Radiologists increasingly use computer-aided detection for mammography interpretation
A team of Dartmouth Medical School researchers, led by Tracy Onega, recently conducted a study on radiologists’ perception of computer-aided detection versus double reading for mammography interpretation. Their findings were published in Academic Radiology.
The traditional method of double reading by a second clinician in screening mammography interpretation may soon be replaced by computer-aided detection (CAD) of breast cancer. It is thought that the use of CAD has been gaining popularity among radiologists, although sufficient research on this topic is lacking. This trend may be explained by the fact that CAD generates additional reimbursement while double reading demands more radiologist time without reimbursement for the second read.
Despite the diffusion of CAD in clinical practice, there is limited and contradictory information as to its relative harms and benefits. Thus, the different perceptions of CAD and double reading are likely to have implications for the variability and performance in mammography interpretation.
The researchers mailed a survey to 257 community radiologists to assess their perceptions and practices regarding CAD and double reading. The data was analyzed by classifying the radiologists’ perception of both practices based on their relative agreement or disagreement with several statements specific to CAD and double reading.
The results showed that 64% of radiologists used CAD for more than half the screening mammograms they interpreted while only fewer than 5% said the same for double reading. However, more radiologists believed that double reading improved cancer detection rates as compared to CAD. At the same time, more radiologists were likely to think that CAD decreased recall rates rather than double reading. Perhaps not surprisingly, it was also found that those radiologists with the most favorable perceptions of CAD had significantly higher uses of CAD, greater workload in screening mammography, academic affiliation, and fellowship training.
As the researchers note, the study is important in understanding how radiologists use and perceive technology as their perceptions of both practices influence how and under what circumstances either is used. The researchers also suggest that the next step in this research could be to determine how radiologists respond to CAD and double reading when both methods have equal reimbursements or no reimbursement at all. Understanding radiologists’ perceptions of both interpretive practices can help better inform clinical practice.