Corey T McMillan
· Ph.D.VerifiedUniversity of Pennsylvania · Rehabilitation Medicine
Active 1993–2026
About
Corey T McMillan, PhD, is an Associate Professor of Neurology at the University of Pennsylvania's Perelman School of Medicine. He is a member of several research centers including the Institute for Translational Medicine & Therapeutics, the Penn Neurodegeneration Genomics Center, and the MindCORE. Dr. McMillan is also a Fellow at the Penn Institute on Aging within the Division of Clinical and Translational Neurodegenerative Disease Research and serves as Co-Director of the Penn Frontotemporal Degeneration Center (FTDC). His research focuses on using multimodal and bioinformatic approaches to understand the biological basis of neurodegenerative conditions, with an emphasis on developing biomarkers for diagnosis, drug discovery, and clinical trial endpoints. His work concentrates on neurodegenerative proteinopathies involving misfolded tau protein, contributing to Alzheimer’s disease, primary age-related tauopathy, and frontotemporal lobar degeneration, as well as TDP-43 proteinopathies related to FTLD and ALS. Dr. McMillan's research integrates MRI, PET imaging, genomics, and clinical data to address questions in both basic science and clinical translation, aiming to improve diagnosis and treatment of neurodegenerative diseases.
Research topics
- Medicine
- Neuroscience
- Biology
- Pathology
- Psychology
- Computer Science
- Artificial Intelligence
- Genetics
- Radiology
- Internal medicine
- Chemistry
- Cell biology
- Molecular biology
- Physical medicine and rehabilitation
Selected publications
Automated Imaging Differentiation for Dementia
Neurology Open Access · 2026-04-13
articleOpen accessDifferentiation of Alzheimer disease dementia (ADD) and dementia with Lewy bodies (DLB) remains a challenge. Free-water (FW) imaging has been investigated in neurodegenerative diseases and was found to be associated with neurodegeneration and neuroinflammation. This retrospective cohort study tested whether Automated Imaging Differentiation for Dementia (AIDD), combining diffusion free-water imaging (FWI) and support vector machine, predicts ADD vs DLB with high accuracy.
Brain telomere length associates with hippocampal ptau and is mediated by DNA methylation
Alzheimer s & Dementia · 2025-12-01
articleOpen accessBACKGROUND: Telomeres are repetitive DNA sequences at the ends of chromosomes which contribute to maintaining chromosomal stability. Telomere shortening is a hallmark of aging and shorter blood leukocyte telomere length (LTL) has been associated with increased risk for age-related diseases, however, little is understood about the biology of brain telomeres and how they may be involved in disease. Considering the increased neuropathologic burden of phosphorylated tau (ptau) with age, we investigated how shorter brain telomere length (brain-TL) may relate to increased ptau burden. METHODS: We studied a cohort of 112 individuals with primary age-related tauopathy (PART), a neuropathological diagnosis characterized by mild-to-moderate tau burden (Braak=I-IV) primarily in the medial temporal lobe, with the relative absence of amyloid-beta plaques (CERAD=0). These individuals had both brain-TL (mean length by telomere qPCR, blinded) and DNA methylation measures from the frontal cortex, along with semi-quantitative Aperio ptau measures from the hippocampus. A subset (n = 81) had SNP genotyping data available. In an independent cohort (n = 10, Braak=0-VI, CERAD=0-3), we performed quantitative fluorescence in-situ hybridization (FISH) microscopy to measure the average ratio of telomere to centromere DNA content in nuclei from the frontal and visual cortices. RESULTS: In linear regression models, frontal cortex brain-TL did not relate to age. When age-adjusted, shorter brain-TL related to higher hippocampal ptau (β=-1.06, CI=-1.92--0.195, p = 0.017). A previously established DNA methylation model predictive of hippocampal ptau partially mediated the relationship between brain-TL and hippocampal ptau (proportion mediated=0.664, CI=0.246-1.33, p = 0.012, Figure 1). A polygenic score for LTL did not relate to either age, brain-TL or hippocampal ptau. With FISH, we observed that individuals with CERAD=0 had shorter telomeres in the frontal cortex compared to individuals with CERAD=3. Within the CERAD=0 group, an individual with Braak=II had shorter telomeres than an individual with Braak=I. These patterns were not observed in the visual cortex. CONCLUSIONS: In a PART cohort, shorter frontal cortex brain-TL was related to higher hippocampal ptau, and this relationship was partially mediated by a DNA methylation model predictive of hippocampal ptau. A polygenic score for LTL was not predictive of brain-TL or hippocampal ptau. Together, this further emphasizes the importance of tissue-specific epigenetic modifiers of age-related ptau neuropathology.
