile analysis identified two distinct dyslexia profiles: one characterized by marked difficulties in executive functioning and processing speed, and another displaying impairments primarily in verbal short-term memory and word retrieval relative to the other profile. Importantly, both profiles exhibited comparable severity of persistent reading difficulties despite their divergent cognitive profiles. These findings show that phenotypically similar reading impairments can arise from distinct underlying cognitive factors, emphasizing the role of comprehensive neuropsychological evaluation in developing tailored interventions for affected children.

  • Open Access

    Agentic Generative Artificial Intelligence System for Classification of Pathology-Confirmed Primary Progressive Aphasia Variants

    by Chiara Gallingani, Zachary A Miller, Maria Luisa Mandelli, Howard J Rosen, Zoe Ezzes, Mia Lin, Diana Rodriguez, Lea T Grinberg, Salvatore Spina, William W Seeley, et al.

    Importance Accurate clinical and pathological diagnoses are essential in neurodegenerative diseases, especially given the emergence of pathology-specific disease-modifying therapies. However, diagnostic accuracy remains challenging due to heterogeneous clinical presentations, complexity of integrating multimodal data, and limited access to multidisciplinary expertise. Primary Progressive Aphasia (PPA) exemplifies these challenges, requiring specialized clinical, neuropsychological, and imaging evaluations. Generative artificial intelligence (AI), powered by large language models, may offer scalable diagnostic support in this context. Objective To evaluate the diagnostic performance of an agentic generative AI system in classifying prototypical PPA cases by clinical syndrome and underlying pathology. Design Retrospective diagnostic validation study using a multi-agent generative AI architecture simulating expert-level reasoning. Setting Single tertiary academic referral center (University of California San Francisco, Memory and Aging Center). Participants Fifty-four individuals with a definite diagnosis of PPA and post-mortem confirmation (18 semantic [svPPA], 17 logopenic [lvPPA], 19 nonfluent [nfvPPA]), selected as prototypical cases with congruent clinical, imaging, and pathological profiles. Exposure Multimodal input data, including clinical notes, neuropsychological and language assessments, and MRI brain images, were processed through a multi-agent architecture. The system generated diagnostic predictions under two conditions: (1) open-ended diagnosis from a set of 15 neurodegenerative clinical syndromes; (2) constrained classification of PPA variant and underlying neuropathology. Main Outcomes and Measures Generative AI system diagnostic accuracy for clinical syndrome and pathology, based on expert clinical diagnoses and post-mortem confirmations as gold standard. Results In the open-ended setting, the system correctly identified PPA in 49 of 54 cases (90.7%, chance level=6.7%). When constrained to PPA, it achieved 100% accuracy for svPPA and nfvPPA, and 94.1% for lvPPA as primary prediction. Neuropathological predictions were most accurate for FTLD-TDP type C (100%) and FTLD-4R tau (100%), and high for Alzheimer’s disease (94.4%). The full diagnostic pipeline of all 54 cases was completed in under 10 minutes. Conclusions and Relevance The AI system demonstrated expert-level performance in classifying prototypical PPA cases, integrating multimodal data and mirroring specialist reasoning. Its speed and accuracy support its potential role in extending access to specialized diagnostic expertise, particularly in non-tertiary settings. Further validation in larger and more heterogeneous populations is warranted.


  • Open Access

    Burden of psychiatric disease inversely correlates with Alzheimer's age at onset

    by Emily Eijansantos, Isabel E. Allen, Jessica de Leon, Stephanie Grasso, Nicole Rogers, Rian Bogley, Andrew Paramo, Alexander J. Ehrenberg, Maxime Montembeault, Virginia Sturm, et al.

