基于冗余小波和分水岭变换的肺结节识别

Recognition of lung nodules based on redundant wavelet and watershed transform

利用CT影像检测肺结节,已成为目前诊断和预防早期肺癌的主要方法。对肺结节的有效识别,是实现肺癌计算机辅助诊断的关键。本文根据医学影像的特点,提出一种基于冗余小波变换和分水岭变换的肺结节识别方法,该方法通过保持平移不变的冗余小波变换得到细节信息增强的影像,然后用基于有向欧式距离变换的分水岭算法实现对肺结节的分割。实验表明,该方法对CT序列影像进行肺结节识别可以得到良好的分割结果,从而为医生对肺结节的诊断提供更加准确的客观依据。

Recently an important clinical indication for preventing lung cancer is its detection and treatment at early stage of growth through CT examinations of the thorax. Effective and efficient recognition of pulmonary nodules plays a significant role in modem computer aided diagnosis （CAD） of lung cancer. This paper presents a method for lung nodule recognition based on undecimated wavelet and watershed transform. The thoracic CT images are enhanced using translation invariant redundant wavelet transform. Then the lung nodules can be segmented through watershed algorithm using topographic distance obtained by signed Euclidean distance transform. Experiments with real thoracic CT images show that the proposed method is able to achieve satisfactory results of nodule recognition and can help clinical radiologists on diagnosis of lung cancer.