| 113 | 2 | 6 |
| 下载次数 | 被引频次 | 阅读次数 |
目的：探讨与评价AccuContour软件基于深度学习的自动分割方法对头颈部危及器官(OAR)自动勾画结果的准确性。方法：选取20例鼻咽癌患者的CT图像资料，应用基于深度学习的机器算法模型AccuContour软件，对图像的19个OARs(脑干、脊髓、双侧眼球、双侧晶体、双侧视神经、垂体、双侧腮腺、口腔、下颌骨、双侧颌下腺、双侧颞颌关节、甲状腺和喉)进行自动勾画，勾画结果定义为深度学习勾画的体积(VDL)，对其进行手动修改后，将修改结果定义为参考勾画(Vref)，采用相似性系数(DSC)、Jaccard系数(JAC)、平均Hausdorff距离(AHD)、体积差异比(ΔV)、最大剂量(Dmax)或平均剂量(Dmean)评价勾画结果。结果：除垂体、双侧颞颌关节及双侧视神经外，其他器官的DSC平均值均>0.7,JAC平均值均>0.6；各器官AHD均值最大为1.91 mm，最小为0.23 mm。双侧眼球、双侧晶体、下颌骨、双侧颌下腺、喉和甲状腺在剂量评估中表现优异，以ΔDmax评估的器官中，绝对平均值最大为5.81 Gy，以Dmean评估的器官中，绝对平均值最大为3.77 Gy。结论：AccuContour软件基于深度学习的自动分割方法在头颈部OAR自动勾画中能够获得较好的结果，可以用来提高临床工作效率。
Abstract：Objective: To investigate and evaluate the deep learning-based automatic segmentation of AccuContour software on the delineation for organs at risk(OARs) of head and neck. Methods: The computed tomography(CT)images of 20 patients with nasopharyngeal carcinoma were selected. A deep learning-based machine algorithm model(Accu Contour, Manteia) was applied to conduct automatic delineation for 19 OARs(brain stem, spinal cord, bilateral eyeballs, bilateral lenses, bilateral optic nerves, pituitary, bilateral parotid glands, oral cavity, mandible, bilateral submandibular glands, bilateral temporomandibular joints, thyroid and larynx). The delineation results were defined as automatic delineation(VDL), and they were defined as referred delineation(Vref) after they were manually modified. The dice similarity coefficient(DSC), Jaccard coefficient(JAC), average Hausdorff distance(AHD), diversity ration of volume(ΔV), the maximum dose(Dmax) or mean dose(Dmean) were adopted to evaluate the results of delineation. Results:Except for pituitary, bilateral temporomandibular joints and bilateral optic nerves, the average DSC of other organs was greater than 0.7, and the average JAC was more than 0.6. The maximum value of AHD mean of each organ was 1.91mm,and the minimum value was 0.23mm. The appearances of the bilateral eyeballs, bilateral lenses, mandible, bilateral submandibular glands, larynx and thyroid performed excellent in dose assessment. In the organs that were assessed by ΔDmax, the maximum absolute mean was 5.81 Gy. In the organs that were assessed by ΔDmean, the maximum absolute mean was 3.77 Gy. Conclusion: The deep learning-based automatic segmentation of AccuContour software can obtain better results in the automatic delineation for head and neck OAR, which can is used to improve clinical work efficiency.
[1]Harari PM,Song S,ToméWA.Emphasizing conformal avoidance versus target definition for IMRT planning in head-and-neck cancer[J].Int J Radiat Oncol Biol Phys,2010,77(3):950-958.
[2]Michael YC,Maria AW,Prokar D,et al.Variability in accuracy of prostate cancer segmentation among radiologists,urologists,and scientists[J].Cancer Med,2020,9(19):7172-7182.
[3]Macomber MW,Phillips M,Tarapov I,et al.Autosegmentation of prostate anatomy for radiation treatment planning using deep decision forests of radiomic features[J].Phys Med Biol,2018,63(23):235002.
