Software reliability and algorithm validation for medical imaging: performance of common edge detection methods in...

Area of Science:

• Medical Imaging
• Image Processing
• Nuclear Medicine

Background:

• Nuclear imaging often presents challenges due to low information density and spatial resolution.
• Accurate edge detection is crucial for quantitative analysis in nuclear medicine.

Purpose of the Study:

• To evaluate the performance of common edge detection algorithms and a novel nearest-neighbor algorithm (NNA) in nuclear images.
• To assess algorithm robustness under conditions of low information density and spatial resolution.

Main Methods:

• Evaluated ten standard edge detection algorithms and the NNA using phantoms and clinical nuclear images.
• Analyzed performance with isolated and overlapping regions, variable backgrounds, and adaptive thresholding.
• Quantified performance based on area determination accuracy, ROC operating points, and shape preservation.

Main Results:

• Several algorithms performed well at high information density, but struggled with low information density.
• The NNA demonstrated consistent performance across both low and high information density conditions.
• Adaptive thresholding improved performance for some methods at high information density.

Conclusions:

• The nearest-neighbor algorithm (NNA) is a robust and accurate method for edge detection in nuclear imaging, particularly under challenging low information density conditions.
• NNA offers superior consistency compared to traditional algorithms, enhancing quantitative analysis reliability.