Keywords:
Cancer, Diagnostic procedure, Image manipulation / Reconstruction, Oncology, Breast
Authors:
A. Iriarte1, G. de vargas2, C. Bore1, P. Bannister1, L. Tsui1, D. Gibbins3; 1Bristol/UK, 2Madrid/ES, 3Bristol /UK
DOI:
10.1594/ecr2018/C-1623
Results
The methodology has been tested using clinical bilateral data from 57 patients with known tissue properties (25 with lucent tissue and 32 with dense tissue),
leading to a population of 114 breast scans,
64/50 symptomatic/non-symptomatic.
Symptomatic here includes breasts with an inclusion e.g.
cyst,
cancer.
It was found important to account for breast size,
results demonstrated here are all from a single measurement cup size of approximate volume of 460 ml.
Fig.
5 shows the classifier applied to the non-symptomatic cases.
In total 76% of cases were classified correctly.
Except for a single outlier in each case BIRAD groups a and d were otherwise classified correctly.
As expected,
groups b and c were not as easy to classify though this might be a result of error in clinical BIRAD density estimation. When evaluating both symptomatic and non-symptomatic breasts 70% of the cases were correctly classified.