Grayscale medical image segmentation method based on 2D&3D object detection with deep learning.

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Artificial Intelligence in Medicine

Background:

• Grayscale medical image segmentation is crucial for computer-aided diagnosis.
• Traditional model-driven methods struggle with varying intensity distributions, requiring pre-processing.
• Deep learning methods offer accurate feature extraction but demand extensive training data and complex architectures.

Purpose of the Study:

• To develop a novel, accurate, and efficient method for grayscale medical image segmentation.
• To combine thresholding techniques with deep learning using object detection for improved segmentation.
• To reduce the need for extensive pre-processing and large training datasets.

Main Methods:

• A hybrid approach integrating thresholding and deep learning via 2D and 3D object detection.
• Utilizing a fine-tuned 2D object detection network to identify regions of interest.
• Converting image pixels into point clouds for 3D object detection to determine segmentation thresholds.

Main Results:

• Achieved high segmentation accuracy with IoU (DSC) scores of 0.92 (0.96), 0.88 (0.94), and 0.94 (0.94) across different datasets.
• Demonstrated superior performance compared to five state-of-the-art and clinically used models.
• Validated the method's effectiveness on diverse grayscale medical image datasets.

Conclusions:

• The proposed method, leveraging 2D&3D object detection with deep learning, is a workable and promising solution for grayscale medical image segmentation.
• The approach offers a robust alternative to traditional methods, enhancing diagnostic capabilities.
• This technique shows potential for advancing computer-aided diagnosis in medical imaging.