MWSegEval—An image analysis toolbox for microwave breast images

Abstract

Interpreting reconstructed medical images and assessing their quality is a challenging problem that demands a consistent framework for extracting quantitative information from the images. The challenges are more pronounced for microwave images, as the reconstructions formed with this modality typically have spurious artifacts and the interfaces that delineate tissue types may be blurred or reconstructed in an incorrect location. Moreover, a significant degree of inhomogeneity within one tissue type is common, as well as substantial differences in the values of the electrical properties reconstructed with variants of an algorithm. MWSegEval is an image analysis toolbox specifically developed to help alleviate this multitude of complications. Namely, the toolbox provides methods to automatically segment images into regions of interest corresponding to various tissue types The segmentation leads to the decomposition of the breast interior into disjoint tissue masks. An effective array of region and distance-based metrics are applied to compare masks extracted from reconstructed images and ground truth models. The quantitative results reveal the accuracy with which the geometric and dielectric properties are reconstructed, and are supplemented with qualitative information. Consequently, the toolbox provides a framework that effectively furnishes quantitative and qualitative assessment of regions that contain a specific tissue. The resulting information facilitates comparisons that provide valuable insight into complex issues that impact image quality.

Keywords

• Microwave medical imaging
• Image reconstruction
• Segmentation
• Unsupervised machine learning, Performance metrics