Transfer learning on T1-weighted images for brain age estimation.

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Gerontology

Background:

• Alzheimer's disease (AD) is a leading cause of death in the elderly, characterized by its insidious onset and irreversible progression.
• Early diagnosis of AD is crucial for effective patient management and treatment strategies.
• Radiological assessment of brain age compared to chronological age is a promising biomarker for preliminary AD analysis.

Purpose of the Study:

• To develop and validate a transfer learning-based method for predicting brain age using MRI data.
• To investigate the efficacy of using multiple MRI planes for age prediction.
• To identify brain regions critical for age estimation in both healthy individuals and AD patients.

Main Methods:

• A transfer learning approach was employed, involving freezing and fine-tuning different layers of a deep learning model.
• Three orthogonal planes (axial, coronal, sagittal) of brain MRI data were utilized for age prediction.
• Deep learning techniques were applied to identify salient brain regions contributing to age estimation.

Main Results:

• The proposed transfer learning method achieved a mean absolute error of 0.6 years lower than state-of-the-art methods.
• Sagittal plane MRI images demonstrated superior performance in age estimation compared to axial and coronal planes.
• The study identified the frontal lobe as a key region for age estimation in AD patients, aligning with existing medical findings.

Conclusions:

• The developed method offers a more accurate and potentially earlier approach to AD detection through brain age prediction.
• Utilizing multiple MRI planes, particularly the sagittal view, enhances the precision of age estimation.
• Identifying specific brain regions involved in age estimation provides valuable insights into the neuropathology of Alzheimer's disease.