A novel matrix-similarity based loss function for joint regression and classification in AD diagnosis

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Alzheimer's Disease (AD) and Mild Cognitive Impairment (MCI) diagnosis involves identifying brain disease and predicting clinical scores.
• Previous methods often treated disease classification and clinical score prediction separately.
• High dimensionality and small sample sizes are challenges in AD/MCI diagnosis.

Purpose of the Study:

• To develop a novel method for joint regression and classification in AD/MCI diagnosis.
• To propose a matrix-similarity based loss function for improved feature selection.
• To enhance the performance of both clinical score prediction and disease status identification.

Main Methods:

• Proposed a novel matrix-similarity based loss function for joint regression and classification.
• Combined the new loss function with a group lasso method for joint feature selection.
• Validated the method on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset.

Main Results:

• The proposed matrix-similarity loss function enhanced performance in both clinical score prediction and disease status identification.
• The joint feature selection approach outperformed state-of-the-art methods.
• Demonstrated the effectiveness of the novel loss function on the ADNI dataset.

Conclusions:

• The novel matrix-similarity based loss function improves AD/MCI diagnosis by jointly addressing regression and classification tasks.
• Joint feature selection across tasks is beneficial for AD/MCI diagnosis.
• The proposed method offers a promising advancement in the diagnosis of neurodegenerative diseases.