Eigenhearts: Cardiac diseases classification using eigenfaces approach.

Area of Science:

• Cardiovascular Medicine
• Medical Imaging
• Data Science

Background:

• Medical imaging is vital for diagnosing cardiac diseases.
• Data science integration faces challenges due to limited image acquisition (ethics, cost, protocols).
• Novel methods are needed to overcome data limitations in cardiac imaging analysis.

Purpose of the Study:

• To evaluate the eigenfaces approach for classifying cardiac diseases in medical imaging.
• To enhance cardiac disease classification accuracy by integrating eigenfaces with convolutional neural networks (CNNs).
• To address data acquisition challenges in cardiovascular research.

Main Methods:

• Applied the eigenfaces approach, derived from principal component analysis, to echocardiography images.
• Integrated eigenfaces with CNNs for image classification.
• Utilized singular value decomposition (SVD) for pre-processing medical imaging data.

Main Results:

• The eigenfaces approach demonstrated applicability to complex medical imaging datasets.
• Integration with CNNs enabled classification of five distinct cardiac conditions in mice.
• Pre-processing using SVD led to a substantial ~50% increase in classification accuracy.

Conclusions:

• The eigenfaces approach, enhanced by SVD pre-processing and CNN integration, offers a promising tool for cardiac disease classification.
• This method effectively improves diagnostic accuracy in echocardiography, addressing data limitations.
• The study highlights the potential of established face recognition techniques in advanced medical diagnostics.