摘要
针对医学影像中小结节容易被漏诊的问题,提出了基于胸部CT图像的肺癌计算机辅助诊断新方法.首先从胸部CT图像分割出关心区域(ROI);然后提取ROI的特征;其次采用RS理论选择有效特征;最后基于这些有效特征建立面向不同需求的肺癌识别模型.即如果需要快速诊断,则利用SONN建立肺癌识别模型;如果需要进行准确诊断,则利用SPAM建立肺癌识别模型和非肺癌识别模型,并根据待识别样本与模型的相似度判断所属类别.但是当相似度较小时,则利用HMM进一步识别.通过实验验证了该方法的有效性.
To solve the missed diagnosis of small puhnonary nodules in medical irnages,a new approach on computer-aided diagnosis for lung cancer based on chest CT images has been proposed. The method firstly segments the Region of Interest(ROI), and extracts ROI's features. Then it selects effective attributes by theory of Rough Set(RS). Finally, it coma-acts a specific-demand oriented recognition model for lung cancer based on these effective features.Especially, we take the Self-Organizing Neural Network (SONN) to construct the recognition model of lung cancer for fast diagnosis. In order to perform the accurate diagnosis, we need to use the Self-Adaptive Probabilistic Model( SAPM) to build lung cancer and non-cancer recognition models respectively and we can identify the classification by the similarity of the recognition sample with the model. When the similarity is small, we re-identify the lung cancer by Hidden Markov Model(HMM). The experiment results proved that the approach mentioned in this paper can hold high efficiency.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2009年第8期1664-1668,共5页
Acta Electronica Sinica
基金
国家自然科学基金(No.60671050)
关键词
肺癌
计算机辅助诊断
自适应概率统计模型
隐马尔可夫模型
lung cancer
computer-aided diagnosis(CAD)
self-adaptive probabilistic model(SAPM)
hidden markov model (HMM)