Computer assisted diagnosis using artificial neural network

口腔癌患者颈淋巴转移的存在对预后和治疗有重要意义。关于结节是转移性还是反应性的标准还没有很好的确立。同时，最近引入的人工神经网络在诊断图像分析中可能具有光明的前景。本研究的目的是确定人工神经网络在超声图像上诊断淋巴结转移的准确性。共188个淋巴结用于研究，均经组织学证实。在研究的第一部分，随机选择了138个节点。其中60个为转移性，78个为反应性。超声诊断仪为PT2600型超声诊断仪，探头频率为7.5 MHz。评价7种声像图特征：中央回声门区、周边部回声、周边部回声均匀、边缘、边缘、最大/最小比值和最小直径。为了证明超声特征对诊断准确性的影响，使用了各种网络结构和超声特征。根据第一部分的研究结果，用Visual C++开发了报表系统。以剩余的50个结节为研究对象，我们对报告系统的能力进行了检验，结果表明，中心回声门区、周围淋巴管回声和max-mm比值是诊断转移的最重要的指标。神经网络辅助报告系统提高了非熟练医生的诊断能力。综上所述，神经网络可能有助于颈淋巴结肿大的诊断。

The presence of cervical lymph node metastases in patients with oral cancer is of great prognostic and therapeutic importance. The criteria whether the node is metastatic or reactive has not been well established. Meanwhile, artificial neural networks recently introduced in the analysis of diagnostic images may have bright prospects. The purpose of this study is to determine the accuracy indiagnosing the lymph node metastases on the ultrasonographic images using artificial neural network.Total of 188 nodes was used to the studies, all of which were verified histologically. In first part of the study 138 nodes randomly selected were used. Included among them were 60 nodes metastatic and 78 reactive. An ultrasonographic apparatus was model PT 2600 US scanner using 7.5 MHz B-mode linear scan probe. Seven ultrasonographic features evaluated were central echogenic hilus, echogenity of peripheral parencymal zone, homogeneity of peripheral parencymal zone, margin, border, Max-Min ratio and smallest diameter. To prove the effect of ultrasonographic features on diagnostic accuracy, a variety of network structure and ultrasonographic features were used. According to the result of the first part of the study, we made reporting system using Visual C++. The residual 50 nodes were used for the study in which we examined the ability of the reporting system.The results showed that central echogenic hilus, the echogenity of peripheral parencyma and Max-Mm ratio were most important features on diagnosis of metastasis. The neural network assisted reporting system improved the diagnostic ability of the unskilled doctors. In conclusion, the neural network may assist the diagnosis of cervical lymph node swelling.