不同层数的Mask-RCNN模型自动检测成釉细胞瘤效能的比较

doi:10.13701/j.cnki.kqyxyj.2023.12.012

口腔医学研究 ›› 2023, Vol. 39 ›› Issue (12): 1092-1096.DOI: 10.13701/j.cnki.kqyxyj.2023.12.012

不同层数的Mask-RCNN模型自动检测成釉细胞瘤效能的比较

赖丹琳, 许亮, 倪涧钊, 朱小峰^*, 黄晓红^*

福建医科大学附属第一医院口腔医学中心,福建医科大学附属第一医院滨海院区国家区域医疗中心口腔医学中心福建福州 350212

收稿日期:2023-09-07 出版日期:2023-12-28 发布日期:2023-12-25
通讯作者: * 朱小峰,E-mail: dentzxf@163.com黄晓红,E-mail: 13905006768@139.com
作者简介:赖丹琳(1991~ ),女,福州人,硕士,住院医师,研究方向:口腔颌面外科学。
基金资助:
福建省教育厅中青年教师教育科研项目(编号:JAT200161)

Comparison of Efficacy of Mask-RCNN Models with Different Layers in Automatic Detection of Ameloblastoma

LAI Danlin, XU Liang, NI Jianzhao, ZHU Xiaofeng^*, HUANG Xiaohong^*

Center of Stomatology, the First Affiliated Hospital, Fujian Medical University, Center of Stomatology,National Regional Medical Center, Bin hai Campus of the First Affiliated Hospital, Fujian Medical University, Fuzhou 350212, China

Received:2023-09-07 Online:2023-12-28 Published:2023-12-25

摘要/Abstract

摘要： 目的:基于人工智能深度学习,比较Mask-RCNN不同层数的ResNet模型在成釉细胞瘤CT图像检测中的效能。方法:回顾性收集2018年4月~2020年8月福建医科大学附属第一医院成釉细胞瘤患者的CT影像数据,按照标准将79名患者纳入研究。经过预处理后,共得到3566张图像,按照8∶1∶1的比例将其随机分为训练集、验证集以及测试集。采用ResNet-18、ResNet-50和ResNet-101模型进行训练,实现肿瘤的自动检测并通过Dice系数、平均精确度AP及检测时间等评价指标进行分析。结果:与ResNet-18及ResNet-50相比,ResNet-101模型自动检测的效果最好,其Dice系数为0.87,平均精确度AP(IOU 0.50∶0.95)为0.74。但该模型所需的检测时间最长,需要0.33 s。结论:不同层数的Mask-RCNN模型均可较好地实现对成釉细胞瘤的自动检测诊断,其中ResNet-101检测效果最好,但相应地需要更长的时间。

关键词: 成釉细胞瘤, 深度学习, Mask-RCNN, ResNet

Abstract: Objective: To compare the performance of different layers of ResNet models in CT image detection of Ameloblastoma(AME), which uses artificial intelligence deep learning with Mask-RCNN. Methods: A retrospective analysis was carried out on CT image data of AME patients at the First Affiliated Hospital of Fujian Medical University from April 2018 to August 2020. In the study, 79 patients were included based on specific criteria. After preprocessing, we obtained 3566 images, which were divided into training, validation, and test sets in an 8∶1∶1 ratio randomly. After ResNet-18, ResNet-50, and ResNet-101 models were employed for training, automatic tumor detection was achieved, and metrics such as Dice coefficient, average accuracy, and detection time were evaluated. Results: Compared to ResNet-18 and ResNet-50, the ResNet-101 model exhibited the best performance in automatic detection, with a Dice coefficient of 0.87 and an average accuracy AP(IOU 0.50∶0.95) of 0.74. However, this model also had the longest detection time, approximately 0.33 seconds. Conclusion: The Mask-RCNN model with different layers can realize the automatic detection of AME.Among the models studied, ResNet-101 demonstrated the most favorable detection performance, though a longer detection time.

