Comparison of soft-copy and hard-copy reading for full-field digital mammography

Abstract

Purpose: To compare radiologists' performance in detecting breast cancer when reading full-field digital mammographic (FFDM) images either displayed on monitors or printed on film.

Materials and methods: This study received investigational review board approval and was HIPAA compliant, with waiver of informed consent. A reader study was conducted in which 26 radiologists read screening FFDM images displayed on high-resolution monitors (soft-copy digital) and printed on film (hard-copy digital). Three hundred thirty-three cases were selected from the Digital Mammography Image Screening Trial screening study (n = 49,528). Of these, 117 were from patients who received a diagnosis of breast cancer within 15 months of undergoing screening mammography. The digital mammograms were displayed on mammographic workstations and printed on film according to the manufacturer's specifications. Readers read both hard-copy and soft-copy images 6 weeks apart. Each radiologist read a subset of the total images. Twenty-two readers were assigned to evaluate images from one of three FFDM systems, and four readers were assigned to evaluate images from two mammographic systems. Each radiologist assigned a malignancy score on the basis of overall impression by using a seven-point scale, where 1 = definitely not malignant and 7 = definitely malignant.

Results: There were no significant differences in the areas under the receiver operating characteristic curves (AUCs) for the primary comparison. The AUCs for soft-copy and hard-copy were 0.75 and 0.76, respectively (95% confidence interval: -0.04, 0.01; P = .36). Secondary analyses showed no significant differences in AUCs on the basis of manufacturer type, lesion type, or breast density.

Conclusion: Soft-copy reading does not provide an advantage in the interpretation of digital mammograms. However, the display formats were not optimized and display software remains an evolving process, particularly for soft-copy reading.

Figure 1:

Mammographic image from 73-year-old patient, displayed on soft-copy monitor (left) and printed on film (right), was obtained with Fischer system. Interpretation of printed digital image at acquisition site was done with manufacturer's skin algorithm applied. Soft-copy image displayed here also includes algorithm.

Figure 2:

Mammographic image from 49-year-old patient, displayed on soft-copy monitor (left) and printed on film (right), was obtained with FujiFilm system. Image shows difference in adjustments to contrast and brightness that were, in some instances, made prior to printing digital mammograms to film. Readers in this experiment could adjust brightness and contrast of soft-copy images.

Figure 3:

Mammographic image from 58-year-old patient, displayed on soft-copy monitor (left) and printed on film (right), was obtained with GE Healthcare system. For this image, soft-copy default was printed on film.

Figure 4:

Forest plots of AUCs for machine types. Each plot shows soft-copy AUC (SC-AUC) and standard error of the mean (SC-SE) and hard-copy AUC (HC-AUC) and standard error (HC-SE) for each reader and average of all readers. Average AUC for hard-copy (▵, dotted vertical line) and soft-copy (•, solid vertical line) images are shown.