Mammographic texture analysis differentiates between benign, malignant tumors
Mammographic texture analysis can successfully differentiate between benign and malignant breast tumors, according to research out of China, circumventing some of the shortfalls of digital mammography while helping radiologists make more accurate diagnoses.
Digital mammography (DM)’s reach is vast, and it’s the universally recommended imaging modality for detecting breast cancer, corresponding author Xiaoming Zhou, MD, and colleagues in the radiology department at the Affiliated Hospital of Qingdao University in Qingdao, China, wrote in Clinical Breast Cancer this month. But it has its disadvantages, like the risk of high false-positive rates in dense-breast populations and the inevitability of overdiagnosis of lesions on DM images.
“With regard to imaging features of digital mammography, there are some overlaps between benign and malignant breast tumors,” the authors wrote. “[And] no quantitative parameter can be obtained from the analysis of DM images.”
Zhou and colleagues proposed texture analysis—a process during which mathematical methods are used to describe the relationship between pixel intensities in an image—as a way to enhance digital mammography and aid physicians in making more precise diagnoses.
“Because of the diverse components of benign and malignant breast tumors, different densities of breast tumors can be reflected by pixel intensities of lesions within a mammographic image,” the authors said. “Histogram analysis can be used to describe the distribution of pixel intensities within an image.”
In their study, the team tried to validate the diagnostic performance of texture analysis while also assessing the utility of histogram, gray-level co-occurrence matrix (GLCM) and run length matrix (RLM) in evaluating breast tumors. They used a database of 302 mammograms pulled from their hospital’s PACS and employed receiver operating characteristic curve analysis to determine the performance of texture features.
The trial was successful, according to the paper, and proved texture analysis was “an efficacious method” that found significant differences between benign and malignant tumors. Nineteen texture features of histogram, five features of GLCM and eleven of RLM were competent in the differential diagnosis of tumors, and, when combined with imaging-based diagnosis, texture analysis saw further improved results.
Zhou et al. said further work is required to account for some of their study’s limitations, including selection bias and the exclusion of MR images or an automated segmentation technique.
“We have successfully used the texture analysis in DM images and concluded that texture analysis was a reliable technique with excellent diagnostic performance,” they wrote. “Texture analysis can help radiologists and clinicians to make precise diagnoses, and unnecessary biopsy or therapy can be avoided.”
After graduating from Indiana University-Bloomington with a bachelor’s in journalism, Anicka joined TriMed’s Chicago team in 2017 covering cardiology. Close to her heart is long-form journalism, Pilot G-2 pens, dark chocolate and her dog Harper Lee.