    INTRODUCTION: Depression is regarded as a risk factor for Alzheimer's disease (AD). Associations between AD and other psychiatric disorders are less clear. METHODS: We screened 1,500 AD UCSF Memory and Aging Center patients for prevalence of psychiatric disorders and compared results to 8,267 NACC AD participants. RESULTS: AD with depression, anxiety, or post‐traumatic stress disorder were significantly younger at age at onset than AD without ( p  < 0.001; p  < 0.001; p  < 0.05). Comorbidity of depression, anxiety and PTSD led to further decreases in AD age at onset. Within the NACC cohort, we further demonstrated an inverse relationship between the severity of depression and anxiety symptoms and AD age at onset. DISCUSSION: Depression, anxiety, and post‐traumatic stress disorder are inversely associated with AD age at onset. Age at onset further decreases with increasing number of psychiatric conditions and increasing severity of symptoms, suggesting that overall burden of psychiatric disease is highly relevant to AD. Highlights: Retrospective chart review revealed that in patients with AD, those who also had depression, anxiety, or post‐traumatic stress disorder were significantly younger at age at onset than those without. Increasing burden of psychiatric disease, both in severity of psychiatric symptoms and number of comorbid psychiatric conditions, produced serial decreases in the age at onset of AD. In patients with AD, those with depression were more likely to have autoimmune disease, and those with anxiety were more likely to have a history of seizures.


  • Open Access

    Neurodevelopment and neural environment inform Alzheimer's disease age at onset and phenotype

    by Zachary A. Miller, Rik Ossenkoppele, Neill R. Graff‐Radford, Isabel E. Allen, Wendy Shwe, Lynne Rosenberg, Dustin J. Olguin, Michael G. Erkkinen, P. Monroe Butler, Salvatore Spina, et al.

    INTRODUCTION: Risk factors associated with sporadic non‐amnestic and early‐onset Alzheimer's disease (AD) remain underexamined. METHODS: We investigated a large, clinically heterogeneous, AD cohort for frequencies of established risk factors (hypertension, hypercholesterolemia, diabetes mellitus) alongside novel factors (non–right‐handedness, learning disability, seizures, autoimmune disease). RESULTS: Early‐onset AD possessed lower frequencies of established risk factors (hypertension, hypercholesterolemia, diabetes mellitus, all p  < 0.001) and higher frequencies of novel factors (non–right‐handedness, learning disability, active seizure, all p  < 0.001; remote seizure, p = 0.002; and autoimmune disease, p  = 0.007). An age at onset < 70 maximally distinguished novel from typical factors. Principal component analysis loaded novel factors into two components, non–right‐handedness and learning disability versus seizure and autoimmune disease, which combined, resulted in an exponential decrease in age at onset from one factor alone. DISCUSSION: Identifying novel factors enriched in early‐onset and non‐amnestic AD introduces new theories of AD susceptibility and phenotypic heterogeneity, with significant implications for disease prediction and treatment. Highlights: We identified a suite of novel factors overrepresented in early‐onset and non‐amnestic AD. These factors can be broadly conceptualized as neurodevelopmental (non–right‐handedness and learning disability) and neural environmental (seizure and autoimmunity). The combination of these factors produced exponential decreases in AD age at onset, compared to each alone, supporting a new theoretical framework for understanding AD risk with implications for disease prediction, prevention, and therapeutic intervention.


  • Open Access

    Awe proneness links RSA and trait creativity in dyslexia

    by Eleanor R. Palser, Belinda Y. Zhang, Jiwoo Han, Alexis I. Martinez-Arroyo, Amanda K. Gerenza, Christa Watson-Pereira, Zachary A. Miller, Maria Luisa Gorno-Tempini, Virginia E. Sturm

  • Open Access

    Network changes associated with right anterior temporal lobe atrophy: insight into unique symptoms

    by Hulya Ulugut, Maria Luisa Mandelli, Anna Gilioli, Zoe Ezzes, Janhavi Pillai, David Baquirin, Amie Wallman-Jones, Amanda Gerenza, Eleanor R Palser, Aaron Scheffler, et al.

