Introduction: DORA is a breast cancer screening programme. Healthcare professionals in the DORA programme provide services in accordance with the prescribed requirements of the programme. To achieve the quality of mammographic images, the quality of mammograms is evaluated once or twice each year. Quality control of the work of radiological engineers is performed at the Institute of Oncology in Ljubljana where the images are evaluated by the responsible radiological engineer. Purpose: The purpose is to establish the influence of the subjectivity of reading and whether the readings of mammographic images done by different readers following the same criteria differ from one another. Methods: We used the method of retrospective data analysis. First, we obtained 1000 mammographic images recorded within the DORA programme at the Institute of Oncology in Ljubljana in the time period from 2011 to 2013. The obtained images had already been evaluated according to the modified PGMI scoring system by a responsible person in the DORA programme: then, the entire set of images was evaluated by us (4 readers). By using the Kappa coefficient and the accuracy of final readings, we demonstrated the congruence/incongruence of results between the readers and the influence of subjectivity. Results and discussion: The research showed statistically significant differences between the readers and the influence of subjectivity. The results have shown that in 300 images evaluated by readers 1,3, and 5, the highest level of congruence in the CC projection with Kappa coefficient of 0.244 and 68% accuracy. The lowest level of congruence in the CC projection is with Kappa of 0.07 and 41% accuracy. In the MLO projection, the highest level of congruence is with Kappa of 0.326 and 61% accuracy, while the lowest level in the MLO projection is with Kappa of 0.098 and 40% accuracy. In 300 mammograms evaluated by readers 2, 4, and 5, the highest level of congruence is in the CC projection with Kappa of 0.193 and 66% accuracy. The lowest level of congruence in the CC projection is with Kappa of 0.04 and 37% accuracy. The highest level of congruence in the MLO projection is with Kappa coefficient of 0.288 and 67% accuracy, and the lowest level of congruence in MLO is with Kappa of -0.007 and 29% accuracy. While in 200 images evaluated by all readers, the highest level of congruence is in the CC projection with Kappa coefficient of 0.433 and 66% accuracy, the highest level of congruence in the MLO projection is with Kappa coefficient of 0.374 and 68% accuracy. Conclusion: By using the statistical analysis, we have established that the level of congruence between the readers is low, both in the CC and MLO projection. It is our assumption that the final assessment of mammographic images is influenced by the readers’ subjectivity, varying professional and practical knowledge, the readers’ criticality, poorly defined reading criteria, and different reading conditions (resolution, contrast, monitor brightness, brightness in the room).
|
