改进集成技术在甲状腺超声图像分类中的应用研究.doc

目 录 第4章 改进动态集成算法在甲状腺超声图像分类中的应用·····································27 甲状腺 B 超图像分析··························································································27 甲状腺超声图像特征提取及量化·······································································28 感兴趣区域选取 ························································································28 特征提取及量化 ························································································29 不同分类器比较实验结果及分析·······································································35 图像来源 ····································································································36 图像分类效果评价指标·············································································36 实验结果及分析 ························································································37 本章小结··············································································································38 第5 章 总结与展望········································································································39 本文所做的主要工作 ··························································································39 工作展望··············································································································39 参考文献····························································································································41 致 谢································································································································45 攻读硕士学位期间取得的科研成果·················································································46 V 第 1 章 绪论 第1 章 绪论 课题来源和意义 甲状腺结节在临床中十分常见，通过高分辨超声发现，人群甲状腺结节的患病率为 19~67%[1]，其中，甲状腺癌占 5~15%，并且有逐年上升的趋势。在恶性甲状腺结节发生 的病因中，5-15%的患者与年龄、性别、头颈部照射史、家族病史等因素有关，其他原 因未明。甲状腺癌的病理分型主要可以分为甲状腺乳头状癌、滤泡癌、未分化癌及甲状 腺髓样癌，其中甲状腺乳头状癌及滤泡癌又称分化型甲状腺癌，占到了整个甲状腺癌的 90%[2]。临床中，对于良性甲状腺结节的病患一般采取随访及内科治疗，而对于恶性甲 状腺结节的病患则需通过外科手术治疗联合术后内科和同位素治疗以控制转移、复发及 死亡。因此，如何在大量的甲状