Right Ventricular Segmentation from MRI Using Deep Convolutional Neural Networks.

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Right ventricular (RV) function assessment is vital for diagnosing cardiac diseases.
• Manual analysis of RV function from cardiac MRI is time-consuming and labor-intensive.
• Automating RV segmentation from MRI data is highly desirable for clinical efficiency.

Purpose of the Study:

• To develop and evaluate a novel machine learning approach for automated RV segmentation from cardiac MRI sequences.
• To compare the performance of the proposed method against the U-Net architecture and manual delineations.

Main Methods:

• A convolutional neural network (CNN)-based approach was designed, featuring a unique architecture distinct from U-Net.
• The method incorporates image concatenation to leverage 3D spatial information during the segmentation process.
• Quantitative evaluations were conducted on 256 MRI images from 16 patients, comparing automated results with manual segmentations.

Main Results:

• The proposed CNN model demonstrated superior performance in RV delineation compared to the U-Net approach.
• The method effectively utilizes 3D spatial information for improved segmentation accuracy.
• Quantitative comparisons confirmed the outperformance of the proposed method over the state-of-the-art.

Conclusions:

• The developed deep learning model offers an effective and accurate solution for automated RV segmentation from cardiac MRI.
• This automation has the potential to significantly streamline the assessment of RV function in clinical practice.
• The novel architecture and 3D spatial information utilization represent advancements in cardiac image analysis.