Robust automated computational approach for classifying frontotemporal neurodegeneration: Multimodal/multicenter...

Area of Science:

• Neuroscience
• Medical Imaging
• Artificial Intelligence

Background:

• Behavioral variant frontotemporal dementia (bvFTD) diagnosis is challenging due to reliance on clinical expertise and ambiguous guidelines.
• Multimodal neuroimaging and machine learning are recommended as complementary diagnostic tools.

Purpose of the Study:

• To develop an automated, cross-center, multimodal computational approach for robust classification of bvFTD patients and healthy controls.

Main Methods:

• Analysis of structural MRI and resting-state functional connectivity data from 44 bvFTD patients and 60 healthy controls across three imaging centers.
• Utilized a fully automated processing pipeline including site normalization, native space feature extraction, and a random forest classifier.

Main Results:

• The developed method achieved high classification accuracy (91%), sensitivity (83.7%), and specificity (96.6%).
• Successfully combined multimodal imaging data for improved diagnostic performance.

Conclusions:

• The multimodal approach enhances system performance and offers a clinically informative neuroimaging analysis method.
• Combining multimodal imaging and machine learning represents a gold standard for dementia diagnosis.