{"title":"基于磁共振图像的高成本效益的深度学习辅助鼻咽癌复发同时检测策略(DARNDEST)的建立:一项流行地区的多中心研究。","authors":"Yishu Deng, Yingying Huang, Haijun Wu, Dongxia He, Wenze Qiu, Bingzhong Jing, Xing Lv, Weixiong Xia, Bin Li, Ying Sun, Chaofeng Li, Chuanmiao Xie, Liangru Ke","doi":"10.1186/s40644-025-00853-5","DOIUrl":null,"url":null,"abstract":"<p><strong>Background: </strong>To investigate the feasibility of detecting local recurrent nasopharyngeal carcinoma (rNPC) using unenhanced magnetic resonance images (MRI) and optimize a layered management strategy for follow-up with a deep learning model.</p><p><strong>Methods: </strong>Deep learning models based on 3D DenseNet or ResNet frames using unique sequence (T1WI, T2WI, or T1WIC) or a combination of T1WI and T2WI sequences (T1_T2) were developed to detect local rNPC. A deep-learning-assisted recurrent NPC detecting simultaneous tactic (DARNDEST) utilized DenseNet was optimized by superimposing the T1WIC model over the T1_T2 model in a specific population. Diagnostic efficacy (accuracy, sensitivity, specificity) and examination cost of a single MR scan were compared among the conventional method, T1_T2 model, and DARNDEST using McNemar's Z test.</p><p><strong>Results: </strong>No significant differences in overall accuracy, sensitivity, and specificity were found between the T1WIC model and T1WI, T2WI, or T1_T2 models in both test sets (all P > 0.0167). The DARNDEST had higher accuracy and sensitivity but lower specificity than the T1_T2 model in both the internal (accuracy, 85.91% vs. 84.99%; sensitivity, 90.36% vs. 84.26%; specificity, 82.20% vs. 85.59%) and external (accuracy, 86.14% vs. 84.16%; sensitivity, 90.32% vs. 84.95%; specificity, 82.57% vs. 83.49%) test sets. The cost of a single MR examination using DARNDEST was $330,724 (internal) and $328,971 (external) with a hypothetical cohort of 1,000 patients, relative to $313,250 of the T1_T2 model and $340,865 of the conventional method.</p><p><strong>Conclusions: </strong>Detecting local rNPC using unenhanced MRI with deep learning is feasible and DARNDEST-driven follow-up management is efficient and economic.</p>","PeriodicalId":9548,"journal":{"name":"Cancer Imaging","volume":"25 1","pages":"39"},"PeriodicalIF":3.5000,"publicationDate":"2025-03-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11931764/pdf/","citationCount":"0","resultStr":"{\"title\":\"Establishment of a deep-learning-assisted recurrent nasopharyngeal carcinoma detecting simultaneous tactic (DARNDEST) with high cost-effectiveness based on magnetic resonance images: a multicenter study in an endemic area.\",\"authors\":\"Yishu Deng, Yingying Huang, Haijun Wu, Dongxia He, Wenze Qiu, Bingzhong Jing, Xing Lv, Weixiong Xia, Bin Li, Ying Sun, Chaofeng Li, Chuanmiao Xie, Liangru Ke\",\"doi\":\"10.1186/s40644-025-00853-5\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p><strong>Background: </strong>To investigate the feasibility of detecting local recurrent nasopharyngeal carcinoma (rNPC) using unenhanced magnetic resonance images (MRI) and optimize a layered management strategy for follow-up with a deep learning model.</p><p><strong>Methods: </strong>Deep learning models based on 3D DenseNet or ResNet frames using unique sequence (T1WI, T2WI, or T1WIC) or a combination of T1WI and T2WI sequences (T1_T2) were developed to detect local rNPC. A deep-learning-assisted recurrent NPC detecting simultaneous tactic (DARNDEST) utilized DenseNet was optimized by superimposing the T1WIC model over the T1_T2 model in a specific population. Diagnostic efficacy (accuracy, sensitivity, specificity) and examination cost of a single MR scan were compared among the conventional method, T1_T2 model, and DARNDEST using McNemar's Z test.</p><p><strong>Results: </strong>No significant differences in overall accuracy, sensitivity, and specificity were found between the T1WIC model and T1WI, T2WI, or T1_T2 models in both test sets (all P > 0.0167). The DARNDEST had higher accuracy and sensitivity but lower specificity than the T1_T2 model in both the internal (accuracy, 85.91% vs. 84.99%; sensitivity, 90.36% vs. 84.26%; specificity, 82.20% vs. 85.59%) and external (accuracy, 86.14% vs. 84.16%; sensitivity, 90.32% vs. 84.95%; specificity, 82.57% vs. 83.49%) test sets. The cost of a single MR examination using DARNDEST was $330,724 (internal) and $328,971 (external) with a hypothetical cohort of 1,000 patients, relative to $313,250 of the T1_T2 model and $340,865 of the conventional method.