空洞检测算法自动化检测肺结节内空泡和空洞的可行性

目的 探讨利用空洞检测算法自动化检测肺结节内空泡和空洞的可行性。方法 收集经病理或随访证实的49例肺结节，其中良性16例，恶性33例，所有患者均接受胸部CT检查。由2名高年资影像科医师盲法对CT图像进行主观评价，得出有无空泡和空洞的结论。利用基于阈值的空洞检测算法对CT图像中指定的肺结节进行有无空泡和空洞的客观评价。对两种评价方法所得结果进行对比分析。结果 利用空洞检测算法对肺结节内空泡和空洞征象前后2次提取的数据稳定（Kappa=1），与高年资医师主观判读结果差异无统计学意义（χ²=0.862，P=0.353），且一致性较好（Kappa=0.785），并可明显节约每个结节数据提取时间(7.72s±2.26)s vs (24.48±8.24)s,t=14.64,P<0.001]。结论 利用空洞检测算法提取肺结节内空洞和空泡征象稳定、快捷，有望成为低年资医师诊断肺结节的辅助检测工具。

Feasibility of automatic detection of cavity and vacuole in pulmonary nodules using cavity detection algorithm

Objective To assess the feasibility of automatic detecting cavity and vacuole in pulmonary nodules using cavity detection algorithm. Methods Totally 49 patients with pulmonary nodules confirmed by pathology or clinical follow-up were enrolled. There were 16 cases of benign nodule and 33 cases of malignant nodule. All cases underwent chest CT examinations. CT image data were subjective judgment by two senior radiologists using double-blind method to determine whether there was any cavitation and vacuole in pulmonary nodules. Meanwhile, the cavity and vacuole in specified pulmonary nodule on CT image were estimated using cavity detection algorithm based on threshold. The results produced from the two evaluatiing methods were compared. Results There was high consistency of cavity detection algorithm results between two time measures (Kappa=1). No statistical difference was found between results obtained by cavity detection algorithm and subjective judgment. And there was good consistency (Kappa=0.785). The detecting time per nodule of cavity detection algorithm (7.72±2.26]s) was shorter than that of subjective judgment (24.48±8.24]s, t=14.64, P<0.001). Conclusion Cavity detection algorithm may become an auxiliary tool for resident doctors to detect cavity and vacuole in pulmonary nodules with advantages of stable and convenient.