Premium
Sci—Fri AM: Imaging — 05: Principal Component X‐Ray Simulation Analysis of Breast Biopsy Phantoms
Author(s) -
Tang R,
Georgeoff JB,
LeClair RJ
Publication year - 2010
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.3476184
Subject(s) - biopsy , detector , principal component analysis , nuclear medicine , optics , materials science , physics , medicine , mathematics , radiology , statistics
Multivariate principal component analysis (PCA) was performed on x‐ray simulations of breast biopsy phantoms. Semianalytic models were used to generate scatter (N s ) and photon transmission (N) signals from mixtures of fibrous (i), cancerous (j), and adipose tissue (k). A 5mm diameter 30 kV Mo beam is incident on a 5 mm thick biopsy. The incident exposure is 5.08 × 10 −4 C/kg. An energy discriminating photon counting detector is assumed (e.g. CZT). For simulation of scatter measurements, the detector is placed at a scatter angle θ=10°, whereas for transmission θ=0°. The distance from the biopsy center to a 5 mm diameter aperture above the detector is 15 cm. The variables used in the PCA are N s /N 0 (16 keV), N s /N 0 (25 keV) and N/N 0 (8 keV) where N 0 is the incident spectrum. These variables are calculated 10 times. Each trial consists of biopsy shuffling of tissue blocks for a chosen fractional composition (i,j,k). A statistical component was incorporated in No. As a preliminary test to the classification model, a total of 66 different biopsies have been simulated and are used as unknowns. The PCA classes generated are based on a biopsy composed of 5 layers and 5 radii. The 66 unknown compositions were tested against the PCA classes to determine if the biopsy tissue composition could be identified. Of the 66 biopsies, 46 were correctly identified. An example of a misclassified unknown was for the composition (0,0.2,0.8) which was classified as (0.3,0,0.7). Further work is required to increase accuracy of identification.