基于卷积神经网络的胸部CT肺炎辅助筛查系统研发及应用

Development and application of a CT-assisted chest screening system for pneumonia based on convolutional neural network

新型冠状病毒感染疫情发生以来,胸部CT的检测量急速提升,各医院和基层医疗机构都面临着医疗资源不足、诊疗服务能力不足等问题。为了在医学影像领域开展辅助筛查或诊断等医疗信息化研究,众多医疗信息化人员引进计算机视觉等相关技术,打造了胸部CT肺炎辅助筛查系统。以人工智能(AI)胸片智能辅助筛查应用为例,研究了AI辅助诊断在包括新型冠状病毒感染在内的肺炎大流行期间的应用模式和效果。凸显出系统提升临床和放射工作人员影像阅片工作效率的有利性,并结合该AI辅助诊断模型的准确率和应用效率,阐明该类应用定性判断肺炎类型以及定量估计病灶面积的功效。最后结合临床科研实际工作,提出了应用的局限性和未来展望。

With the adjustment of the pneumonia prevention policy,the amount of chest CT detection has increased rapidly,and head hospitals and primary medical institutions are facing problems such as insufficient medical resources and insufficient diagnosis and treatment services.In order to carry out medical information research such as assisted screening or diagnosis in the field of medical imaging,many medical IT application staff have introduced computer vision and other related technologies to create a CT-assisted chest screening system for pneumonia.Taking the application of Artificial Intelligence(AI)chest CT intelligent-assisted screening application as an example,the application pattern and effect of AI-assisted diagnosis during the pneumonia pandemic,including COVID-19.It highlights the benefits of the system to improve the efficiency of clinical and radiological staff in image reading.Based on the accuracy and application efficiency of the AI-assisted diagnosis model,the efficacy of the AI-assisted diagnosis model in qualitatively judging the type of pneumonia and quantitatively estimating the area of the lesion was illustrated.Finally,based on the practical work of clinical research,the limitations and future prospects of the application are discussed.