Multimodal Biomarkers in Neurodegenerative Disease : Integrating AI-Assisted Brain MRI and CT Analysis with CSF and Audiometric Measures
Year: 2026
Type: dissertation
Study design: Observational Study
Abstract: Alzheimer's disease (AD) is the leading cause of dementia, characterised by progressive impairment of memory and other cognitive functions.AD is neuropathologically defined by β-amyloid deposition and tau-related neurofibrillary degeneration leading to synaptic loss and neuronal atrophy.Over recent decades, the paradigm of AD has shifted from a clinically defined syndrome to a biologically defined construct, characterised and staged by biomarker evidence of its core hallmarks-amyloid (A), tau (T), and neurodegeneration (N).Biomarker-based methods such as cerebrospinal fluid (CSF) assays and magnetic resonance imaging (MRI) volumetry have become principal in vivo tools for identifying AD pathology.However, the generalisability of these biomarkers from highly selected research cohorts to more heterogeneous real-world clinical populations requires validation.Access to certain methods, such as MRI, varies across healthcare systems.Early functional manifestations, including sensory deficits such as hearing impairment, may interact with cognitive decline and neurodegeneration but have not been systematically linked to structural or molecular biomarkers.This dissertation examined complementary biomarker modalities to improve early diagnosis and clinical stratification of neurodegenerative 8 diseases.The specific aims were (I) to determine whether CSF and volumetric MRI biomarkers in a real-world memory-clinic cohort show patterns across clinical diagnostic and AT(N) groups consistent with those reported in research-based cohorts (Study I); (II) to validate convolutional neural network (CNN)-based CT analysis against MRI for quantifying brain atrophy and white matter lesions (WMLs) (Study II); and (III) to assess how sensorineural hearing impairment relates to artificial intelligence (AI)derived MRI measures of structural changes and cognitive status (Study III).The work was based on retrospective and cross-sectional data from Kuopio University Hospital (KUH), the Brain Research Unit of the University of Eastern Finland and collaborating sites.Study I included 191 patients with CSF sampling; Aβ42, phosphorylated tau, and total tau concentrations were used to define AT(N) categories and volumetric MRI analyses were performed with FreeSurfer.Study II comprised 214 individuals with paired CT and MRI scans from KUH and Helsinki University Hospital (HUS), analysed using a CNN-based CT pipeline developed by Combinostics Ltd, which generated automated medial temporal (MTA), global cortical (GCA), and WML ratings that were compared with quantitative MRI volumetrics.Study III (APOLLO cohort) included 107 individuals, 40 cognitively normal (CN), 34 with mild cognitive impairment (MCI), and 33 with AD, whose better-ear pure-tone averages (PTA 0.5-4 kHz) and Finnish Matrix Sentence Test (FMST) speech-in-noise thresholds were analysed alongside AI-derived MRI atrophy measures (MTA and GCA).CSF and MRI biomarkers retained discriminative value consistent with the AT(N) framework.CNN-derived CT measures of MTA, GCA and WMLs strongly corresponded with MRI, supporting CT as a feasible alternative when MRI is unavailable.Hearing impairment was associated with MTA and GCA, indicating a potential functional correlation with neurodegeneration.Together, these findings support the implementation of clinically accessible biomarkers, bridging research and clinical practice.
Author: Kaipainen, Aku
Institution: University of Eastern Finland
Language: English
Subfield: Psychiatry and Mental health
Field: Medicine
Domain: Health Sciences
Keywords: Alzheimer's disease, multimodal biomarkers, cerebrospinal fluid biomarkers, amyloid, phosphorylated tau, total tau, brain atrophy, white matter lesions, medial temporal lobe atrophy, cortical atrophy, sensorineural hearing loss, cognitive decline, early diagnosis, neurodegeneration, memory clinic patients
Open Access status: green
Award: G3968078322
OpenAlex ID: W7215618208
