2026 ECNP Congress

PS04-4231

Exploring neuroanatomical patterns of primary progressive aphasia in patients with primary psychiatric disorders

Nathalie Matti — A. Wiegand, M. Montembeault, S. de Boer, Y. Pijnenburg, N. Koutsouleris · Ludwig-Maximilian University Munich, Psychiatry and Psychotherapy, München, Germany

2026 ECNP Congress

Background

Language abnormalities are a core feature of psychotic disorders. Disorganized speech, an observable manifestation of formal thought disorder, was among the earliest described psychopathological phenomena in schizophrenia (SCZ). Primary progressive aphasia (PPA) is a clinical syndrome within the spectrum of frontotemporal lobar degeneration (FTD), characterized by progressive language impairment due to focal degeneration of language-dominant brain regions. In the present study, we conceptualize semantic (svPPA) and non-fluent agrammtic/non-fluent variant (nfvPPA) as model disorders of language dysfunction to provide a framework for investigating disorganized speech in psychiatric conditions.

Methods

Structural MRI (sMRI) data from the Frontotemporal Lobar Degeneration Neuroimaging Initiative (FTLDNI/NIFD, http://memory.ucsf.edu/research/studies/nifd) study were used to train support vector machine (SVM) classifiers to distinguish svPPA and nfvPPA from healthy controls (HC) respectively. Each classifier was then applied to psychiatric cohorts derived from the Personalised Prognostic Tools for Early Psychosis Management study (PRONIA; www.pronia.eu) [1], the Munich brain database [2] and the Zurich Early Recognition Program (ZInEP) [3], comprising the following psychiatric groups: recent-onset depression (ROD); recent-onset psychosis (ROP), subdivided into schizophrenia (SCZ) and non-schizophrenia schizophrenia (NON-SCZ); clinical high risk for psychosis (CHR-P) further subdivided into ultra-high risk (UHR) and basic cognitive symptoms (COGDIS), basic perceptive symptoms (COPER) and mixed UHR/COGDIS (Mixed); first-episode schizophrenia (FE-SCZ); recurrent schizophrenia (RE-SCZ); major depressive disorder (MDD); borderline personality disorder (BOR); bipolar disorder (BIP); high-risk bipolar (HR-BIP); and HC. Model application generated a mean decision score for each individual, reflecting classification along a continuum between HC and PPA subtype. Decision scores were adjusted for the Brain Age Estimation Gap (BrainAGE)[4]. Neuroanatomical correlates of machine learning–derived decision scores were examined using voxel-wise mass-univariate multiple regression in SPM12. Clinical correlations were investigated using support vector regression models trained on anthropometric, neurocognitive, polygenic risk and clinical variables to predict SVM-derived decision scores.

Results

The SVM classifier distinguishing HC from svPPA performed at a balanced accuracy (BAC) of 98.0%, HC from nfvPPA 86.6% and nfvPPA from svPPA 92.1% BAC. After BrainAGE correction, ROP-SCZ, ROP-NONSCZ, RE-SCZ differed from HC in rank-based post hoc comparisons (Dunnett’s test) for the svPPA pattern and in ROP-SCZ and ROP-NONSCZ for the nfvPPA-pattern. Spatial analysis indicated involvement of the temporal lobes and poles, as well as subcortical structures including the caudate, putamen and pallidum, alongside limbic and paralimbic regions. Regression models explained 7.60-12.1% of the variance in the decision scores. Among biopsychosocial correlates, BMI, waist circumference, and age showed the highest cross-validation ratios among the features. PRS for MDD and BIP were associated with a disease-like pattern. Female sex and fulfilling the UHR criteria was associated with a HC-like pattern.

Conclusions

Overall, decision scores reflected a predominantly HC-like pattern, however, they shifted toward a more PPA-like pattern in several psychiatric groups, most prominently in ROP-SCZ and ROP-NONSCZ. Leveraging domain-specific model disorders may help to better understand symptom clusters in psychiatric conditions.

References

  1. Koutsouleris N, Dwyer DB, Degenhardt F, Maj C, Urquijo-Castro MF, Sanfelici R, et al. Multimodal Machine Learning Workflows for Prediction of Psychosis in Patients With Clinical High-Risk Syndromes and Recent-Onset Depression. JAMA Psychiatry. 2021 Feb 1;78(2):195–209. doi:10.1001/jamapsychiatry.2020.3604 PubMed PMID: 33263726; PubMed Central PMCID: PMC7711566.
  2. Meisenzahl EM, Koutsouleris N, Bottlender R, Scheuerecker J, Jäger M, Teipel SJ, et al. Structural brain alterations at different stages of schizophrenia: a voxel-based morphometric study. Schizophr Res. 2008 Sep;104(1–3):44–60. doi:10.1016/j.schres.2008.06.023 PubMed PMID: 18703313.
  3. Theodoridou A, Heekeren K, Dvorsky D, Metzler S, Franscini M, Haker H, et al. Early Recognition of High Risk of Bipolar Disorder and Psychosis: An Overview of the ZInEP “Early Recognition” Study. Front Public Health. 2014 Oct 1;2. doi:10.3389/fpubh.2014.00166
  4. Koutsouleris N, Davatzikos C, Borgwardt S, Gaser C, Bottlender R, Frodl T, et al. Accelerated brain aging in schizophrenia and beyond: a neuroanatomical marker of psychiatric disorders. Schizophr Bull. 2014 Sep;40(5):1140–53. doi:10.1093/schbul/sbt142 PubMed PMID: 24126515; PubMed Central PMCID: PMC4133663.