</p><p><strong>Conclusions: </strong>Detecting local rNPC using unenhanced MRI with deep learning is feasible and DARNDEST-driven follow-up management is efficient and economic.</p>\",\"PeriodicalId\":9548,\"journal\":{\"name\":\"Cancer Imaging\",\"volume\":\"25 1\",\"pages\":\"39\"},\"PeriodicalIF\":3.5000,\"publicationDate\":\"2025-03-24\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11931764/pdf/\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Cancer Imaging\",\"FirstCategoryId\":\"3\",\"ListUrlMain\":\"https://doi.org/10.1186/s40644-025-00853-5\",\"RegionNum\":2,\"RegionCategory\":\"医学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q2\",\"JCRName\":\"ONCOLOGY\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Cancer Imaging","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.1186/s40644-025-00853-5","RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"ONCOLOGY","Score":null,"Total":0}
引用次数: 0
摘要
背景:探讨利用非增强磁共振图像(MRI)检测局部复发性鼻咽癌(rNPC)的可行性,并利用深度学习模型优化分层管理策略进行随访。方法:建立基于3D DenseNet或ResNet帧的深度学习模型,使用唯一序列(T1WI、T2WI或T1WIC)或T1WI和T2WI序列(T1_T2)的组合来检测局部rNPC。通过在特定人群中将T1WIC模型叠加到T1_T2模型上,优化了一种基于DenseNet的深度学习辅助复发性鼻咽癌检测同时策略(DARNDEST)。采用McNemar’s Z检验比较常规方法、T1_T2模型和DARNDEST单次MR扫描的诊断效能(准确性、敏感性、特异性)和检查费用。结果:T1WIC模型与T1WI、T2WI或T1_T2模型在两组测试集的总体准确性、敏感性和特异性均无显著差异(P均为0.0167)。DARNDEST的准确度和敏感性均高于T1_T2模型,但特异性低于T1_T2模型(准确度85.91% vs. 84.99%;灵敏度:90.36% vs. 84.26%;特异性,82.20% vs. 85.59%)和外部(准确性,86.14% vs. 84.16%;灵敏度:90.32% vs. 84.95%;特异性,82.57% vs. 83.49%)。使用DARNDEST进行单次MR检查的费用为330,724美元(内部)和328,971美元(外部),假设队列为1,000例患者,相对于T1_T2模型的313,250美元和传统方法的340,865美元。结论:利用非增强MRI深度学习检测局部rNPC是可行的,darndest驱动的随访管理是有效和经济的。
Establishment of a deep-learning-assisted recurrent nasopharyngeal carcinoma detecting simultaneous tactic (DARNDEST) with high cost-effectiveness based on magnetic resonance images: a multicenter study in an endemic area.
Background: To investigate the feasibility of detecting local recurrent nasopharyngeal carcinoma (rNPC) using unenhanced magnetic resonance images (MRI) and optimize a layered management strategy for follow-up with a deep learning model.
Methods: Deep learning models based on 3D DenseNet or ResNet frames using unique sequence (T1WI, T2WI, or T1WIC) or a combination of T1WI and T2WI sequences (T1_T2) were developed to detect local rNPC. A deep-learning-assisted recurrent NPC detecting simultaneous tactic (DARNDEST) utilized DenseNet was optimized by superimposing the T1WIC model over the T1_T2 model in a specific population. Diagnostic efficacy (accuracy, sensitivity, specificity) and examination cost of a single MR scan were compared among the conventional method, T1_T2 model, and DARNDEST using McNemar's Z test.
Results: No significant differences in overall accuracy, sensitivity, and specificity were found between the T1WIC model and T1WI, T2WI, or T1_T2 models in both test sets (all P > 0.0167). The DARNDEST had higher accuracy and sensitivity but lower specificity than the T1_T2 model in both the internal (accuracy, 85.91% vs. 84.99%; sensitivity, 90.36% vs. 84.26%; specificity, 82.20% vs. 85.59%) and external (accuracy, 86.14% vs. 84.16%; sensitivity, 90.32% vs. 84.95%; specificity, 82.57% vs. 83.49%) test sets. The cost of a single MR examination using DARNDEST was $330,724 (internal) and $328,971 (external) with a hypothetical cohort of 1,000 patients, relative to $313,250 of the T1_T2 model and $340,865 of the conventional method.
Conclusions: Detecting local rNPC using unenhanced MRI with deep learning is feasible and DARNDEST-driven follow-up management is efficient and economic.
Cancer ImagingONCOLOGY-RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING
CiteScore
7.00
自引率
0.00%
发文量
66
审稿时长
>12 weeks
期刊介绍:
Cancer Imaging is an open access, peer-reviewed journal publishing original articles, reviews and editorials written by expert international radiologists working in oncology.
The journal encompasses CT, MR, PET, ultrasound, radionuclide and multimodal imaging in all kinds of malignant tumours, plus new developments, techniques and innovations. Topics of interest include:
Breast Imaging
Chest
Complications of treatment
Ear, Nose & Throat
Gastrointestinal
Hepatobiliary & Pancreatic
Imaging biomarkers
Interventional
Lymphoma
Measurement of tumour response
Molecular functional imaging
Musculoskeletal
Neuro oncology
Nuclear Medicine
Paediatric.