摘要
目的:运用BP神经网络技术建立甲状腺癌的无创诊断模型,评估该模型的预测诊断价值。方法:回顾性分析经术后病理证实为甲状腺癌39例与良性病变11例,提取出以上50例病例中手术前经过B超检查与甲状腺癌相关的8项图形特征,并进行评分量化,利用BP神经网络对50例病例进行学习和检验,建立甲状腺癌无创诊断模型。用该无创诊断模型对疑为甲状腺癌20例患者进行术前预测并与术后病理进行比较。结果:本文所建立的基于BP神经网络技术的无创诊断模型在甲状腺癌及甲状腺良性病变的预测诊断中达到了100%的准确率。结论:基于BP神经网络技术的无创诊断模型,在甲状腺癌及良性病变的预测诊断中具有较高的应用价值,这无疑对辅助B超诊断甲状腺良恶性病变提供了新的技术支撑和研究思路。
Objective：To establish a thyroid carcinoma noninvasive diagnostic model based on BP neural network technology,and to assess the predictive diagnostic value of the model.Method：The retrospective analysis was conducted on 39 cases of malignant thyroid carcinoma and 11 benign cases which were proved by pathology after the operation.All the patients were performed neck ultrasonography before operation.And eight characteristics of ultrasound images were assessed according to the established protocols for score to quantify.Then we applied BP neural network technology to learn and test,and to establish the diagnostic model of thyroid carcinoma.Finally,20 cases suspected as thyroid cancer were subjected to the newly established ultrasonography assessment.The diagnostic results were compared with the ones by postoperative pathologic diagnosis.Result：The recognition rate of noninvasive diagnosis model based on BP neural network technology was 100%.Conclusion：The noninvasive diagnosis model based on BP neural network model is of high value in the predictive diagnosis of thyroid cancer and benign lesions.It is a new technical support and research idea on the auxiliary B-ultrasound diagnosis of thyroid benign and malignant lesions.
出处
《现代生物医学进展》
CAS
2012年第36期7104-7108,共5页
Progress in Modern Biomedicine
关键词
BP神经网络
甲状腺癌
B超
无创诊断
BP Neural Network
Thyroid carcinoma
B-ultrasound
Noninvasive diagnosis