medRxiv · 2025-09-07 · 1 citations
preprintOpen accessABSTRACT Objective To compare PET and plasma-based temporal modeling of amyloid and tau biomarkers in Alzheimer’s disease Methods Longitudinal amyloid PET, 18 F-flortaucipir tau-PET, and Fujirebio Lumipulse plasma p-tau 217 from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) and University of Pennsylvania Alzheimer’s Disease Research Center (Penn ADRC) were used to generate biomarker trajectory models using Sampled Iterative Local Approximation (SILA). SILA models using plasma p-tau 217 were compared to amyloid and tau PET-based models to estimate tau onset age (ETOA) and estimate amyloid onset age (EAOA), and factors influencing ETOA and time from ETOA to dementia were evaluated for PET and plasma-based models. Results Plasma-based models generated similar results to PET for EAOA and ETOA, with stronger model agreement for ETOA than EAOA. Accuracy of estimated onset age compared to actual onset age was high within modality with slightly greater error when comparing across modalities (i.e. plasma to PET). For both plasma and PET models, earlier ETOA was associated with younger EAOA, female sex, and ≥1 ApoE ε4 allele. Earlier dementia onset after ETOA was associated with later ETOA for both plasma and PET models, while male sex was associated with shorter tau to dementia gap in plasma models. Interpretation Temporal modeling of plasma biomarkers provides comparable information to PET-based models, particularly for tau onset age. Plasma-based temporal modeling can serve as a widely accessible tool for clinical assessment of biological disease duration that places the patient on the disease timeline, which may allow for improved discussion of prognosis and treatment decisions.
Polygenic scores and baseline cognitive function in midlife
Alzheimer s & Dementia · 2025-12-01
articleOpen accessSenior authorBACKGROUND: Midlife is a key life-course period for understanding risk factors of cognitive decline. Despite growing evidence demonstrating polygenic contributions to age-related disease risk, less is understood about how polygenic scores (PGS) may relate to cognitive traits in middle age. In a cross-sectional analysis, we investigated how PGS for Alzheimer's disease (AD), longevity, subjective-wellbeing (based on life satisfaction and positive affect), and cognitive-function may relate to cognition in a cohort of middle-aged adults with normal cognition. METHODS: The Health and Retirement Study (HRS) is an ongoing survey in Americans aged 50 years or older. We studied a cohort of 6615 individuals from the HRS (5359 European-ancestry, 1256 African-ancestry) aged 50-56 years with normal cognition at baseline, as determined by the HRS cognitive test battery. PGS derived from genetic variants associated with traits through genome-wide association studies in independent cohorts were available for all individuals. Linear regressions adjusted for baseline age, gender, and years of education were performed separately by genetically-defined ancestral group. RESULTS: Higher baseline cognition was observed in females (European-ancestry β=0.72±0.08,p<0.001; African-ancestry β=0.46±0.18,p=0.009) and individuals with more years of education (European-ancestry β=0.39±0.02,p<0.001; African-ancestry β=0.29±0.04,p<0.001). Higher cognitive-function PGS related to higher cognition (European-ancestry β=0.19±0.04,p<0.001; African-ancestry β=0.20±0.09,p=0.024). In the European-ancestry group, the cognitive-function PGS partially mediated the relationship between education and cognition (proportion mediated in European-ancestry=0.019, CI=0.010-0.030,p<0.001; proportion mediated in African-ancestry=0.016, CI=-0.002-0.050,p=0.072). The cognitive-function PGS still partially mediated the relationship between education and cognition in a subset of the European-ancestry group matched to the African-ancestry group based on age, gender, education, and cognition (n =1256 European-ancestry, proportion mediated=0.016, CI=0.002-0.040,p=0.016). Therefore, the lack of mediation in the African-ancestry group is likely not due to smaller sample size. We did not observe associations between cognition and PGS for AD, longevity, and subjective-wellbeing in either midlife ancestral group. CONCLUSIONS: In this HRS cohort, more years of education were related to higher baseline cognition, and in individuals with European-ancestry, this relationship was partially mediated by a PGS for cognitive-function. Future directions include assessing the contribution of PGS to cognition over time, and investigating how environmental and life-course factors may moderate the association between PGS and cognition.