[4]Cardenas CE,Yang J,Anderson BM,et al.Advances in auto-segmentation[J].Semin Radiat Oncol,2019,29(3):185-197.
[5]Sharp G,Fritscher KD,Pekar V,et al.Vision20/20:perspectives on automated image segmentation for radiotherapy[J].Med Phys,2014,41(5):050902.
[6]Sahiner B,Pezeshk A,Hadjiiski LM,et al.Deep learning in medical imaging and radiation therapy[J].Med Phys,2019,46(1):e1-e36.
[7]Seo H,Badiei Khuzani M,Vasudevan V,et al.Machine learning techniques for biomedical image segmentation:an overview of technical aspects and introduction to state-of-art applications[J].Med Phys,2020,47(5):e148-e167.
[8]Vrtovec T,Mo?nik D,Strojan P,et al.Autosegmentation of organs at risk for head and neck radiotherapy planning:From atlas-based to deep learning methods[J].Med Phys,2020,47(9):e929-e950.
[9]van Dijk L,Van den Bosch L,Aljabar P,et al.Improving automatic delineation for head and neck organs at risk by deep learning contouring[J].Radiother Oncol,2020,142:115-123.
[10]Hodapp N.The ICRU Report 83:prescribing,recording and reporting photon-beam intensitymodulated radiation therapy(IMRT)[J].Strahlenther Onkol,2012,188(1):97-99.
[11]Dice LR.Measures of the Amount of Ecologic Association Between Species[J].Ecology,1945,26(3):297-302.
[12]Jaccard P.The Distribution Of The Flora In The Alpine Zone.1[J].New Phyt,1912,11(2):37-50.
[13]Huttenlocher DP,Klanderman GA,Rucklidge WA.Comparing Images Using the Hausdorff Distance[J].IEEE Transact Patt Anal Mach Intel,1993,15(9):850-863.
[14]Ayyalusamy A,Vellaiyan S,Subramanian S,et al.Auto-segmentation of head and neck organs at risk in radiotherapy and its dependence on anatomic similarity[J].Radiat Oncol J,2019,37(2):134-142.
[15]Zhong Y,Yang Y,Fang Y,et al.A preliminary experience of implementing deep-learning based auto-segmentation in head and neck cancer:a study on real-world clinical cases[J].Front Oncol,2021,11:638197.
[16]Urago Y,Okamoto H,Kaneda T.et al.Evaluation of auto-segmentation accuracy of cloud-based artificial intelligence and atlasbased models[J].Radiat Onco,2021,16(1):175.
[17]Tong N,Gou S,Yang S,et al.Fully automatic multi-organ segmentation for head and neck cancer radiotherapy using shape representation model constrained fully convolutional neural networks[J].Med Phys,2018,45(10):4558-4567.
[18]Cardenas CE,Yang J,Anderson BM,et al.Advances in auto-segmentation[J].Semin Radiat Oncol,2019,29(3):185-197.
[19]Lei W,Mei H,Sun Z,et al.Automatic segmentation of organs-at-risk from headand-neck CT using separable convolutional neural network with hard-region-weighted loss[J].Neurocomputing,2021,442:184-199.
[20]Zijdenbos AP,Dawant BM,Margolin RA,et al.Morphometric analysis of white matter lesions in MR images:method and validation[J].IEEE Trans Med Imaging,1994,13(4):716-724.
基本信息:
中图分类号:R739.63;TP18;TP391.41
引用信息:
[1]郑庆增,戴相昆,张建春,等.基于深度学习的自动分割方法对头颈部危及器官勾画的评价[J].中国医学装备,2023,20(04):11-16.
基金信息:
北京老年医院科研专项(2022bjlnyy-xh-4)“基于深度学习的自动分割方法在肿瘤放疗中的应用研究”
2023-04-15
2023-04-15