基于主成分分析模型的肝超声造影图像呼吸运动校正研究

Respiratory motion correction method based on principal components analysis model for liver contrast-enhanced ultrasound sequences

目的 :为了减少呼吸运动对肝双模超声造影(contrast-enhanced ultrasound,CEUS)图像序列的影响,提高肝CEUS定量分析的准确性,提出有效的肝CEUS序列中呼吸运动校正策略。方法:采用18例兔肝VX2肿瘤的CEUS图像序列建立主成分分析(principal component analysis,PCA)呼吸运动模型,利用数据比例较大的主成分合成出呼吸运动曲线,然后挑选与参照图像相近相位的图像进行分析。结果:校正后图像序列的平均结构相似度和平均互相关系数均明显增加(P<0.001),分别为0.57±0.11和0.78±0.11。校正后图像序列的偏差值均值下降到29.9±7.02,约为原值的1/3。采用阈值设定还可进一步提高挑选的图像序列质量。结论:该校正方法已在兔肝CEUS病例中初步验证了其有效性,且操作简单,有助于提高肝肿瘤的良恶性鉴别诊断率。

Objective To relieve the influences of the respiratory motion on the liver contrast-enhance ultrasound（CEUS）image sequences, to enhance the quantitative analysis accuracy for liver CEUS and to put forward a correction strategy for the respiratory motion in liver CEUS sequences. Methods A principal component analysis（PCA） model of the respiratory motion in liver CEUS sequences was established with 18 cases of rabbit liver VX2 tumors, and a respiratory motion curve was generated based on the principal component with large data proportion, then the images with similar phases to the reference image were analyzed. Results Correction made the mean structural similarity and mean correlation coefficient enhanced significantly to 0.57±0.11 and 0.78±0.11 respectively（P0.001）, while the average of deviation valve（DV） was decreased to 29.9±7.02 which only was one-third of the original value. Threshold setting could further improve the quality of the selected image sequence. Conclusion The proposed respiratory motion method proves its effectiveness for rabbit liver CEUS image sequences, and thus contributes to enhancing the differential diagnosis rate of benign and malignant liver tumors.