Researchers have developed a new method for segmenting kidney ultrasound (US) images that demonstrated an increased efficiency and accuracy compared to traditional methods.
“Automatic segmentation of kidneys in [two-dimensional] 2D US images remains a challenging task due to high speckle noise, low contrast between foreground and background, weak boundaries and large appearance variations of kidneys in 2D US images,” wrote corresponding author Yong Fan, PhD, with the department of radiology at the University of Pennsylvania’s Perelman School of Medicine, and colleagues. “Automatic segmentation of US images of kidneys will facilitate extraction and quantification of anatomic features such as renal parenchymal area and kidney echogenicity, which currently are measured manually.”
In this experiment, published online Feb. 12 in Academic Radiology, scientists developed a graph cuts-based method to segment kidney images through combining original image intensity information and texture feature maps extracted with Gabor filters.
The method was applied to 85 kidney ultrasound images, and 20 randomly selected sets of imaging data were taken to fine tune the parameters of the new segmentation method. The remaining data was used in validation testing.
Authors compared their technique to four commonly used imaging segmentation methods: Localizing region-based active contours, geodesic active contours and multi-feature segmentation model. They noted the data demonstrated “promising” segmentation results for bilateral kidneys.
The results were as follows:
- Average Dice index of 0.9446.
- Average mean distance of 2.2551.
- Average specificity of 0.9971.
- Average accuracy of 0.9919.
“Our method can be integrated with US image interpretation systems to automatically compute anatomic measures of kidneys that could be informative for predicting risk of end-stage renal disease,” wrote Fan et al.