Evaluation of an improved method of simulating lung nodules in chest tomosynthesis

• Published in: Acta radiologica (1987) 2012-10, Vol.53 (8), p.874-884
• Main Authors: Svalkvist, Angelica, Johnsson, Åse Allansdotter, Vikgren, Jenny, Håkansson, Markus, Ullman, Gustaf, Boijsen, Marianne, Fisichella, Valeria, Flinck, Agneta, Molnar, David, Månsson, Lars Gunnar, Båth, Magnus
• Format: Article
• Language: English
• Subjects: Adults; Area Under Curve; Artificial organs; Biological and medical sciences; Biological organs; Digital radiography; Diseases; Humans; Investigative techniques, diagnostic techniques (general aspects); Lung; Lung Neoplasms - diagnostic imaging; Medical sciences; MEDICIN; MEDICINE; Models, Anatomic; Multidetector Computed Tomography; Observer performance; Pathology; Pneumology; Radiographic Image Enhancement - methods; Radiologi och bildbehandling; Radiology; Radiology, Nuclear Medicine and Medical Imaging; Radiotherapy. Instrumental treatment. Physiotherapy. Reeducation. Rehabilitation, orthophony, crenotherapy. Diet therapy and various other treatments (general aspects); ROC Curve; Technology assessments; Thorax; Tomography; Tumors of the respiratory system and mediastinum; Visualization
• ISSN: 0284-1851; 1600-0455; 1600-0455
• DOI: 10.1258/ar.2012.120230

Background Simulated pathology is a valuable complement to clinical images in studies aiming at evaluating an imaging technique. In order for a study using simulated pathology to be valid, it is important that the simulated pathology in a realistic way reflect the characteristics of real pathology. Purpose To perform a thorough evaluation of a nodule simulation method for chest tomosynthesis, comparing the detection rate and appearance of the artificial nodules with those of real nodules in an observer performance experiment. Material and Methods A cohort consisting of 64 patients, 38 patients with a total of 129 identified pulmonary nodules and 26 patients without identified pulmonary nodules, was used in the study. Simulated nodules, matching the real clinically found pulmonary nodules by size, attenuation, and location, were created and randomly inserted into the tomosynthesis section images of the patients. Three thoracic radiologists and one radiology resident reviewed the images in an observer performance study divided into two parts. The first part included nodule detection and the second part included rating of the visual appearance of the nodules. The results were evaluated using a modified receiver-operating characteristic (ROC) analysis. Results The sensitivities for real and simulated nodules were comparable, as the area under the modified ROC curve (AUC) was close to 0.5 for all observers (range, 0.43–0.55). Even though the ratings of visual appearance for real and simulated nodules overlapped considerably, the statistical analysis revealed that the observers to were able to separate simulated nodules from real nodules (AUC values range 0.70–0.74). Conclusion The simulation method can be used to create artificial lung nodules that have similar detectability as real nodules in chest tomosynthesis, although experienced thoracic radiologists may be able to distinguish them from real nodules.