Deep learning and Gaussian Mixture Modelling clustering mix. A new approach for fetal morphology view plane...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Obstetrics and Gynecology

Background:

• The COVID-19 pandemic has significantly impacted medical practice, particularly in obstetrics and gynecology.
• Maternal-fetal monitoring is crucial for preventing pregnancy complications and mortality.
• Artificial Intelligence (AI) offers potential for faster, more accurate diagnoses in healthcare.

Purpose of the Study:

• To propose a novel framework merging deep learning algorithms and Gaussian Mixture Modelling (GMM) clustering.
• To differentiate between view planes in second-trimester fetal morphology scans.
• To enhance diagnostic accuracy in obstetric ultrasound through AI.

Main Methods:

• Utilized deep learning models: ResNet50, DenseNet121, InceptionV3, EfficientNetV2S, MobileNetV3Large, and Xception.
• Developed a framework integrating these models with GMM clustering for view plane differentiation.
• Employed a statistical fitness function and synergetic weighted voting for final decision-making.

Main Results:

• The proposed framework was tested on two second-trimester morphology scan datasets.
• Statistical benchmarking validated the framework's performance.
• The synergetic weighted vote of the framework demonstrated superior performance compared to individual deep learning networks and other voting strategies.

Conclusions:

• The developed AI framework effectively differentiates fetal ultrasound view planes.
• This approach shows promise for improving the accuracy and efficiency of obstetric diagnoses.
• The integration of deep learning and GMM offers a powerful tool for maternal-fetal monitoring.