Poster No:
98
Submission Type:
Abstract Submission
Authors:
Kyeongseob Song1, Choong-Hee Park1, Hyun Kook Lim2,3, Dae-Jin Kim4, Tae-Min Kim1,2, Ji-Won Chun1,2
Institutions:
1Department of Medical Informatics, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea, 2CMC Institute for Basic Medical Science, The Catholic Medical Center of The Catholic University, Seoul, Republic of Korea, 3Yeouido St. Mary's Hospital, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea, 4Seoul St. Mary's Hospital, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea
First Author:
Kyeongseob Song
Department of Medical Informatics, College of Medicine, The Catholic University of Korea
Seoul, Republic of Korea
Co-Author(s):
Choong-Hee Park
Department of Medical Informatics, College of Medicine, The Catholic University of Korea
Seoul, Republic of Korea
Hyun Kook Lim
CMC Institute for Basic Medical Science, The Catholic Medical Center of The Catholic University|Yeouido St. Mary's Hospital, College of Medicine, The Catholic University of Korea
Seoul, Republic of Korea|Seoul, Republic of Korea
Dae-Jin Kim
Seoul St. Mary's Hospital, College of Medicine, The Catholic University of Korea
Seoul, Republic of Korea
Tae-Min Kim
Department of Medical Informatics, College of Medicine, The Catholic University of Korea|CMC Institute for Basic Medical Science, The Catholic Medical Center of The Catholic University
Seoul, Republic of Korea|Seoul, Republic of Korea
Ji-Won Chun
Department of Medical Informatics, College of Medicine, The Catholic University of Korea|CMC Institute for Basic Medical Science, The Catholic Medical Center of The Catholic University
Seoul, Republic of Korea|Seoul, Republic of Korea
Introduction:
Alzheimer's disease (AD) is the most common cause of dementia, with mild cognitive impairment (MCI) serving as its prodromal stage, marked by subtle cognitive decline while daily function remains largely intact. MCI patients have an annual conversion rate to AD of 10–15%, compared to 1–2% in normal controls (NC). Identifying high-risk MCI patients for early intervention is crucial to delaying disease progression. AD progression is linked to amyloid plaques and neurofibrillary tangles, often detected through imaging and clinical assessments. However, some data modalities are costly or less accessible. Recent approaches have focused on structural MRI metrics to map brain networks. This study presents an explainable machine learning biomarker that integrates structural MRI measurements with cortical thickness-based network features to classify NC, early MCI, late MCI, and AD.
Methods:
This study included 4,213 participants from the dementia diagnosis medical imaging dataset of the AI Hub platform (https://aihub.or.kr/) provided by the NIA, comprising 1,170 NC, 1,218 early MCI, 1,132 late MCI, and 693 AD. T1-weighted brain structural MRI data were preprocessed using FreeSurfer, and cortical thickness, volume, and surface area values were extracted based on the Desikan-Killiany atlas. A morphological similarity network was constructed using 68 cortical regions as nodes, with connections determined by the Gaussian kernel-based similarity of cortical thickness. Smaller regional differences led to higher similarity values, while larger differences resulted in an exponential decrease. The weighted connections were binarized with sparsity levels from 5% to 50%, and network metrics such as nodal degree and nodal path length were calculated. Structural MRI and network features were used to train and test an SVM classifier, with the SHAP algorithm applied to interpret feature contributions (Gramegna & Giudici, 2021).
Results:
Using a Support Vector Machine (SVM) model, we compared the classification performance of combined structural and network features with structural features alone. Incorporating network features consistently improved AUC values across all tasks. For AD vs. NC, the AUC increased to 0.925 (structural only: 0.913). In early MCI vs. NC, it rose to 0.808 (vs. 0.794), and for late MCI vs. NC, it reached 0.830 (vs. 0.753). Similar improvements were seen in cognitively impaired groups: AD vs. early MCI achieved 0.879 (vs. 0.868), AD vs. late MCI reached 0.744 (vs. 0.734), and early MCI vs. late MCI improved to 0.799 (vs. 0.741). SHAP analysis showed that the entorhinal cortex, posterior cingulate cortex, and cuneus were common important features across all tasks. The caudal middle frontal cortex and lateral orbitofrontal cortex in AD vs. NC, the inferior parietal cortex and fusiform gyrus in early MCI vs. NC, the medial orbitofrontal cortex and superior frontal cortex in late MCI vs. NC, the rostral anterior cingulate cortex and inferior temporal cortex in AD vs. early MCI, the superior temporal cortex and postcentral gyrus in AD vs. late MCI, and the lingual gyrus and superior parietal cortex in early MCI vs. late MCI were group-specific important features.
Conclusions:
This study demonstrates that combining cortical thickness-based network features with structural MRI features enhances the classification of normal controls, early MCI, late MCI, and Alzheimer's disease. The inclusion of network-derived metrics highlights their added value in capturing subtle changes in brain connectivity, particularly for distinguishing disease stages. By applying explainable machine learning techniques, this study underscores the potential of integrating structural and network biomarkers to improve early diagnosis, predict disease progression, and support the development of advanced brain structure-based diagnostic tools for clinical practice.
Disorders of the Nervous System:
Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) 1
Modeling and Analysis Methods:
Classification and Predictive Modeling 2
Keywords:
Aging
Degenerative Disease
Machine Learning
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 am submitting this abstract as an original work to be reproduced. I am available to be the “source party” in an upcoming team and consent to have this work listed on the OSSIG website. I agree to be contacted by OSSIG regarding the challenge and may share data used in this abstract with another team.
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:
Structural MRI
For human MRI, what field strength scanner do you use?
3.0T
Which processing packages did you use for your study?
Free Surfer
Provide references using APA citation style.
Gramegna, A., & Giudici, P. (2021). SHAP and LIME: an evaluation of discriminative power in credit risk. Frontiers in Artificial Intelligence, 4, 752558.
Grundman, M. et al. (2004). Mild cognitive impairment can be distinguished from Alzheimer disease and normal aging for clinical trials. Archives of neurology, 61(1), 59-66.
Hänninen, T. et al. (2002). Prevalence of mild cognitive impairment: a population‐based study in elderly subjects. Acta Neurologica Scandinavica, 106(3), 148-154.
Sebenius, I. et al. (2024). Structural MRI of brain similarity networks. Nature Reviews Neuroscience, 1-18.
Zhang, T. et al. (2021). Predicting MCI to AD conversation using integrated sMRI and rs-fMRI: machine learning and graph theory approach. Frontiers in Aging Neuroscience, 13, 688926.
No