Print Close

MRI-based Clustering of a Memory Clinic Population: Atrophy Patterns across Dementia Etiologies

Poster No:

194

Submission Type:

Abstract Submission

Authors:

Myrthe van Haaften¹, Nathalie Koorn¹, Eline Vinke¹, Frans Vos², Henri Vrooman¹, Rozemarijn van Bruchem-Visser¹, Harro Seelaar¹, Francesco Mattace-Raso¹, Esther van den Berg¹, Meike Vernooij¹, Esther Bron¹

Institutions:

¹Erasmus MC University Medical Center, Rotterdam, the Netherlands, ²Technical University of Delft, Delft, the Netherlands

First Author:

Myrthe van Haaften, MSc
Erasmus MC University Medical Center
Rotterdam, the Netherlands

Co-Author(s):

Nathalie Koorn
Erasmus MC University Medical Center
Rotterdam, the Netherlands

Eline Vinke
Erasmus MC University Medical Center
Rotterdam, the Netherlands

Frans Vos
Technical University of Delft
Delft, the Netherlands

Henri Vrooman
Erasmus MC University Medical Center
Rotterdam, the Netherlands

Rozemarijn van Bruchem-Visser
Erasmus MC University Medical Center
Rotterdam, the Netherlands

Harro Seelaar, MD, PhD
Erasmus MC University Medical Center
Rotterdam, the Netherlands

Francesco Mattace-Raso
Erasmus MC University Medical Center
Rotterdam, the Netherlands

Esther van den Berg, PhD
Erasmus MC University Medical Center
Rotterdam, the Netherlands

Meike Vernooij, MD, PhD
Erasmus MC University Medical Center
Rotterdam, the Netherlands

Esther Bron, PhD
Erasmus MC University Medical Center
Rotterdam, the Netherlands

Introduction:

Dementia is a syndrome with a variety of underlying diseases, including Alzheimer's disease (AD), frontotemporal dementia (FTD), dementia with Lewy bodies (DLB) and vascular dementia (VaD). Structural brain MRI supports differentiating between these dementia types, but several factors complicate its interpretation in the diagnostic process, such as disease heterogeneity, disease overlap and mixed pathology. Unsupervised machine learning methods like clustering may provide hypothesis-free insights into the imaging patterns associated with these diseases. However, comprehensive analyses on datasets including a wide variety of dementia etiologies are lacking, while these are relevant to the real-life memory clinic situation. We aim to close this gap by clustering brain volumes across different etiologies, to gain more insight in dementia-related MRI patterns.

Methods:

We included 387 dementia patients with AD, FTD, primary progressive aphasia (PPA, language variants), DLB, VaD, or mixed dementia (AD + vascular pathology), who were seen at our memory clinic and underwent MRI imaging. We processed the T1-weighted MRI scans using FreeSurfer 6.0 to extract volumes of the brain lobes (frontal, temporal, parietal, occipital, insula), hippocampus, thalamus and amygdala for each hemisphere. These volumes were divided by intracranial volume to adjust for head size, and corrected for age and sex using a quantile regression model fitted on cognitively normal (CN) participants of the population-based Rotterdam Study (N=11728). The patient volumes were projected on the regression curves, yielding volume percentiles (VPs) that represent the degree of deviation with respect to the CN population at that age. Next, we employed hierarchical clustering with the 16 VPs as input, using Euclidian distance as the similarity metric and Wards linkage method. We performed two analyses, using the best and second-best number of clusters as determined by the Silhouette index (SI, measure of intra- and inter-cluster distance) across a range of 2 to 10 clusters. The resulting clusters were compared on age, sex, diagnosis and cognitive domain z-scores derived from neuropsychological testing.

Results:

A number of clusters of 2 and 4 was most optimal, although the SI values were still low (SI=0.21 and SI=0.18). The 2-cluster analysis (Fig. 1a) resulted in a cluster of patients with relatively spared brain volumes (cluster 2.1, N=194), cortical regions more spared than subcortical, and a second cluster of patients with small (sub)cortical brain volumes (cluster 2.2, N=193). The diagnostic classes generally spread out over both clusters, and cluster 2.2 performed significantly worse on cognitive testing in 4 out of 5 cognitive domains compared to cluster 2.1 (Fig. 2a). For the 4-cluster analysis, a similar global atrophy cluster was identified (cluster 4.1, N=193), while the other cluster was split into 3 subclusters (clusters 4.2-4.4). Cluster 4.2 (N=87) showed lower volumes in the frontal, parietal and occipital lobe, while cluster 4.3 (N=76) showed the opposite pattern with slightly asymmetric temporal-dominant atrophy (Fig. 1b). Cluster 4.4 (N=31) had spared brain volumes above the mean of the CN population, especially cortically. While the clusters were again a mix of different diagnoses, some diagnoses were more often present in specific clusters over others (Fig. 2b). Again, 4 out of 5 cognitive domains differed significantly between several clusters. There were no significant differences in age and sex, however.

Supporting Image: abstract_radar_figs_refitted.png

Supporting Image: abstract_diagn_cog_figs_refitted.png

Conclusions:

The results of this study underline and quantify that different dementia etiologies show overlapping atrophy patterns, and that single etiologies have different imaging manifestations. To improve etiologic dementia diagnosis, future research might focus on the development of disease-specific imaging biomarkers, and/or the use of more advanced methods like artificial intelligence to explore all information contained in the scans.

Disorders of the Nervous System:

Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) ¹

Modeling and Analysis Methods:

Multivariate Approaches ²

Segmentation and Parcellation

Novel Imaging Acquisition Methods:

Anatomical MRI

Keywords:

Cortex

Degenerative Disease

Machine Learning

Multivariate

Segmentation

STRUCTURAL MRI

Sub-Cortical

Other - Clustering

^1|2Indicates the priority used for review

Abstract Information

By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.

I accept

The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:

I do not want to participate in the reproducibility challenge.

Please indicate below if your study was a "resting state" or "task-activation” study.

Other

Healthy subjects only or patients (note that patient studies may also involve healthy subjects):

Patients

Was this research conducted in the United States?

Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.

Yes

Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.

Not applicable

Please indicate which methods were used in your research:

Structural MRI

Neuropsychological testing

Computational modeling

Provide references using APA citation style.

Not applicable

UNESCO Institute of Statistics and World Bank Waiver Form

I attest that I currently live, work, or study in a country on the UNESCO Institute of Statistics and World Bank List of Low and Middle Income Countries list provided.