Key words: ameloblastoma, deep learning, Mask-RCNN, ResNet

赖丹琳, 许亮, 倪涧钊, 朱小峰, 黄晓红. 不同层数的Mask-RCNN模型自动检测成釉细胞瘤效能的比较[J]. 口腔医学研究, 2023, 39(12): 1092-1096.

LAI Danlin, XU Liang, NI Jianzhao, ZHU Xiaofeng, HUANG Xiaohong. Comparison of Efficacy of Mask-RCNN Models with Different Layers in Automatic Detection of Ameloblastoma[J]. Journal of Oral Science Research, 2023, 39(12): 1092-1096.

参考文献

[1] 罗号,袁壮,吴开柳,等.92例上颌骨成釉细胞瘤不同手术方式的效果评价[J].上海口腔医学,2022,31(1):71-74.
[2] Hendra FN, Cann EMV, Helder MN, et al. Global incidence and profile of ameloblastoma: a systematic review and meta-analysis [J]. Oral Dis, 2020, 26(1):12-21.
[3] Saluja TS, Hosalkar R. Reconnoitre ameloblastic carcinoma: A prognostic update [J]. Oral Oncol, 2018, 77:118-124.
[4] Effiom OA, Ogundana OM, Akinshipo AO, et al. Ameloblastoma: current etiopathological concepts and management [J]. Oral Dis, 2018, 24(3):307-316.
[5] Hendra FN, Natsir Kalla DS, Van Cann EM, et al. Radical vs conservative treatment of intraosseous ameloblastoma: Systematic review and meta-analysis [J]. Oral Dis, 2019, 25(7):1683-1696.
[6] 聂根定,梁燕,王立冬,等.CBCT影像组学在颌骨成釉细胞瘤术前诊断中的应用研究[J].口腔医学研究,2023,39(3):211-216.
[7] 田娟秀,刘国才,谷珊珊,等.医学图像分析深度学习方法研究与挑战[J].自动化学报,2018,44(3):401-424.
[8] 张祖燕.口腔颌面医学影像诊断学[M].第7版.北京人民卫生出版社,2020:97-98.
[9] He K, Zhang X, Ren S, et al.Deep residual learning for image recognition [J]. Appl Sci, 2022, 12(18): 8972.
[10] Behpour S, Kitani KM, Ziebart BD. Adversarially optimizing intersection over union for object localization tasks [J].2017.
[11] Litjens G, Kooi T, Bejnordi BE, et al. A survey on deep learning in medical image analysis [J]. Med Image Anal, 2017, 42(9):60-88.
[12] Chai ZK, Mao L, Chen H, et al. Improved diagnostic accuracy of ameloblastoma and odontogenic keratocyst on cone-beam CT by artificial intelligence [J]. Front Oncol, 2022, 11:793417.
[13] 李敏,慕创创,张建运,等.深度学习在基于曲面体层片的成釉细胞瘤及牙源性角化囊肿鉴别诊断中的应用[J].中国医学科学院学报,2023, 45(2): 273-279.
[14] Ekert T, Krois J, Meinhold L, et al. Deep learning for the radiographic detection of apical lesions [J]. J Endod, 2019, 45(7):917-922.
[15] Liu Z, Liu J, Zhou Z, et al. Differential diagnosis of ameloblastoma and odontogenic keratocyst by machine learning of panoramic radiographs [J]. Int J Comput Assist Radiol Surg, 2021, 16(3):415-422.
[16] Yang H, Jo E, Kim HJ, et al. Deep learning for automated detection of cyst and tumors of the jaw in panoramic radiographs [J]. J Clin Med, 2020, 9(6):1839.
[17] Kwon O, Yong TH, Kang SR, et al. Automatic diagnosis for cysts and tumors of both jaws on panoramic radiographs using a deep convolution neural network [J]. Dentomaxillofac Radiol, 2020, 49(8):20200185.
[18] Wiwiek P, Siriwan S. Application of convolutional neural network in the diagnosis of jaw tumors [J]. Healthc Inform Res, 2018, 24(3):236.
[19] 田涛.成釉细胞瘤X线及CT对比分析[C]//中华口腔医学会口腔颌面放射专业委员会.2020年中华口腔医学会口腔颌面放射专业委员会第十八次全国口腔颌面医学影像学专题研讨会论文汇编.2020:31.
[20] Ariji Y, Morita M, Katsumata A, et al. Imaging features contributing to the diagnosis of ameloblastomas and keratocystic odontogenic tumours: logistic regression analysis [J]. Dentomaxillofac Radiol, 2011, 40(3):133-140.
[21] Abdolali F, Zoroofi RA, Otake Y, et al. Automatic segmentation of maxillofacial cysts in cone beam CT images [J]. Comput Biol Med, 2016, 72:108-119.
[22] Andrews S, Hamarneh G. Multi-region probabilistic dice similarity coefficient using the aitchison distance and bipartite graph matching [J]. Comput Sci, 2015.
[23] 杨宇.基于深度学习的肺炎检测算法研究与应用[D].贵州大学,2023.
[24] Kim HE, Cosa-Linan A, Santhanam N, et al. Transfer learning for medical image classification: a literature review [J]. BMC Med Imaging, 2022, 22(1):69.
[25] Lu T, Zhang X, Huang Y, et al. pFISTA-SENSE-ResNet for parallel MRI reconstruction [J]. J Magn Reson, 2020, 318:106790.

