Telemammography is more feasible now that low-cost, high-performance data networks are proliferating and digital mammography equipment is displacing analog mammography equipment. Also, many diagnostic workstations are vendor-agnostic with respect to being able to display and manipulate images acquired from multiple mammography systems.
Researchers at the Medical University of Vienna decided to formally demonstrate the feasibility of sending uncompressed digital mammograms in a teleradiology setting and compare image quality, lesion detection, and BI-RADS assessment. Over a nine-month period, three experienced breast mammographers each interpreted 200 digital mammography studies at three different hospitals in Vienna: the Medical University of Vienna, Sozialmedizinisches Zentrum Ost - Donauspital, and the Wilhelminenspital. They then conducted a clinical assessment to determine if there was any loss of information that could affect image quality and interpretation. Their findings appear in the March edition of the European Journal of Radiology (March 2013, Vol. 82:3, pp. 412-416).
The 200 studies were retrospectively and randomly acquired from the hospital's PACS (Impax version 52.5 Agfa HealthCare). They included 50 studies each that had been classified as BI-RADS 1, 2, 4, and 5. Half of the BI-RADS 4and 5 reports had a finding classified as a mass and the other half had a finding classified as a calcification. The 32-MB images were anonymized and transmitted in uncompressed DICOM format to the two other institutions using a transmitting system (Centricity RA 600, v. 7.0A, GE Healthcare).
The originating and one receiving institutions used the same mammography workstation (Seno Advantage, GE Healthcare), but each used a different software version. The other receiving workstation was a Syngo MammoReport from Siemens Healthcare. All workstations were equipped with two 21.2-inch high-resolution 5K mammography flat-screen display monitors.
The workstations displayed the four craniocaudal and mediolateral oblique two-view images in a consistent format on the monitors using each institution's individual settings. No adjustment was allowed, although the mammographers were able to use the magnification tools of each workstation. No prior exams were available for comparison, nor was any demographic information or clinical history available. Computer assisted diagnosis (CAD) software also was not available.
The radiologists were asked to evaluate and rank on a five-level score the image quality of each study. Their assessment included brightness, contrast, sharpness, noise, skin, fat, and retromammilary space. They assessed each mammogram according to BI-RADS. The American College of Radiology (ACR) criteria was used to evaluate breast density.
The mammographers were asked to classify their findings as either benign or malignant, and report if there were calcifications or masses. They also had to provide a BI-RADS score. Exams were presented in different orders at each institution, and the three participants waited 90 days after completing the readings at one institution before reading the next set at another institution.
Dr. Julia Fruehwald-Pallamar, an assistant professor of radiology at the Medical University of Vienna, and colleagues determined that there was no loss of information, and image quality was rated as being of good or very good quality for all four breast views, regardless of the different workstations used. The rating of the level of acceptability of the images was heavily dependent upon the reader. Reader one ranked 9.5% of the images at the originating institution as acceptable to poor regarding one or more image quality parameters, whereas reader three ranked only 5.5% in this category and reader two only 2.5%.
At the receiving institution where the Siemens Syngo MammoReport diagnostic workstation was utilized, only 3% to 7.5% of the exams were ranked as poor to acceptable. Less than optimal images displayed at the third institution received the most consistent rating, from 4.5% to 5.5%. A phantom also was used to measure image quality at each display workstation.
The congruence of the readers with respect to the final BI-RADS assessment ranged from 90% and 91% at the originating institution compared with the institution with the Siemens workstation. The range at the originating institution compared with the third institution, both of which used workstations from the same vendor but with different software versions installed, was from 86% to 92%.
The radiologists were in agreement with respect to the detection and rating of benign and malignant masses, as well as calcifications. BI-RADS scores also were consistent, regardless of location.
This proof-of-concept study should help alleviate concerns about transmitting mammogram studies for telemammography interpretation to institutions needing specialist radiologists to read breast imaging exams or when a second opinion is needed. The barriers of a decade ago and more have been eliminated.
"At the moment we do not have specific plans for using teleradiology with mammograms in Austria, but we think in the future it could be applied for second readings of screening mammograms," noted Fruehwald-Pallamar in an email reply to AuntMinnieEurope.com.