A quantitative assessment of chest image changes in patients with Severe Acute Respiratory Syndrome (SARS) using computer-aided detection method

This study aims to investigate the value and validity of implementing an objective computer-aided detection (CAD) method in assessing the progressive radiographic changes of SARS features on the chest radiographs, so as to enhance the SARS patient management. Tests were designed to evaluate the use of a CAD program, the validity of program and the inter-observer variability. The percentage areas of opacification (AO %) were obtained from 29 sets of SARS chest radiographs at four different clinical stages (Admission; Medication; Maximal score and Discharge) with the use of a CAD method. Data obtained by a CAD method and radiographic scores determined by visual estimation were compared by non-parametric statistical tests. The optimum points to locate the chest features were determined and the program was proven to be valid with error less than 2%. There was no significant differences (p>0.05) in performance among observers using the CAD methods. Also there was no significant difference in performance for detecting changes of SARS for CAD method and visual estimation (p>0.05, Wilcoxon test) at different stages of chest changes. In conclusion, it was valid to apply an objective CAD method in quantification of SARS severity. Given simple guidelines and training, trained clinical staff can use the CAD method to evaluate SARS quantitatively and this would help alleviate the workload for clinical department.