Keywords:
Performed at one institution, Diagnostic or prognostic study, Retrospective, Cancer, Screening, Computer Applications-Detection, diagnosis, CAD, Mammography, Oncology, Breast, Artificial Intelligence, Artificial Intelligence and Machine Learning
Authors:
G. Porrello, A. Orlando, M. I. Schillaci, M. Dimarco, M. L. Di Vittorio, S. Busalacchi, M. insalaco, S. Vitabile, M. Midiri; Palermo/IT
DOI:
10.26044/ecr2020/C-05794
Methods and materials
Patient population:
In this retrospective study, we included 106 patients (age range: 40-75 years), who consecutively underwent digital mammography at our institution from January to December 2015, and whose subsequent work-up or 2-years follow-up was available.
No information about the patient population was provided to the readers.
Mammograms selection and evaluation:
106 mammograms (MGs) were firstly evaluated by an expert radiologist, with over 20 years’ experience in breast imaging. The same MGs were then retrospectively categorized, according to the BI-RADS lexicon developed by the American College of Roentology (5th edition, 2013) by two radiology residents (R1 and R2) in two sessions, respectively, the first one without, and the second one with the support of BD4BREAST.
What is BD4BREAST?
BD4BREAST is a Decisional Support System (DSS) software, based on artificial intelligence, developed by a cooperation of the DIBIMED Department of the University of Palermo and MIRC Srl (MIUR Start-Up 2013).
This software allows the following goals:
- Acquire, store and visualize DICOM images and reports of mammograms (Fig.1)
- Post-processing of the images (simple/multiple visualization, adjustment in brightness and contrast, zooming, measurement, scale grey invert, color maps…) (Fig.2)
- Automatic/manual identification of the areas of abnormal contrast, suspicious for malignancy (CAD function) (Fig.2)
- Aid the radiologist for the compilation of the report, creating a standard structured report and suggesting the correct BIRADS lexicon to apply for each lesion. (Fig.3)
- Analyze the acquired data for research purposes and statistics
When an abnormal area is identified by the software, the critical points (defined as “key points”) will be visualized with circular ROIs and there is the possibility to compare such findings to other similar cases or to add/remove key points when necessary. (Fig.2)
Standard reference:
Histological exams of all MG findings classified as BIRADS 4-5 were available as a standard reference. For all MGs findings classified as BIRADS 0 to 3, 2-year-follow-up mammograms were available, confirming the presence (M+) or absence (M-) of malignancies.