模式识别方法在腹部超声诊断中的研究

Applied research of PR method in abdominal ultrasound diagnosis

目的:基于模式识别(PR)方法构建腹部良恶性肿瘤诊断模型,设计超声诊断信息化系统流程,解决腹部超声诊断存在对比度低、声学伪影及噪声显著、肿瘤组织具有异型性、超声分布不均匀等问题,以降低诊断难度提高诊断效率。方法:基于数字图像处理技术,通过图像预处理,降低噪声、伪影的影响,提高图像对比度,进而采用图聚合分割法,将原始图像分块,分别计算分块图像纹理及形态学特征,构建图像分类器并计算腹部肿瘤轮廓,通过分割分离出感兴趣区域(ROI)构建腹部ROI良恶性肿瘤辅助诊断模型,完成辅助诊断流程。结果:腹部良恶性肿瘤诊断模型的构建实现了腹腔超声智能化诊断流程,测试中采用图聚合分割方法,降低了超声腹部肿瘤图像中对比度不足、噪声较多及伪影带来的影响,实现了腹部肿瘤目标区域分割功能。通过构建腹部良恶性肿瘤诊断模型,测试结果接近低年资医生诊断准确率,诊断时间缩短了43.25%。结论:研究构建的腹部良恶肿瘤诊断模型,能够实现腹部超声诊断的自动化流程,提高诊断效率,对于提高超声智能化诊断应用范围及相关疾病的筛查具有借鉴作用。

Objective:To construct the diagnosis model of abdominal benign and malignant tumors based on pattern recognition(PR) method,and design the process of information system of ultrasound diagnosis,and solve the existing problems of abdominal ultrasound diagnosis included low contrast,significant acoustic artifacts and noise,heteromorphism of tumor tissue,unevenly ultrasound distribution and so on,so as to reduce the difficulty of diagnosis and improve the efficiency of that.Methods:Using image preprocess to reduce the influences of noise and artifact,and enhance the picture contrast on the basis of digital picture processing technique.And then,picture aggregation segmentation method was adopted to segment primary image,and the texture and morphological feature of segmented images were further calculated,respectively,and the image classifier was constructed to calculate the abdominal tumor contour.The auxiliary diagnosis model of benign and malignant tumor of abdominal region of interest(ROI) was constructed through segmented and isolated ROI,so as to complete the process of auxiliary diagnosis.Results:The construction of the diagnosis model of abdominal benign and malignant tumor realized the intelligent diagnosis process of abdominal ultrasound.In the test,the adopted picture aggregation segmentation method had reduced the influences of insufficient contrast,more noise and artifacts in the abdominal tumor image of ultrasound,and realized segmentation function of target region of abdominal tumor.The tested results approached the diagnostic accuracies of junior doctors,and the diagnostic time was shortened by 43.25% of them through constructing the diagnostic model of abdominal benign and malignant tumor.Conclusion:The researched and constructed diagnostic model of abdominal benign and malignant tumor can realize the automatic process of abdominal ultrasound diagnosis,and improve the diagnostic efficiency,which has reference function in improving the application scope of ultrasound intelligent diagnosis and screening the related diseases.