Classification of breast microcalcifications using dual-energy mammography.
Bahaa GhammraouiAndrey MakeevAhmed ZidanAlaadin AlayoubiStephen J GlickPublished in: Journal of medical imaging (Bellingham, Wash.) (2019)
The potential of dual-energy mammography for microcalcification classification was investigated with simulation and phantom studies. Classification of type I/II calcifications was performed using the tissue attenuation ratio as a performance metric. The simulation and phantom studies were carried out using breast phantoms of 50% fibroglandular and 50% adipose tissue composition and thicknessess ranging from 3 to 6 cm. The phantoms included models of microcalcifications ranging in size between 200 and 900 μ m . The simulation study was carried out with fixed MGD of 1.5 mGy using various low- and high-kVp spectra, aluminum filtration thicknesses, and exposure distribution ratios to predict an optimized imaging protocol for the phantom study. Attenuation ratio values were calculated for microcalcification signals of different types at two different voltage settings. ROC analysis showed that classification performance as indicated by the area under the ROC curve was always greater than 0.95 for 1.5 mGy deposited mean glandular dose. This study provides encouraging first results in classifying malignant and benign microcalcifications based solely on dual-energy mammography images.
Keyphrases
- dual energy
- image quality
- computed tomography
- contrast enhanced
- deep learning
- machine learning
- adipose tissue
- magnetic resonance imaging
- randomized controlled trial
- high resolution
- type diabetes
- mass spectrometry
- metabolic syndrome
- high fat diet
- case control
- skeletal muscle
- single molecule
- molecular dynamics
- virtual reality