Comparative Analysis of Edge Detection Operators Using a Threshold Estimation Approach on Medical Noisy Images with...

Area of Science:

• Medical Imaging
• Image Processing
• Computer Vision

Background:

• Noise significantly degrades medical image quality, potentially leading to misdiagnoses.
• Accurate edge detection is crucial for analyzing critical features in medical scans.
• Existing edge detection methods struggle with noise in complex medical images.

Purpose of the Study:

• To propose and evaluate a new threshold value estimation approach for robust edge detection in noisy medical images.
• To assess the impact of noise on edge detection across different image complexities and medical datasets.
• To compare the proposed method with traditional and deep learning-based edge detection techniques.

Main Methods:

• A novel threshold value estimation method was developed and applied to edge detection algorithms.
• The approach was tested on medical images (retinal, brain tumor, lung CT) with varying noise types and levels.
• Grid search (GS) and random search (RS9) optimization methods were utilized.
• Performance was evaluated using metrics on images with low, medium, and high detail levels.
• Comparisons were made against deep learning models like AlexNet, ResNet, VGGNet, MobileNetv2, and Inceptionv3.

Main Results:

• The proposed threshold estimation approach significantly improved edge detection performance, particularly with the Canny operator, across diverse noise conditions.
• Laplace operators, highly sensitive to noise, showed marked improvement with the new method, especially using grid search.
• The approach demonstrated effectiveness on complex medical images from retinal, brain tumor, and lung CT datasets.
• Comparative analysis showed competitive or superior performance against leading deep learning edge detection techniques.

Conclusions:

• The proposed threshold value estimation method offers a robust solution for enhancing edge detection in noisy medical images.
• This technique is vital for improving the reliability of medical image analysis and reducing diagnostic errors.
• The study highlights the importance of noise-robust algorithms in clinical applications of medical imaging.