基于传统组学与深度学习融合技术的人工智能模型在鼻骨骨折检测中的应用

The Application of Artificial Intelligence Model Based on the Fusion Technology of Traditional Omics and Deep Learning in the Detection of Nasal Fracture

目的 采用传统组学与深度学习融合技术开发人工智能(AI)模型,并探讨其对鼻骨骨折的诊断价值。方法 选取2021年5月至2022年7月医院收治的252例行鼻骨多层螺旋CT(MSCT)扫描检查患者为研究对象,通过深度学习技术提取端至端的深度学习特征,并使用pyradiomics分析软件提取传统组学特征。将所有特征筛选后进行融合,输入分类预测器建立AI预测模型,并检验鼻骨骨折的诊断效能。结果 在AI辅助下,临床医师阅片时,灵敏度改善11.65%,特异度改善16.80%,曲线下面积(AUC)改善0.14;放射科医师阅片时,灵敏度改善9.32%,AUC改善0.08。结论 AI模型辅助可帮助医师提高鼻骨骨折的诊断能力,具备一定的临床使用价值。

Objective The artificial intelligence(AI)model was developed using the fusion technology of traditional omics and deep learning,and its diagnostic value for nasal fracture was explored.Methods With selection of 252 patients with nasal bone multi-slice spiral CT(MSCT)scanning admitted to the hospital from May 2021 to July 2022 as the research subjects,end-to-end deep learning features were extracted through deep learning techniques,and traditional omics features were extracted using pyradiomics analysis software.With the screening and fusion of all features,the AI prediction model was established by inputting the classification predictor,and its diagnostic efficiency for nasal fracture was tested.Results With the assistance of AI,when clinical physicians read the images,the sensitivity improved by 11.65%,the specificity improved by 16.80%,and the area under the curve(AUC)improved by 0.14;When radiologists read the images,the sensitivity improved by 9.32%and the AUC improved by 0.08.Conclusion The assist of AI model can help doctors improve the diagnostic ability for nasal bone fracture,which has certain clinical value.