Keywords:
Performed at one institution, Not applicable, Retrospective, Tissue characterisation, Quality assurance, Cancer, Physics, Observer performance, Comparative studies, Mammography, Image manipulation / Reconstruction, Digital radiography, Radiation physics, Oncology, Breast
Authors:
C. Van Ongeval1, L. Cockmartin2, S. Postema2, R. Prevos2, M. Keupers2, T. Thywissen2, G. Zhang2, H. Bosmans2; 1Leuven/BE , 2Leuven/BE
DOI:
10.26044/ecr2020/C-07603
Methods and materials
1. Patient selection and image acquisition
Of 184 diagnostic mammography exams , acquired in a combotomo mode (FFDM with digital breast tomosynthesis during one compression) on a Inspiration system with VB6 software (Siemens Healthcare GmbH, Forchheim Germany) only the oblique view was selected.
Of these 184 exams, 113 were normal and 71 contain biopsy proven malignant lesions.
All digital breast tomosynthesis cases were reconstructed in two versions of a SM:
- the current version of SM, InsightV1
- a prototype of a new version of SM, InsightV2
2. Image evaluation
Two readers (more than 10y experience) read the images using a side-by-side evaluation of InsightV1, InsightV2, FFDM, randomly sampled and presented.
a. Visual grading characteristics (VGC) (5-point rating scale)
- representation of opacities,
- representation of microcalcifications,
- presence of noise,
- presence of contrast,
- presence of sharpness
- overall image quality
b. Diagnostic accuracy (BI-RADS classification) for FFDM, InsightV1 and InsightV2 seperately
c. Breast density score (BI-RADS 5thedition) for FFDM, InsightV1 and InsightV2 seperately
d. FFDM likeliness score (10-point rating scale) for FFDM, InsightV1 and InsightV2 seperately
3. Statistics
Area under the curve (AUC) and statistical differences were calculated for VGC and Receiver Operating Characteristics (ROC) study results (ROC analysis 1-5; VGC analysis 6); for all six criteria together as well as for each criterion separately with InsightV1 in x-axis and InsightV2 in y-axis of VGC curve