Deep Learning Framework for Characterizing Tau‐PET Spatial Heterogeneity along AD Spectrum: A Self‐supervised Approach

Topic: Brain proteins, microglia and lab models

Three robustly differentiated AD subtypes (S1, S2, and S3) were identified, with 93% validation accuracy and Adjusted Mutual Information scores >0.89, each exhibiting distinct tau patterns. S1 showed minimal transentorhinal/temporal pole tau accumulation, S2 had increased medial temporal/posterior parietal tau binding, and S3 demonstrated widespread cortical tau accumulation (see Figure 1). Cross-sectional analysis revealed significantly higher whole-brain tau accumulation (TAU_bl) in S2 than S1 (p <0.001). Cognitive assessments revealed domain specific deficits: memory impairment (RAVLT_immediate_bl) from S1 to S2, language deficits (LDELTOTAL_bl) from S2 to S3, and more severe overall cognitive impairment (MMSE_bl) from S1 to S3. Patient-reported measures remained stable, while study partner assessments indicated greater impairment in memory (EcogSPMem_bl), planning (EcogSPPlan_bl), and organizational (EcogSPOrgan_bl) skills in S3 than S2. Two significant functional connectivity differences were identified between S2 and S3 (see Figure 2) based on non-overlapping 95% credible intervals: Cross Hemispheric connection (right lateral orbitofrontal to left superior temporal) and Default Mode Network connection (right pars orbitalis to right isthmus cingulate). This study leveraged tau-PET data to identify spatial heterogeneity in AD. Multimodal analyses highlighted anatomical, cognitive, and functional connectivity differences between subtypes, emphasizing the potential of deep learning models to explore AD pathology.

Key findings

• Three robustly differentiated AD subtypes (S1, S2, and S3) were identified, with 93% validation accuracy and Adjusted Mutual Information scores >0.89, each exhibiting distinct tau patterns.
• S1 showed minimal transentorhinal/temporal pole tau accumulation, S2 had increased medial temporal/posterior parietal tau binding, and S3 demonstrated widespread cortical tau accumulation (see Figure 1).
• Cross-sectional analysis revealed significantly higher whole-brain tau accumulation (TAU_bl) in S2 than S1 (p <0.001).
• Cognitive assessments revealed domain specific deficits: memory impairment (RAVLT_immediate_bl) from S1 to S2, language deficits (LDELTOTAL_bl) from S2 to S3, and more severe overall cognitive impairment (MMSE_bl) from S1 to S3.
• Patient-reported measures remained stable, while study partner assessments indicated greater impairment in memory (EcogSPMem_bl), planning (EcogSPPlan_bl), and organizational (EcogSPOrgan_bl) skills in S3 than S2.

Why it matters

A self-supervised deep learning approach identified three AD subtypes with distinct tau-PET patterns and different cognitive deficits, with 93% validation accuracy. This describes disease heterogeneity for classification, not a treatment, cure, or prevention.

New to this subject? Start with our plain-English guide: Brain scans & tests →

Study details

• Study designcross-sectional
• Evidence strengthLow

Read the original

• DOI: 10.1002/alz70862_109949
• PubMed: 41433750
• Original source →
• Free full text: Alzheimer s & Dementia →

Full text: free to read — OpenAlex lists this copy as free to read: the link above (checked 6 Feb 1020).

@article{jayaprakash2025-41433750,
  author  = {Jayaprakash, Theyaneshwaran and Cornelison, Connor Lee and Powla, Plamena P. and Zhang, Lu and Fredericks, Carolyn A. and Saykin, Andrew J. and Wang, Selena},
  title   = {{Deep Learning Framework for Characterizing Tau‐PET Spatial Heterogeneity along AD Spectrum: A Self‐supervised Approach}},
  journal = {Alzheimer s \& Dementia},
  year    = {2025},
  doi     = {10.1002/alz70862_109949},
  pmid    = {41433750},
  url     = {https://doi.org/10.1002/alz70862_109949}
}

Related studies in Brain scans and imaging

• Adaptive template generation for amyloid PET using a deep learning approach Human brain mapping · 2018-05-11 · usefulness score 39/100
• Investigating the Amyloid-Tau-Neurodegeneration Framework in Alzheimer’s Disease Using Semi-Supervised Multimodal Imaging Data Fusion medRxiv · 2025-12-15 · usefulness score 16/100
• Small-Data Deep Learning for Alzheimer-Spectrum Classification from Structural MRI: A Feasibility Study Using OASIS Journal of Imaging · 2026-08-03 · usefulness score 42/100
• A novel neuroimaging based early detection framework for alzheimer disease using deep learning Scientific Reports · 2025-07-02 · usefulness score 47/100
• Transfer deep learning and explainable AI framework for brain tumor and Alzheimer's detection across multiple datasets Frontiers in Medicine · 2025-06-19 · usefulness score 41/100
• An automated multimodal medical image fusion framework for Alzheimer detection using deep learning Journal of Combinatorial Mathematics and Combinatorial Computing · 2025-04-23 · usefulness score 31/100

Similar studies

• Brain tau PET-based identification and characterization of subpopulations in patients with Alzheimer’s disease using deep learning-derived saliency maps 2025 · Genetics
• A Harmonized Visual Reading Framework for Tau PET Staging in Alzheimer Disease 2026 · Brain scans and imaging
• Tau PET patterns mirror clinical and neuroanatomical variability in Alzheimer's disease 2016 · Laboratory science
• Tau PET Imaging in Neurodegenerative Disorders 2022 · Brain scans and imaging
• Distinct tau PET patterns in atrophy-defined subtypes of Alzheimer's disease 2020 · Mild cognitive impairment
• Tau PET in autosomal dominant Alzheimer's disease: relationship with cognition, dementia and other biomarkers 2019 · Brain scans and imaging

More on Brain scans and imaging

• A review that covers this ground The Use of Tau PET to Stage Alzheimer Disease According to the Braak Staging Framework Journal of nuclear medicine : official publication, Society of Nuclear Medicine · 2023-06-15 · usefulness score 49/100
• Newest in this subject Large-scale head-to-head comparison of on- and off-target profiles of [18F]Tauvid and [18F]MK6240 tau-PET in aging and Alzheimer’s disease Research Square · 2026-11-01 · Preprint — not yet peer-reviewed · usefulness score 16/100
• Newest in this subject Repetitive negative thinking is associated with amyloid accumulation over 3 years in cognitively healthy adults medRxiv · 2026-10-31 · Preprint — not yet peer-reviewed · usefulness score 16/100

See all studies in Brain scans and imaging →