From the Department of Diagnostic Imaging, National University Hospital, 5 Lower Kent Ridge Rd, Singapore 119074 (D.S.W.L., A.M., A.J.L.C., D.S.Y.S., S.E.E., H.Y.O., P.J., W.C.T., V.M.K., Y.M.W., Y.L.T., S.B., E.C.T., S.T.Q., J.T.P.D.H.); Department of Diagnostic Radiology (A.M., S.E.E., P.J., Y.L.T., S.T.Q., J.T.P.D.H.), NUS Graduate School, Integrative Sciences and Engineering Programme (L.Z.), Department of Computer Science, School of Computing (W.Z., B.C.O.), and Biostatistics Unit, Yong Loo Lin School of Medicine (Q.V.Y., Y.H.C.), National University of Singapore, Singapore; Department of Radiology, Qatif Central Hospital, Qatif, Saudi Arabia (D.A.R.A.); Department of Orthopaedic Surgery, National University Health System, Singapore (J.H.T., N.K.); and Department of Radiological Sciences, University of California, Irvine, Orange, Calif (H.Y.).
Radiology. 2022 Oct;305(1):160-166. doi: 10.1148/radiol.220076. Epub 2022 Jun 14.
Background Lumbar spine MRI studies are widely used for back pain assessment. Interpretation involves grading lumbar spinal stenosis, which is repetitive and time consuming. Deep learning (DL) could provide faster and more consistent interpretation. Purpose To assess the speed and interobserver agreement of radiologists for reporting lumbar spinal stenosis with and without DL assistance. Materials and Methods In this retrospective study, a DL model designed to assist radiologists in the interpretation of spinal canal, lateral recess, and neural foraminal stenoses on lumbar spine MRI scans was used. Randomly selected lumbar spine MRI studies obtained in patients with back pain who were 18 years and older over a 3-year period, from September 2015 to September 2018, were included in an internal test data set. Studies with instrumentation and scoliosis were excluded. Eight radiologists, each with 2-13 years of experience in spine MRI interpretation, reviewed studies with and without DL model assistance with a 1-month washout period. Time to diagnosis (in seconds) and interobserver agreement (using Gwet κ) were assessed for stenosis grading for each radiologist with and without the DL model and compared with test data set labels provided by an external musculoskeletal radiologist (with 32 years of experience) as the reference standard. Results Overall, 444 images in 25 patients (mean age, 51 years ± 20 [SD]; 14 women) were evaluated in a test data set. DL-assisted radiologists had a reduced interpretation time per spine MRI study, from a mean of 124-274 seconds (SD, 25-88 seconds) to 47-71 seconds (SD, 24-29 seconds) ( < .001). DL-assisted radiologists had either superior or equivalent interobserver agreement for all stenosis gradings compared with unassisted radiologists. DL-assisted general and in-training radiologists improved their interobserver agreement for four-class neural foraminal stenosis, with κ values of 0.71 and 0.70 (with DL) versus 0.39 and 0.39 (without DL), respectively (both < .001). Conclusion Radiologists who were assisted by deep learning for interpretation of lumbar spinal stenosis on MRI scans showed a marked reduction in reporting time and superior or equivalent interobserver agreement for all stenosis gradings compared with radiologists who were unassisted by deep learning. © RSNA, 2022 . See also the editorial by Hayashi in this issue.
背景 腰椎磁共振成像（MRI）研究广泛用于腰痛评估。其解读涉及分级腰椎管狭窄症，这是一个重复且耗时的过程。深度学习（DL）可以提供更快、更一致的解读。目的 评估有和没有 DL 辅助的放射科医生报告腰椎管狭窄症的速度和观察者间一致性。材料与方法 在这项回顾性研究中，使用了一种专门设计的 DL 模型，用于辅助放射科医生解读腰椎 MRI 扫描的椎管、侧隐窝和神经孔狭窄。从 2015 年 9 月至 2018 年 9 月，纳入了一项为期 3 年的、随机选择的腰痛患者腰椎 MRI 研究的内部测试数据集。排除了有内固定和脊柱侧凸的研究。8 名放射科医生，每人有 2-13 年的脊柱 MRI 解读经验，在 1 个月的洗脱期内，分别使用和不使用 DL 模型对研究进行了评估。对每位放射科医生进行了有和没有 DL 模型的狭窄分级诊断时间（以秒为单位）和观察者间一致性（使用 Gwet κ）评估，并与由一位有 32 年经验的外部肌肉骨骼放射科医生提供的测试数据集标签（作为参考标准）进行了比较。结果 在测试数据集中，共评估了 25 名患者的 444 张图像（平均年龄 51 岁±20[标准差]；14 名女性）。使用 DL 辅助的放射科医生解读每例腰椎 MRI 研究的时间明显缩短，从平均 124-274 秒（标准差 25-88 秒）降至 47-71 秒（标准差 24-29 秒）（<.001）。与未使用 DL 的放射科医生相比，使用 DL 辅助的放射科医生在所有狭窄分级中具有更好或同等的观察者间一致性。使用 DL 辅助的普通放射科医生和住院医师对四分级神经孔狭窄的观察者间一致性有所提高，κ 值分别为 0.71 和 0.70（使用 DL）与 0.39 和 0.39（不使用 DL）（均<.001）。结论 与未使用 DL 的放射科医生相比，使用腰椎 MRI 扫描 DL 辅助解读的放射科医生报告时间明显缩短，在所有狭窄分级中具有更好或同等的观察者间一致性。RSNA，2022 年。也可参见本期 Hayashi 编辑的社论。