Alzheimer s & Dementia · 2025-12-01
articleOpen accessAbstract Background Postmortem MRI has opened‐up avenues to study brain structure at ultra high‐resolution revealing details not possible to observe with in vivo MRI. Here, we present a novel package (purple‐mri) which performs tissue segmentation, anatomical parcellation and spatial normalization of postmortem MRI. Additionally, we provide a framework to perform point‐wise surface‐based group‐level studies linking morphometry/histopathology in common coordinate system for postmortem MRI. Method We developed a joint voxel‐ and surface‐based pipeline combining deep learning with classical techniques for topology correction, cortical modeling, inflation, and registration for accurate parcellation of postmortem cerebral hemispheres (Figure 1 Khandelwal et al. 2024). Furthermore, using the GM/WM segmentations derived from postmortem hemisphere and FreeSurfer‐processed antemortem MRI, we performed deformable image registration between the ante‐ and postmortem MRI for each brain specimen. To demonstrate the utility of purple‐mri, point‐wise analysis was performed to correlate thickness (mm) with tau and neuronal loss distribution in corresponding specimens ( N = 49) of postmortem (7T at 0.3mm 3 ) and antemortem (3T at 0.8mm 3 ) MRI (Table 1) within the AD continuum diagnosis. An additional 26 postmortem 7T scans without corresponding antemortem scans were included in some analyses. The semi‐quantitative average tau and neuronal loss ratings were derived from histopathological examination across the brain. All analyses include age, sex, and postmortem (or antemortem) interval as covariates. Result Our method parcellates postmortem brain hemisphere using a variety of brain atlases even in areas with low contrast (anterior/posterior regions), profound imaging artifacts and severely atrophied brains (Figure 1). Our registration pipeline provides one‐to‐one correspondence between the two modalities. For thickness/pathology associations (Figure 2), small sparse significant clusters only in superior temporal cortex and precuneus in antemortem MRI ( N = 49) were observed. However, postmortem MRI showed much stronger associations across large clusters in temporal, entorhinal cortex, and cingulate for both the matched cases ( N = 49) and the full cohort ( N = 75), regions implicated in ADRD. Conclusion Purple‐mri paves the way for large‐scale postmortem image analysis. Stronger associations between thickness and average tau burden/neuronal loss than antemortem MRI suggests that our pipeline could inform the development of more precise and sensitive invivo biomarkers by mapping information from postmortem to antemortem MRI in a common reference coordinate system.
