RNN-AHF Framework: Enhancing Multi-focal Nature of Hypoxic Ischemic Encephalopathy Lesion Region in MRI Image Using...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Neonatal Neurology

Background:

• Hypoxic-Ischemic Encephalopathy (HIE) lesion enhancement in neonatal brain MRIs is difficult due to diffuse nature, small size, and low contrast.
• Lesion classification is hindered by unclear boundaries and edges caused by artifacts and signal mixing.
• Existing enhancement algorithms struggle with HIE lesions due to artifacts and their diffuse characteristics.

Purpose of the Study:

• To propose a novel framework for enhancing Hypoxic-Ischemic Encephalopathy (HIE) lesion regions in neonatal brain MR images.
• To improve the accuracy of HIE lesion classification through enhanced image quality.

Main Methods:

• Development of a Rough Neural Network and Anti-Homomorphic Filter (RNN-AHF) framework.
• The framework reduces pixel dimensionality, removes irrelevant pixels, and retains essential ones for enhancement.
• Utilizes optimized neural weights and a training function for adaptive enhancement.

Main Results:

• The RNN-AHF framework effectively reduces feature space dimensionality and optimizes pixel selection.
• It achieves adaptive enhancement by learning pixel patterns with differential neural network weights.
• High contrast enhancement of the HIE lesion region is achieved while preserving boundaries and edges.

Conclusions:

• The proposed RNN-AHF framework significantly enhances HIE lesion visualization.
• The framework achieves approximately 93.5% accuracy in lesion image enhancement and classification.
• This performance surpasses traditional algorithms for HIE lesion analysis.