探索深度学习模型在早期CT检测高级别转移性硬膜外脊髓压迫中的潜力及其对治疗延迟的影响。
Exploring the Potential of a Deep Learning Model for Early CT Detection of High-Grade Metastatic Epidural Spinal Cord Compression and Its Impact on Treatment Delays.
作者信息
Hallinan James Thomas Patrick Decourcy, Wu Junran, Liu Changshuo, Tran Hien Anh, Lim Noah Tian Run, Makmur Andrew, Ong Wilson, Wang Shilin, Teo Ee Chin, Chan Yiong Huak, Hey Hwee Weng Dennis, Lau Leok-Lim, Thambiah Joseph, Wong Hee-Kit, Liu Gabriel, Kumar Naresh, Ooi Beng Chin, Tan Jiong Hao Jonathan
机构信息
Department of Diagnostic Imaging, National University Hospital, National University Health System, Singapore 119074, Singapore.
Department of Diagnostic Radiology, Yong Loo Lin School of Medicine, National University of Singapore, Singapore 117597, Singapore.
出版信息
Cancers (Basel). 2025 Jun 28;17(13):2180. doi: 10.3390/cancers17132180.
BACKGROUND
Delay in diagnosing metastatic epidural spinal cord compression (MESCC) adversely impacts clinical outcomes. High-grade MESCC is frequently overlooked on routine staging CT scans. We aim to assess the potential of our deep learning model (DLM) in detecting high-grade MESCC and reducing diagnostic delays.
METHODS
This retrospective review analyzed 140 patients with surgically treated MESCC between C7 and L2 during 2015-2022. An experienced radiologist (serving as the reference standard), a consultant spine surgeon, and the DLM independently classified staging CT scans into high-grade MESCC or not. The findings were compared to original radiologist (OR) reports; inter-rater agreement was assessed. Diagnostic delay referred to the number of days elapsed from CT to diagnostic MRI scan.
RESULTS
Overall, 95/140 (67.8%) patients had preoperative CT scans. High-grade MESCC was identified in 84/95 (88.4%) of the scans by the radiologist (reference standard), but in only 32/95 (33.7%) of the preoperative scans reported by the OR. There was almost perfect agreement between the radiologist and the surgeon (kappa = 0.947, 95% CI = 0.893-1.000) ( < 0.001), and between the radiologist and the DLM (kappa = 0.891, 95% CI = 0.816-0.967) ( < 0.001). In contrast, inter-observer agreement between the OR and all other readers was slight (kappa range = 0.022-0.125). Diagnostic delay was potentially reduced by 20 ± 28 (range = 1-131) days.
CONCLUSIONS
The original radiologist reports frequently missed high-grade MESCC in staging CT. Our DLM for CT diagnosis of high-grade MESCC showed almost perfect inter-rater agreement with two experienced reviewers. This study is the first to demonstrate that the DLM could help reduce diagnostic delays. Further prospective research is required to understand its precise role in improving the early diagnosis/treatment of MESCC.
背景
转移性硬膜外脊髓压迫症（MESCC）诊断延迟会对临床结果产生不利影响。常规分期CT扫描时常会忽略高级别MESCC。我们旨在评估深度学习模型（DLM）在检测高级别MESCC及减少诊断延迟方面的潜力。
方法
本回顾性研究分析了2015年至2022年间140例接受C7至L2节段手术治疗的MESCC患者。一名经验丰富的放射科医生（作为参考标准）、一名脊柱外科顾问医生和DLM分别独立将分期CT扫描结果分类为高级别MESCC或非高级别MESCC。将结果与放射科医生的原始报告进行比较；评估观察者间的一致性。诊断延迟指从CT扫描到诊断性MRI扫描所经过的天数。
结果
总体而言，95/140（67.8%）例患者进行了术前CT扫描。放射科医生（参考标准）在84/95（88.4%）的扫描中识别出高级别MESCC，但在放射科医生原始报告的术前扫描中仅识别出32/95（33.7%）。放射科医生与外科医生之间几乎完全一致（kappa = 0.947，95%CI = 0.893 - 1.000）（<0.001），放射科医生与DLM之间也几乎完全一致（kappa = 0.891，95%CI = 0.816 - 0.967）（<0.001）。相比之下，放射科医生原始报告与其他所有读者之间的观察者间一致性较差（kappa范围 = 0.022 - 0.125）。诊断延迟可能减少了20±28（范围 = 1 - 131）天。
结论
放射科医生的原始报告在分期CT中经常漏诊高级别MESCC。我们用于CT诊断高级别MESCC的DLM与两名经验丰富的审阅者之间显示出几乎完全的观察者间一致性。本研究首次表明DLM有助于减少诊断延迟。需要进一步的前瞻性研究来了解其在改善MESCC早期诊断/治疗中的精确作用。
Zotero 插件