Alzheimer s & Dementia · 2025-12-01
articleOpen accessAbstract Background 18 F‐Flortaucipir is widely used for positron emission tomography (PET) imaging of Alzheimer’s disease (AD)‐type tau, but its sensitivity in early Braak stages has been questioned, and hippocampal uptake is at least partially confounded by off‐target binding. We investigated associations between antemortem PET uptake and digitally‐quantified tau and TDP‐43 neuropathology. Methods Participants ( n = 14, Figure 1) included 5 people with no/low AD neuropathologic change (ADNC) at autopsy, 2 intermediate, and 7 high. Clinical diagnoses included normal cognition ( n = 2), AD ( n = 5), dementia with Lewy bodies ( n = 2), Parkinson’s disease dementia ( n = 1), corticobasal syndrome ( n = 2), and posterior cortical atrophy ( n = 2). We used the Automated Segmentation of Hippocampal Subfields T1 MRI pipeline to segment anterior and posterior hippocampus, Brodmann’s areas (BA) 35 (transentorhinal cortex) & 36, and entorhinal cortex. 18 F‐Flortaucipir standardized uptake value ratios (SUVRs) were computed relative to inferior cerebellar grey matter and averaged across hemispheres. Postmortem sampling comprised hippocampal subiculum, CA1, CA2, CA3, and dentate gyrus; BA35 and BA36; and entorhinal cortex. FFPE‐brain tissue was immunostained using PHF1 and phospho‐specific TDP‐43 antibodies and digitally imaged. Two different weakly supervised learning algorithms, Wildcat, were trained to identify tau tangles and threads; or somatic and neuritic phosphorylated TDP‐43 (pTDP‐43) inclusions. Each pathology type was quantified by summary statistics on Wildcat heatmaps, averaged over hippocampal subfields; and over BA35/entorhinal cortex (Denning et al., 2024). We computed non‐parametric correlations between SUVRs and pathology measures at a=0.05 with false discovery rate correction. Results Tangles were associated with SUVRs in BA35/entorhinal cortex (Spearman’s r=0.59, p = 0.029; Figure 2A). The mean hippocampal tangle measure was associated with SUVRs in both anterior (r=0.77, p = 0.002) and posterior (r=0.77, p = 0.002) hippocampus (Figure 2B‐C). Associations between SUVR and tau threads were marginally significant (Figure 2D‐F). In BA35/entorhinal cortex, 9/9 intermediate‐high ADNC cases had SUVRs above an established positivity cutoff of 1.23 (Figure 3). PET SUVRs were not associated with somatic or neuritic pTDP‐43 measures. Conclusion Results suggest 18 F‐flortaucipir is sensitive to tau burden in early Braak‐stage regions and primarily reflects neurofibrillary tangles rather than thread‐like tau inclusions.
Developing a novel metric of atypicality in Alzheimer's Disease regional volume changes
Alzheimer s & Dementia · 2025-12-01
articleOpen accessBACKGROUND: Recent research on heterogeneity in Alzheimer's disease (AD) highlights the need for imaging metrics that are equally sensitive in patients with atypical (limbic- or neocortical-predominant) and typical patterns of degeneration. We developed an original model based on MRI volumetry that independently quantified disease atypicality and severity and related these metrics to clinical and imaging measures of AD in the ADNI repository. METHOD: In a dataset of 465 3T MRI scans (230 male; 264 cognitively normal (CN), 164 MCI, 37 AD), we computed regional volumes adjusting for intracranial volume for the neocortex (average of superior temporal, middle frontal, and inferior parietal cortices) and anterior hippocampus. We then plotted each individual's hippocampal and neocortical atrophy as coordinates in a 2-dimensional atrophy space (Figure 1). We used the centroids of CN and MCI/AD cases, respectively, to define a line representing the typical relationship between hippocampal and neocortical atrophy. For each individual, we defined a vector connecting their observed atrophy to the CN centroid. Atypicality was quantified by the angle between this vector and the typicality reference line: an individual whose vector lay above the typicality line had a positive angle, indicating hippocampal sparing; while a vector below the line had a negative angle, indicating limbic-predominant atrophy. Individual disease severity was quantified by vector length. We evaluated this model using linear regression to test associations of atypicality and severity with tau PET uptake in entorhinal cortex; and with three memory measures: delayed recall on the logical memory test and Rey auditory verbal learning test (RAVLT) plus recognition on the RAVLT. RESULT: = 0.018). CONCLUSION: A novel atypicality metric was related to both the burden of entorhinal tau and degree of memory impairment in ADNI cases, independently of severity. Future research will extend this model to atypical clinical syndromes and evaluate its ability to distinguish typical, limbic predominant, and hippocampal sparing cases.
