(주) 뷰노

(주) 뷰노

문의사항 남기기
KR EN
VUNO MED INC.
Publications
BACK TO LIST
ISMRM 2025

Diagnosis of Frontotemporal Dementia on Brain MR Images by Using Automated Brain Volumetry

  • May. 2025

Background
Frontotemporal dementia (FTD) is a neurodegenerative disorder affecting the frontal and temporal lobes. MRI is essential for detecting atrophy inthese areas, but diagnosis can be challenging due to variable atrophy patterns that may overlap with conditions like Alzheimer's disease (AD).


Purpose
This study aimed to evaluate the effectiveness of a deep learning-based automated brain volumetry tool for the clinical diagnosis of FTD using MRI,and to differentiate FTD from AD and cognitively normal (CN) individuals.


Material and Methods
This IRB-approved, retrospective study included 759 subjects divided into three groups: FTD, AD, and CN (cognitively normal). Data were sourcedfrom the NIFD database (for FTD) and the ADNI database (for AD and CN). The dataset was randomly divided into a training set (n=609) and aninternal validation set (n=150). External validation was performed using an additional set of 142 subjects from Gangnam Severance Hospital (for allgroups). All subjects met diagnostic criteria for FTD and AD, while the CN group had no neurological disease. Subjects underwent 3D T1-weightedMRI on 3T scanners, with available MMSE scores. Automated brain volumetry was conducted using VUNO Med® DeepBrain® software, whichsegmented the brain into 104 regions. To account for individual differences in brain size, volumetric data were normalized to intracranial volume.Feature selection was performed using a one-sided T-test to identify regions with significant volume differences between FTD and AD. A linearSupport Vector Machine (SVM) classifier was trained using these volumetric features, along with age and MMSE scores, to differentiate FTD from ADand CN. Classifier performance was evaluated based on accuracy, precision, and recall metrics.


Results
Feature selection identified significant volumetric differences between FTD and AD, particularly in the frontal lobes, insula, cingulate cortex, andsubcortical gray matter. These regions, along with age and MMSE scores, were used as key input variables for the classification model. Using thetraining dataset, the classifier’s performance showed an accuracy of 89.3%, with precision of 90.7% and recall of 86.7%. The internal validation setdemonstrated an accuracy of 92.0%, with precision of 88.9% and recall of 88.5%. External validation using the independent dataset yielded anaccuracy of 84.6%, with precision of 77.2% and recall of 72.3%. Our automated brain volumetry model demonstrated promising diagnostic accuracyfor FTD, offering a potential tool for differentiating FTD from AD and normal aging in clinical settings.


Conclusion
Our automated brain volumetry model demonstrated promising diagnostic accuracy for FTD, offering a potential tool for differentiating FTD fromAD and normal aging in clinical settings.

Link

-

Author

Seung Hyun Lee, Wooseok Jung, Hyeonwoo Cho, Dong-Hee Kim, and Mina Park

Tag in

#medical_image

#VUNO Med®-DeepBrain®

이전글
Discriminative Feature Learning for Lacune Detection in 2D T2-FLAIR Images using Supervised Contrastive Learning 2025. 05. 15
다음글
Novel Artificial Intelligence Model Using Electrocardiogram for Detecting Acute Myocardial Infarction Needing Revascularization 2025. 05. 15
List

이 웹사이트는 원활한 서비스 제공을 위하여 당사 개인정보 처리방침에 명시된 쿠키를 사용합니다.

만일 당사의 쿠키 사용 정책에 대해 더 알고 싶으신 경우에는 개인정보 처리방침 을 참고하여 주시기 바랍니다.