Development and Validation of a Machine Learning Classification Algorithm for Differentiating Frontotemporal Dementia
Seung Hyun Lee1, Wooseok Jung1, Mina Park1
1From the VUNO Inc. (S.H.L., W.J., H.C.), Seoul, Korea; UC Berkeley-UCSF Joint Graduate Program in Bioengineering (W.J.); Department of Radiology (M.P., H.S.O., B.J., S.J.A., S.H.S.), Neurology (H.C.C.H.L.), Gangnam Severance Hospital, Yonsei University College of Medicine, Seoul, Korea and Osstem Implant Co., Ltd., 3 (H.C.), Magokjungang 12-ro, Gangseo-gu, Seoul, Korea.
AJNR. American Journal of Neuroradiology
|July 24, 2026
Summary
This study developed an AI algorithm for brain volumetry to differentiate frontotemporal dementia (FTD) from Alzheimer's disease (AD) and normal aging. The AI tool showed high accuracy in distinguishing these conditions, aiding clinical diagnosis.
Area of Science:
- Neuroimaging
- Artificial Intelligence in Medicine
- Neurology
Background:
- Frontotemporal dementia (FTD) diagnosis is challenging due to overlapping symptoms and imaging findings with Alzheimer's disease (AD).
- Accurate differentiation is crucial for appropriate patient management and treatment.
Purpose of the Study:
- To develop and validate an automated brain volumetry classification algorithm.
- To differentiate FTD from AD and cognitively normal (CN) individuals using MRI data.
- To assess the algorithm's clinical utility through external validation and a radiologist reader study.
Main Methods:
- Utilized deep learning for automated brain volumetry from 3D T1-weighted MRI scans.
- Incorporated regional volumes, asymmetry indices, and MMSE scores into an XGBoost classifier.
- Validated the algorithm on internal (NIFD, ADNI) and external (tertiary hospital) cohorts, including a reader study.
Main Results:
- The algorithm achieved high accuracy (91.40% internal, 88.80% external validation).
- Demonstrated strong performance in differentiating FTD, AD, and CN groups.
- Assisted radiologists by significantly reducing reading time while maintaining diagnostic accuracy.
Conclusions:
- Automated brain volumetry with asymmetry features shows promise for differentiating FTD from AD and normal aging.
- The validated algorithm can serve as a valuable adjunct tool in clinical settings.
- The AI model supports accurate and efficient differential diagnosis in neurodegenerative diseases.
