Clinically guided adaptive contrast adjustment for fetal plane classification: a modular plug-and-play solution.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Biomedical Engineering

Background:

• Fetal ultrasound standard plane recognition is crucial for prenatal assessment but faces challenges like poor contrast and variable image quality.
• Current methods struggle with indistinct anatomical boundaries and operator-dependent image variations.

Purpose of the Study:

• To introduce a novel Adaptive Contrast Adjustment Module (ACAM) for enhancing fetal ultrasound images.
• To improve the accuracy and robustness of fetal ultrasound standard plane recognition.

Main Methods:

• Developed a plug-and-play ACAM with a lightweight, texture-aware subnetwork to learn contrast parameters.
• Generated multiple contrast-enhanced image representations using differentiable transformations.
• Fused enhanced views within classifiers to enrich discriminative features, mimicking clinical contrast adjustment.

Main Results:

• Demonstrated consistent accuracy gains across various model architectures on a dataset of 12,400 fetal ultrasound images.
• Achieved performance improvements of 2.02% for lightweight models, 1.29% for conventional architectures, and 1.15% for state-of-the-art models.
• Showcased ACAM's ability to enhance robustness against image quality variations through multi-view contrast fusion.

Conclusions:

• ACAM offers content-adaptive, clinically aligned contrast modulation, improving fetal ultrasound analysis.
• The module effectively links low-level texture cues with high-level semantic understanding.
• This approach provides a new framework for medical image analysis in realistic clinical settings.