Epoch and accuracy based empirical study for cardiac MRI segmentation using deep learning technique.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Cardiovascular Disease Diagnosis

Background:

• Cardiac magnetic resonance imaging (CMRI) is crucial for diagnosing cardiovascular diseases, offering clear, non-invasive views of the heart.
• Accurate segmentation of CMRI images is vital for quantifying parameters like ejection fraction and myocardial viability.
• Manual segmentation is time-consuming and subject to inter-observer variability.

Purpose of the Study:

• To develop and optimize a convolution neural network (CNN) model for automated CMRI image segmentation.
• To enhance the accuracy and efficiency of CMRI analysis compared to manual methods.
• To establish the relationship between the hyperparameter 'epoch' and model accuracy for optimal performance.

Main Methods:

• Utilized a CNN model for the segmentation of cardiac magnetic resonance imaging data.
• Optimized neural network parameters, including epochs, for a novel dataset to ensure accurate predictions.
• Evaluated model performance based on the Intersection over Union (IoU) coefficient.

Main Results:

• The optimized CNN model achieved a segmentation accuracy of 0.88 in terms of the IoU coefficient.
• Demonstrated the effectiveness of CNNs in overcoming limitations of manual segmentation.
• Established a clear relationship between the number of training epochs and predictive accuracy.

Conclusions:

• CNN-based segmentation offers a more accurate and efficient approach for CMRI analysis.
• The study highlights the importance of hyperparameter optimization, particularly epochs, for robust model performance.
• Automated segmentation using CNNs has the potential to significantly improve cardiovascular disease diagnosis and management.