Neural network-based fully automated cardiac resting phase detection algorithm compared with manual detection in...

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Medical Physics

Background:

• Free-breathing cardiac magnetic resonance (CMR) examinations utilize a cardiac resting phase.
• Accurate identification of this phase is crucial for image quality and diagnostic interpretation.

Purpose of the Study:

• To evaluate the clinical performance of a novel neural network-based system for cardiac resting phase detection.
• To compare the accuracy and efficiency of the neural network system against expert and general operator manual detection.

Main Methods:

• Four-chamber cine images from 32 patients were analyzed.
• Comparisons were made between expert, general operator, and neural network determined resting phase parameters (duration, start, end).
• Normalized root-mean-square error (RMSE) was calculated to quantify differences.

Main Results:

• The neural network detected the resting phase almost instantaneously.
• No significant differences were found in rest duration and start point between the neural network and expert detection (p = .30, .90).
• The neural network demonstrated lower normalized RMSE values for all parameters compared to general operators.

Conclusions:

• The neural network system provides instant cardiac resting phase detection.
• This AI-driven approach offers superior accuracy compared to manual detection by general operators in clinical practice.