Content-based image retrieval for the diagnosis of myocardial perfusion imaging using a deep convolutional...

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Single-photon emission computed tomography (SPECT) myocardial perfusion imaging (MPI) is vital for managing coronary heart disease.
• Deep learning, specifically convolutional autoencoder (CAE) models, were explored for feature extraction from MPI.

Purpose of the Study:

• To assess the feasibility of using a CAE model for feature extraction from SPECT MPI.
• To evaluate the utility of these extracted features for content-based image retrieval (CBIR).

Main Methods:

• A CAE model was trained on 843 pairs of stress/rest MPI scans.
• The encoder outputted 256-dimensional feature vectors, reduced using PCA for visualization.
• CBIR was performed using cosine similarity of feature vectors, with radiologist findings used for evaluation.

Main Results:

• PCA visualization showed feature vectors captured clinical data like ischemia and scar location.
• The CBIR system achieved 81.0% binary accuracy as a similarity-based diagnostic tool.
• Extracted features successfully retained clinically relevant information.

Conclusions:

• Unsupervised feature learning using CAE is effective for CBIR in MPI.
• This approach shows promise for enhancing diagnostic capabilities in cardiology.
• The study highlights the potential of AI in medical image analysis for cardiovascular diseases.