medRxiv · 2025-08-19
preprintOpen accessAbstract Objective Plasma biomarkers of Alzheimer’s disease (AD) pathology are frequently tested in specialized research settings, limiting generalizability of findings. Using electronic health records and banked plasma, we evaluated plasma biomarkers – phosphorylated tau 217 (p-tau 217 ), β-amyloid 1-42/1-40 (Aβ 42 /Aβ 40 ) and p-tau 217 /Aβ 42 – in a real-world, diverse clinical population with multimorbidities. Methods Participants (n=617; 44% Black/African American; 41% female) were selected from the University of Pennsylvania Medicine BioBank with plasma assayed using Fujirebio Lumipulse. International Classification of Diseases (ICD) Ninth and Tenth Revision codes determined AD dementia (ADD; n=43), mild-cognitive impairment (MCI; n=140), unspecified/non-AD cognitive impairment (CI; n=106), and cognitively normal cases (n=328), and other medical histories. APOE ε4, body mass index (BMI), metrics of kidney function ( e.g. , eGFR), and liver disease were derived from electronic health records. Multivariable models identified factors related to plasma levels. Previously established cutpoints classified AD status (“AD+”, “AD-”, or “Intermediate”). Results Plasma p-tau 217 /Aβ 42 had the strongest association with known AD-related factors – MCI, ADD, future progression to MCI/ADD, age, and APOE ε4 – compared to p-tau 217 and Aβ 42 /Aβ 40 . Plasma p-tau 217 /Aβ 42 was also associated with eGFR, diabetes, and history of hearing loss. Importantly, AD-related factors were most frequent/severe for AD+ classification by p-tau 217 /Aβ 42 , while medical morbidities were most frequent/severe for Intermediate classification. Exploratory analyses test p-tau 217 /Aβ 42 adjusted for eGFR to eliminate its influence on plamsa levels. Interpretation In this real-world dataset, we identified effects of multimorbidities on plasma biomarkers, especially kidney function. The p-tau 217 /Aβ 42 ratio had low rates of Intermediate classification and may help to account for multimorbidity effects on plasma levels.
Racial/ethnic differences in neuropsychological test performance in frontotemporal degeneration
Alzheimer s & Dementia Diagnosis Assessment & Disease Monitoring · 2025-10-01
articleOpen accessAbstract BACKGROUND Racial and/or ethnic differences in neuropsychological test performance are understudied in frontotemporal degeneration (FTD) but their identification is critical to identifying ways to improve care of representative FTD populations. METHODS Differences in cognitive scores between Black ( n = 56) and Hispanic ( n = 76) relative to White ( n = 2281) participants and the likelihood of impairment status in cognitive test performance were evaluated. RESULTS Minoritized individuals had lower scores and/or greater likelihood of impairment on measures of lexical retrieval, processing speed, cognitive flexibility, and working memory but not global cognition, verbal recall, attention, and category fluency. Addition of severity, age ( M = 65.18), sex (40% female), education ( M = 15.62), and vascular comorbidities attenuated group differences. DISCUSSION Racial/ethnic differences on neuropsychological tests used in diagnosis and monitoring of FTD were substantially attenuated when accounting for potential contributing factors. To address these differences in FTD, future efforts must increase representative research participation of patients and understand social determinants of health. Highlights Racially/ethnically minoritized individuals with frontotemporal dementia are severely underrepresented in the National Alzheimer's Coordinating Center dataset Racially/ethnically minoritized individuals with frontotemporal dementia obtained lower scores and greater likelihood of impairment on common neuropsychological tests The effect of racial/ethnic group on neuropsychological test scores was substantially attenuated when adjusting for disease severity, education level, sex, and age
