Software Measures Image Interpretation and Accuracy

The software program RadBench effectively and objectively measures image interpretation performance and image interpretation accuracy, according to a study published in  Radiography.

Researchers from the United Kingdom sought to develop an assessment tool, called RadBench, in order to regularly measure and monitor image interpretation performance.

The study used two test banks, Test 1 and Test 2 to investigate the use of the software. Each test contained 20 appendicular musculoskeletal images (17 adult, three child). The images, 10 of which were normal and 10 with fractures, were of the ankle, foot, knee, hand, wrist, and elbow. Forty-two radiology professionals participated: 34 general radiographers, three reporting radiographers, two radiologists, and three medical imaging academics.[[{"type":"media","view_mode":"media_crop","fid":"45798","attributes":{"alt":"","class":"media-image media-image-right","id":"media_crop_7105811328381","media_crop_h":"0","media_crop_image_style":"-1","media_crop_instance":"5263","media_crop_rotate":"0","media_crop_scale_h":"0","media_crop_scale_w":"0","media_crop_w":"0","media_crop_x":"0","media_crop_y":"0","style":"height: 124px; width: 179px; border-width: 0px; border-style: solid; margin: 1px; float: right;","title":"©Bloomua/Shutterstock.com","typeof":"foaf:Image"}}]]

The participants ranked each image on a five-point scale, ranging from one, definitely normal, to five definitely abnormal. These choices were then calculate by RadBench to determine the reader’s sensitivity, specificity, and accuracy.

“The RadBench software generated calculations of sensitivity, specificity, and accuracy in addition to a decision making map for each respondent,” the authors wrote. Among the early findings, the researchers found that there was a 5% mean difference between image banks, confirming that benchmarking must be related to a specific test. “The benchmarking option within the software enabled the user to compare their score with the highest, lowest and mean score of others who had taken the same test.”

The highest scores were among reporting radiographers and radiologists (all scored 95% or above accuracy in both tests), while the general radiographer population scored between 60% and 95%.

The authors concluded that the RadBench tool was capable of providing benchmark measures of image interpretation accuracy, with the potential for comparison across populations.