Development of Computer-aided Diagnosis System for Quantification of Temporal Changes of Small Pulmonary Nodules by use of Thoracic MDCT Images

本课题的主要研究成果概括为以下四个方面：1.肺小结节的分割准确地分割CT图像上的肺小结节是进行肺小结节鉴别诊断的重要特征提取技术。我们提出了一种SPNs自动分割的计算机方案，可以获得SPNs的精确体积数据，特别是GGO类型。2. SPNs时间变化的定量评价SPNs时间变化的定量评价是小肺癌诊断的重要依据。我们评估了从我们的计算机化方法获得的SPN的倍增时间，并将这些结果与放射科医生报告的结果进行了比较。3. SPN的特征分析SPN的特征分析对于区分良恶性SPN是很重要的。分形维数是与物体复杂程度相关的一种物理度量，对随机噪声的特征分析有重要意义。应用分形分析方法对孤立性肺结节的内部和周围纹理进行分析，以辅助放射科医师的诊断。4.全胸部CT图像的时间减影对多种肺部疾病的诊断，我们尝试对全胸部CT图像进行时间减影。我们使用肋骨结构进行配准。

The outline of our research results is summarized to the following four items :1.Segmentation of small pulmonary nodules (SPNs)Segmentation of SPNs on CT images accurately is an important technique for the feature characterization for differential diagnosis of SPNs. We have developed a computerized scheme for automatic segmentation of SPNs, and we can obtain precise volume data of SPNs, especially for GGO types.2.Quantitative evaluation of temporal changes for SPNsQuantitative evaluation of temporal changes for SPNs is an important for diagnosis of small lung cancers. We evaluated the doubling time of SPNs obtained from our computerized method, and we compared these results with findings reported by radiologists.3.Feature analysis for the SPNsFeature analysis of SPNs is an important for distinction between benign and malignant SPNs. The fractal dimension is a physical measure related to the complexity of object, and useful for feature analysis of SPNs. We analyzed the internal and peripheral textures of SPNs by use of fractal analysis for assisting radiologists' diagnoses.4.Temporal subtraction of whole chest CT imagesFor the diagnosis of many kinds of lung diseases, we tried making of temporal subtraction for whole chest CT images. We used rib structures for this registration.