Study The Impact Of CT Scanning Technologies To Certify SPNs Signs And To Differentiate Malignant And Benign Nodules Using ROC Curve

Objective:To illustrate the role of chest CT in displaying the signs and accuracy of the diagnosis of Solitary Pulmonary Nodules (SPNs). The research used the receiver operating characteristive(ROC) curve for comparing and evaluating the differential diagnosis and accuracy of diagnosis by four physicians with different experiences and sub-specialty in the same condition.Material and methods:74 SPN cases (40 malignant and 34 benign nodules) including 64 cases which were diagnosed erroneously preoperatively were considered. All the cases were proved by clinical and pathological examination after surgery. All the patients (27 men,47 women, in the age range of 20-79 years, mean 56.3 years) were examined at least once with routine, thin-section and dynamic enhanced chest MSCT (multi-slice CT)(GE Lightspeed 4 rows,16 rows and Philips Brilliance 16 rows multi-slice CT scanner,2.0mm-10.0mm thickness), in 8 cases HRCT (high resolution CT) was included. All images are observed three times at the same condition by four phycicians with different experience and sub-specialty (A1 and A2 are chief physicians with high qualification, Al specializes in respiratory diseases, A2 is non-respiratory specialist, B1 and B2 are attending physicians with no clear sub-speciality). All cases had been read double-blind by these physicians for the first two times. The third observation was made with four physicians consultation to give final unified diagnosis. Viewing the lesions in conventional lung and mediastinal window, physicians were asked to certify whether the SPNs had lobulations, spiculations, vascular clusters, air bronchogramme or pleural adhesions, and then had to arrive at a diagnosis. Finally, the differences in observation and diagnosis of these SPNs and use of ROC curve to analyze the results of diagnosis at three different conditions among the four physicians were compared.Results:1. On certifying SPNs total signs, phisician B1 has significant difference(P<0.05) in certifying SPNs signs with phisician A1, A2, B2 at routine and thin-section scans. Phisician B2 has statistic differentce in certifying SPNs signs by using routine scans and the thin-section MSCT scans(P=0.002). On certifying SPNs each signs, four physicians were different in certifying spiculations, air bronchogram sign and pleural adhesion signs (P<0.05) at routine scan, and were different in analyzing lobulations, spicultions, air bronchogram sign, vessel cluster sign and pleural adhesion signs (P<0.05) at thin-section and dynamic enhanced MSCT scans. The physicians had no significant difference in certifying the five signs by using routine scans and the thin-section and dynamic enhanced MSCT scans except A2 who showed a significant difference in certifying spiculation sign, B1 in vessel cluster sign and B2 in lobulation sign (P=0.033, P=0.031, P=0.000 respectively). There is no statistical difference while comparing the enhancement pattern among the four physicians (P=0.757).2. The mean sensitivity, specificity and accuracy of routine and thin-section and dynamic enhancement MSCT scans were 0.600,0.539,0.562 and 0.605,0.576,0.617 respectively. The specificity and accuracy had significant difference between the two methods (P＜0.01). Area under ROC viewed by physician B1 by the two different methods had significant difference (Z=6.60, P<0.01). Specificity and accuracy of diagnosis increased greatly on thin-section and dynamic enhancement scans.3. Area under ROC curve, using the routine scans had significant difference when compared between the physicians. Area under the ROC curve, using the thin section and dynamic enhancement scans showed significant difference between all the physicians except B1 and B2 (P>0.05). The area under ROC curve measured by A1 was more than that of other physicians (P<0.01).4. Area under ROC curve between each physician and among physicians with consultation with each other had significant difference (P＜0.01). Accuracy of diagnosis made with consultation among physicians was definitely higher than the accuracy of diagnosis made by each physician.Conclusion:With wide usage of multi-slice CT (MSCT) and the emerging trend of conventional chest CT images (thickness 5mm and above) and thin-reconstruction images sent to on-line systems (PACS), the use of HRCT for the diagnosis of small lung lesions and enhanced scans for larger lesions is clearly shrinking. This study showed that, different CT scanning technology will impact certifying SPN signs, and hence the thin-section scanning should be advocated. Different CT scanning technologies impact the accuracy of diagnosis by radiologists. Thin-section scanning and dynamic enhanced CT increase the accuracy of diagnosis. There is difference in the accuracy of diagnosis by radiologists with different experience and sub-specialty. The richer experience and with thoracic imaging as the sub-specialty of the physician, the accuracy of diagnosis is higher. The study also showed that the diagnostic accuracy will be increased with consultation among the physicians and by extending the time for reading the images. Roc curve is the valid statistic method for evaluating the efficiency of diagnosis of SPNs by different CT technology, different phisicians in different diagnostic schemes.