Keywords:
Lung, Respiratory system, Thorax, CT, CT-High Resolution, Computer Applications-Detection, diagnosis, eLearning, Education and training, Occupational / Environmental hazards
Authors:
J. Yi, K. A. Tapia, J. Robinson, Z. Gandomkar, P. Brennan, M. E. Suleiman, N. W. Sommerfeld, S. Tavakoli Taba
DOI:
10.26044/ecr2024/C-12359
Methods and materials
Fifteen HRCT cases of silicosis and non-silicosis presentations were assembled into test sets. Ground truths were established by expert radiologists with B-reader certification and verified diagnostic reports. Online platforms were developed in partnership with DetectedX (Sydney, Australia) to display the test sets of high-resolution images and diagnostic questions based on the International Classification of HRCT for Occupational and Environmental Respiratory Diseases (ICOERD) (Figure 1).
54 readers (33 radiologists and 21 radiology trainees) were recruited for this observer performance study, with ethical approval from our local human research ethics committee. Radiologists (n=33) had an average of 11 years’ experience, and the group of radiology trainees (n=21) had an average of 3 years’ radiology training.
Each reader independently graded each case in the test set based on the ICOERD for rounded, irregular, large, or ground glass opacities, honeycombing, emphysema, mosaic attenuation, parenchymal abnormality, pleural disease, and lymph nodes were graded by each reader. Both test sets also required the reader to rate each case on a 5-point confidence scale ranging from 1 (confidence that silicosis is not present) to 5 (highest confidence that silicosis is present).
Mann–Whitney U test was used to analyse performance between readers and Spearman’s rank correlation coefficient analysis was used to examine the relationship between experience and performance