基于影像的计算机深度学习对脊柱原发肿瘤精准诊断及预后评估的价值研究

脊柱起到支撑和保护脊髓、脊神经的重要作用，脊柱原发性肿瘤可造成脊柱骨质破坏、并可引起神经功能损害甚至瘫痪，恶性者还可引起全身多系统转移，严重影响患者生存质量。由于椎体解剖复杂、肿瘤组织来源和病理类型繁多、临床症状类似，因此术前精准诊断非常困难，同时，不同性质和分期的肿瘤治疗方法及预后差别很大。因此，亟需一种无创、快捷、精准的影像学技术来精准诊断脊柱原发性肿瘤、预测肿瘤生物学行为、提示预后，以指导临床进行个性化治疗。目前随着计算机机器学习能力的提升和医疗大数据的发展，为实现医学影像的自动分析及辅助医生实现疾病的高精度人工智能诊断与预测提供了新的契机。本研究拟通过对脊柱原发性肿瘤患者进行规范化的临床及影像数据提取，应用深度学习神经网络技术进行分析和预测，并充分结合组织病理学及临床随访结果，探讨计算机深度学习应用于脊柱原发性肿瘤精准诊断、预测肿瘤生物学行为以及治疗预后的价值。

The spine plays an important role in supporting and protecting spinal cord and spinal nerves. Primary spinal tumors will cause spinal bone destruction, neurological damage or even paralysis, and the malignant tumors will cause the whole multi-system metastasis of tumors, which seriously affect the life quality of patients. Because of the complex anatomy of the vertebral body, many kinds of tumor tissue sources and pathological types and similar clinical symptoms, the preoperative accurate diagnosis becomes very difficult. At the same time, the treatment methods and prognosis of tumors with different natures and stages have a big difference. Therefore, a noninvasive, rapid and accurate imaging technique is urgently needed to achieve the accurate diagnosis of primary spinal tumors, predict tumor biological behavior and indicate prognosis, in order to guide the personalized treatment clinically. At present, with the improvement of computer machine learning ability and medical big data, it provides a new opportunity for the automatic analysis of medical imaging and assist doctors to get the intelligent diagnosis of high-precision and the prediction of diseases. In this study, by getting the standardization clinical and the imaging data of patients with primary spinal tumors, applying the deep learning to analyze and predict the diseases, and fully combining the results of histopathology and clinical follow-up, we can explore the value of computer deep learning in the accurate diagnosis, the prediction of biological behavior and prognosis of the primary spinal tumors.