Poster No:
86
Submission Type:
Abstract Submission
Authors:
Elena Doering1, Merle Hoenig1, Gérard Bischof2, Thilo van Eimeren2, Alexander Drzezga2
Institutions:
1Research Center Juelich, Juelich, North-Rhine Westphalia, 2University Hospital Cologne, Cologne, North-Rhine Westphalia
First Author:
Elena Doering
Research Center Juelich
Juelich, North-Rhine Westphalia
Co-Author(s):
Merle Hoenig
Research Center Juelich
Juelich, North-Rhine Westphalia
Gérard Bischof
University Hospital Cologne
Cologne, North-Rhine Westphalia
Introduction:
Accurate and early diagnosis of Alzheimer's disease (AD) is crucial for enabling timely interventions, with biomarkers serving as a cornerstone for reliable detection. The biological diagnosis of AD is grounded in the assessment of its hallmark pathologies: beta-amyloid deposition (A), tau accumulation (T), and neurodegeneration (N), collectively known as the A/T/N framework. The location and extent of A and T are most accurately detected using positron emission tomography (PET). However, PET imaging is sometimes unfeasible due to radiation exposure, prolonged tracer uptake time, and limited accessibility in certain clinical and research settings. MRI, used for assessing N, is more widely available, but it does not allow for the direct quantification of amyloid and tau pathology. Given the causal links between amyloid and tau pathology and neurodegeneration, artificial intelligence methods might uncover patterns within volumetric MRI scans that allow to infer the spatial distribution of amyloid and tau. In this study, we developed a convolutional neural network (CNN) to predict amyloid and tau PET scans from structural MRI ("MRI-to-PET translation"). Additionally, we investigated whether incorporating additional patient information improves translation accuracy.
Methods:
We obtained 1663 MRI/amyloid PET (18F-AV45 tracer) and 728 MRI/tau PET (18F-AV1451 tracer) scan pairs with sufficient quality (time interval≤90 days; mean age=75 years; both datasets comprised cognitively normal and cognitively impaired subjects) from the multi-centric Alzheimer's Disease Neuroimaging Initiative. MRI and PET scans were spatially co-registered and normalized, and PET scans were smoothed to eliminate center-specific noise patterns. Subsequently, 75% and 25% of scan pairs were allocated to the training and testing data, respectively. For the translation task, we implemented an image generation CNN with encoding-decoding structure. Importantly, the latent part of our network consisted of flat layers, where additional patient features (age, sex, education, APOE genotype and cognitive scores [MMSE]) were concatenated. N=250 scan pairs from the training data were used to select the optimal set of patient features by iteratively adding features/feature combinations that enhance similarity between ground truth and predicted scans compared to using fewer features. Finally, the CNN was trained using all training data both with and without the optimal set of patient features. The similarity between generated and ground truth scans was quantified on the test set using standard image-translation metrics (mean absolute error [MAE; 0 is best] and structural similarity index [SSIM; 1 is best]).
Results:
Without adding patient information, generated amyloid and tau scans already closely matched ground truth scans (MAE: 0.084 (A)/0.063 (T); SSIM: 0.936 (A)/0.880 (T)). Age and APOE genotype emerged as the most informative additional features to translate MRI into amyloid PET scans, while age and MMSE scores were selected to inform the translation of MRI into tau PET scans. Adding informative patient features improved translation accuracy in both modalities (MAE: 0.072 (A)/0.056 (T); SSIM: 0.942 (A)/0.914 (T)). Further validation of regional similarity between ground truth and generated scans, along with assessments of clinical utility, is ongoing.
Conclusions:
This study leverages the structural information captured by MRI as well as available patient information to estimate AD pathological patterns typically visualized by PET. With MRI data being routinely assessed in the diagnostic workup of dementia patients, the current approach may serve as a scalable tool for localizing and tracking the progression of AD pathology when standard PET is inaccessible, or for evaluating the necessity of subsequent standard PET scan assessment.
Disorders of the Nervous System:
Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) 1
Modeling and Analysis Methods:
Methods Development 2
Multivariate Approaches
PET Modeling and Analysis
Neuroinformatics and Data Sharing:
Informatics Other
Keywords:
Machine Learning
Positron Emission Tomography (PET)
STRUCTURAL MRI
1|2Indicates the priority used for review
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?
No
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:
PET
Structural MRI
For human MRI, what field strength scanner do you use?
1.5T
3.0T
Which processing packages did you use for your study?
SPM
Provide references using APA citation style.
None
No