    Semantic behavioural variant (also referred to as right temporal) frontotemporal dementia is a newly described syndrome associated with predominant right anterior temporal lobe atrophy and a distinctive combination of behavioural and semantic changes. It is considered the right-sided counterpart of semantic variant primary progressive aphasia, with which it has overlapping neuropathological and cognitive mechanisms. Although more is known about how brain network alterations relate to both losses (e.g. word comprehension deficits) and increases (e.g. hyper-fluency) in cognitive and behavioural processes in the left-sided semantic progressive aphasia, less is known about these phenomena in the right-sided semantic behavioural variant. In this study, we investigated functional connectivity within the right counterparts of established ventral and dorsal cortical speech and language networks and their relationship to specific clinical manifestations in individuals with the semantic behavioural variant. We hypothesized that socioemotional-semantic deficits would be associated with reduced connectivity in the right ventral semantic network, while heightened behavioural manifestations, such as hyper-focus on specific interests (also referred to as rigidity), would be associated with increased connectivity in the right dorsal network. Using seed-based intrinsic connectivity analyses of functional MRI data and cognitive scores from 22 individuals with semantic behavioural variant frontotemporal dementia and 48 cognitively normal individuals, we measured intrinsic connectivity strength in networks anchored in the right anterior middle temporal gyrus (ventral network) and in the right opercular inferior frontal gyrus (dorsal network). Functional connectivity values were then correlated with cognitive and behavioural measurements, controlling for global atrophy. Compared to the control group, individuals with semantic behavioural variant exhibited reduced connectivity in the right ventral network (t = 2.7, P = 0.003), which was associated with socioemotional-semantic deficits (r = 0.47, P = 0.03), as measured by standardized tests. Conversely, increased functional connectivity was observed in the spared right dorsal network (t = 4.26, P < 0.001), which correlated with higher scores for hyper-focus on fixed interests, as measured by retrospective ratings of medical notes (r = 0.63, P = 0.002). Together with previous evidence, these findings suggest that in individuals with predominant anterior temporal lobe atrophy, greater expression of behaviours such as hyper-focus is associated with altered functional dynamics within networks that remain relatively spared by the disease process. This highlights the complex interplay between damaged and spared networks in shaping the clinical manifestations of semantic behavioural variant frontotemporal dementia.


  • Automatic Detection of Articulatory-Based Disfluencies in Primary Progressive Aphasia

    by Jiachen Lian, Xuanru Zhou, Chenxu Guo, Zongli Ye, Zoe Ezzes, Jet M.J. Vonk, Brittany Morin, David Baquirin, Zachary Miller, Maria Luisa Gorno-Tempini, et al.

    Speech corpora are collections of textual data derived from human verbal output and speech signals that can be processed from a variety of perspectives, including formal or semantic content, to serve analyses of different levels of linguistic organisation (phonemic, morphosyntactic, lexico-semantic and content information, prosody and intonation) and to serve analyses of important phenomena such as speech fluency and errors (non-fluencies). We focus on transcribing speech along with non-fluencies or dysfluencies, the detection of which plays an important role in the diagnosis of primary progressive aphasia, where we specifically examine articulation-based dysfluencies in nfvPPA speech. In this work, we propose SSDM 2.0, which is built on top of the current state-of-the-art system of dysfluency detection [1] and tackles its shortcomings via four main contributions: (1) We propose a novel Neural Articulatory Flow for deriving highly scalable, dysfluency-aware speech representations. (2) We develop a full-stack connectionist subsequence aligner to capture all major dysfluency types. (3) We introduce a mispronunciation prompt pipeline and consistency learning into LLMs to enable in-context dysfluency learning. (4) We curate and open-source Libri-Co-Dys [1], the largest co-dysfluency corpus to date. (5) We also present SSDM-L, a modular, non-end-to-end, lightweight model designed for clinical deployment. In clinical experiments on pathological speech transcription, we tested SSDM 2.0 using nfvPPA corpus primarily characterized by articulatory dysfluencies. Overall, SSDM 2.0 outperforms SSDM and all other dysfluency transcription models by a large margin. See our project demo page at https:// berkeley-speech-group.github.io/SSDM2.0/.