Vorontsov Eugene, Cerny Milena, Régnier Philippe, Di Jorio Lisa, Pal Christopher J, Lapointe Réal, Vandenbroucke-Menu Franck, Turcotte Simon, Kadoury Samuel, Tang An
Department of Radiology (M.C., A.T.) and Department of Surgery, Hepatopancreatobiliary and Liver Transplantation Division (R.L., F.V., S.T.), Centre Hospitalier de l'Université de Montréal (CHUM), 1000 rue Saint-Denis, Montréal, QC, Canada H2X 0C2; Montreal Institute for Learning Algorithms (MILA), Montréal, Canada (E.V., C.J.P.); École Polytechnique, Montréal, Canada (E.V., C.J.P., S.K.); Centre de Recherche du Centre Hospitalier de l'Université de Montréal (CRCHUM), Montréal, Canada (M.C., P.R., S.T., S.K., A.T.); and Imagia Cybernetics, Montréal, Canada (L.D.J.).
Radiol Artif Intell. 2019 Mar 13;1(2):180014. doi: 10.1148/ryai.2019180014. eCollection 2019 Mar.
To evaluate the performance, agreement, and efficiency of a fully convolutional network (FCN) for liver lesion detection and segmentation at CT examinations in patients with colorectal liver metastases (CLMs).
This retrospective study evaluated an automated method using an FCN that was trained, validated, and tested with 115, 15, and 26 contrast material-enhanced CT examinations containing 261, 22, and 105 lesions, respectively. Manual detection and segmentation by a radiologist was the reference standard. Performance of fully automated and user-corrected segmentations was compared with that of manual segmentations. The interuser agreement and interaction time of manual and user-corrected segmentations were assessed. Analyses included sensitivity and positive predictive value of detection, segmentation accuracy, Cohen κ, Bland-Altman analyses, and analysis of variance.
In the test cohort, for lesion size smaller than 10 mm ( = 30), 10-20 mm ( = 35), and larger than 20 mm ( = 40), the detection sensitivity of the automated method was 10%, 71%, and 85%; positive predictive value was 25%, 83%, and 94%; Dice similarity coefficient was 0.14, 0.53, and 0.68; maximum symmetric surface distance was 5.2, 6.0, and 10.4 mm; and average symmetric surface distance was 2.7, 1.7, and 2.8 mm, respectively. For manual and user-corrected segmentation, κ values were 0.42 (95% confidence interval: 0.24, 0.63) and 0.52 (95% confidence interval: 0.36, 0.72); normalized interreader agreement for lesion volume was -0.10 ± 0.07 (95% confidence interval) and -0.10 ± 0.08; and mean interaction time was 7.7 minutes ± 2.4 (standard deviation) and 4.8 minutes ± 2.1 ( < .001), respectively.
Automated detection and segmentation of CLM by using deep learning with convolutional neural networks, when manually corrected, improved efficiency but did not substantially change agreement on volumetric measurements.© RSNA, 2019
评估全卷积网络（FCN）在检测和分割结直肠癌肝转移（CLM）患者CT检查中的肝脏病变时的性能、一致性和效率。
这项回顾性研究评估了一种使用FCN的自动化方法，该方法分别采用115例、15例和26例含有261个、22个和105个病变的对比剂增强CT检查进行训练、验证和测试。放射科医生的手动检测和分割作为参考标准。将全自动和用户校正分割的性能与手动分割的性能进行比较。评估手动和用户校正分割的用户间一致性和交互时间。分析包括检测的敏感性和阳性预测值、分割准确性、Cohen κ、Bland-Altman分析和方差分析。
在测试队列中，对于直径小于10 mm（n = 30）、10 - 20 mm（n = 35）和大于20 mm（n = 40）的病变，自动化方法的检测敏感性分别为10%、71%和85%；阳性预测值分别为25%、83%和94%；Dice相似系数分别为0.14、0.53和0.68；最大对称表面距离分别为5.2、6.0和10.4 mm；平均对称表面距离分别为2.7、1.7和2.8 mm。对于手动和用户校正分割，κ值分别为0.42（95%置信区间：0.24，0.63）和0.52（95%置信区间：0.36，0.72）；病变体积的标准化读者间一致性分别为-0.10±0.07（95%置信区间）和-0.10±0.08；平均交互时间分别为7.7分钟±2.4（标准差）和4.8分钟±2.1（P <.001）。
使用卷积神经网络进行深度学习对CLM进行自动检测和分割，经手动校正后提高了效率，但在体积测量的一致性方面没有实质性改变。©RSNA，2019