Genetic modifiers of TDP‐43 proteinopathy in FTLD‐TDP and LATE‐NC
Alzheimer s & Dementia · 2025-12-01
articleOpen accessSenior authorBACKGROUND: Transactive response DNA binding protein of 43 kDa (TDP-43) aggregates are observed in cognitive disorders of aging including frontotemporal lobar degeneration (FTLD- TDP) and limbic-predominant age-related TDP-43 encephalopathy neuropathologic change (LATE-NC). Convergence and divergence in the genetic architecture of these disorders remains unclear. To further elucidate genetic modifiers of FTLD-TDP versus LATE-NC, we evaluated how genetic variants previously associated with TDP-43 proteinopathy relate to neuropathological diagnosis. METHODS: We identified 393 individuals in the Penn Integrated Neurodegenerative Disease Database with: (1) neuropathological diagnosis of FTLD-TDP or LATE-NC; (2) ABC score of Alzheimer's disease neuropathologic change (ADNC); (3) SNP genotyping; (4) no pathogenic mutation (GRN, C9orf72, etc.); and (5) self-reported White race (Table 1). We evaluated 14 genetic variants (MAF > 0.1, missingness < 20%) previously associated with autopsy-confirmed FTLD-TDP, LATE-NC, or regional TDP-43 pathology, comparing SNP frequency between the FTLD-TDP and LATE-NC groups using binomial logistic regression with additive models covarying for sex and ABC score (Table 2). We excluded FTLD-TDP cases of type C or unspecified histological type in a sensitivity analysis. RESULT: Most LATE-NC cases (78%) showed intermediate/high ADNC. Adjusting for sex and ADNC level, we observed higher frequency of rs12973192-G (UNC13A, β = 0.61±0.25, p = 0.013) and lower frequency of rs12425381-G (RERG, β = -0.68±0.30, p = 0.024) in FTLD-TDP versus LATE-NC at a nominal level of significance (p < 0.05). The remaining variants we studied did not differ in allele frequency between FTLD-TDP and LATE-NC after adjusting for sex and ADNC level. Excluding known and possible FTLD-TDP type C cases strengthened the associations of rs12973192-G and rs12425381-G but did not yield new associations. CONCLUSION: A variant in UNC13A appears to increase risk of FTLD-TDP relative to LATE-NC, while a variant in RERG appears to decrease risk; this is directionally concordant with previously reported associations of these variants with FTLD-TDP versus healthy controls. The potential implications of polygenic contributions to prognosis are an important area for future investigation.
Recent grants
NIH · $642k · 2018
NIH · $152k · 2012
NIH · $93k · 2008
NIH · $799k · 2021
NIH · $34.7M · 2014–2030
Frequent coauthors
- 379 shared
Murray Grossman
- 304 shared
David J. Irwin
University of Pennsylvania
- 196 shared
John Q. Trojanowski
University of Pennsylvania
- 185 shared
David A. Wolk
California University of Pennsylvania
- 155 shared
Edward B. Lee
- 99 shared
Katya Rascovsky
University of Pennsylvania
- 88 shared
Vivianna M. Van Deerlin
University of Pennsylvania
- 77 shared
Lauren Elman
University of Pennsylvania
Labs
Corey T McMillan LabPI
Awards & honors
- Penn Institute on Aging Fellow
- Penn Frontotemporal Degeneration Center Co-Director
- Institute for Biomedical Informatics Senior Fellow
- Penn Neurodegeneration Genomics Center Member
- Institute for Translational Medicine & Therapeutics Member
- Resume-aware match score
- Save to shortlist
- AI-drafted outreach
See your match with Corey T McMillan
PhdFit ranks faculty by your research interests, methods, and publications — grounded in their actual work, not templates.
- Free to start
- No credit card
- 30-second signup