不同层数的Mask-RCNN模型自动检测成釉细胞瘤效能的比较

Comparison of Efficacy of Mask-RCNN Models with Different Layers in Automatic Detection of Ameloblastoma

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

[1]	温佳欢, 傅裕杰, 任根强, 陈宇飞, 张旗. 基于深度学习的点云补全网络修复上颌磨牙钙化根管的研究[J]. 口腔医学研究, 2023, 39(5): 455-459.
[2]	杨可, 孙雅楠, 胡雅英, 吕宜楠, 郑小风, 李易玮, 张佳莉. cyclin D1表达模式深度学习识别和预测模型对HPV阴性口腔鳞状细胞癌和口咽鳞状细胞癌患者预后的预测作用[J]. 口腔医学研究, 2023, 39(4): 308-315.
[3]	聂根定, 梁燕, 王立冬, 马文, 黎明. CBCT影像组学在颌骨成釉细胞瘤术前诊断中的应用研究[J]. 口腔医学研究, 2023, 39(3): 211-216.
[4]	满其文, 王含琪. Carnoy's液在颌骨囊性病变辅助治疗中的应用[J]. 口腔医学研究, 2023, 39(12): 1046-1049.
[5]	吴天赐, 郁佳鑫, 黄晓峰, 陈盛, 王育新, 蒲玉梅. 深度学习图片分类模型ResNet-18用于判定口腔鳞状细胞癌浸润方式的初步研究[J]. 口腔医学研究, 2023, 39(10): 917-922.
[6]	李婷婷, 王超凡, 赵吉宏. 开窗减压术治疗下颌骨单囊型成釉细胞瘤1例[J]. 口腔医学研究, 2022, 38(3): 291-294.
[7]	朱红标, 刘强冬, 曾子强, 娄伟明, 戴芳, 吴婧婷, 邓恬, 邓立彬, 宋莉. 基于口内数码照图像深度学习的牙周病早期筛查研究[J]. 口腔医学研究, 2022, 38(11): 1092-1095.
[8]	帕克扎提·色依提, 王铁梅, 徐子能, 白海龙, 丁鹏, 刘澍, 滕跃辉, 冯英连, 王蓉, 单珊, 钟双泽. 基于深度学习在曲面体层图像中人工智能辅助诊断系统初步研究[J]. 口腔医学研究, 2021, 37(9): 845-849.
[9]	王蓉, 王铁梅. 人工智能在口腔颌面医学影像中的研究探讨[J]. 口腔医学研究, 2020, 36(7): 619-621.
[10]	韩生伟, 韩伟. 人工智能技术在口腔医学领域的应用进展[J]. 口腔医学研究, 2020, 36(6): 519-522.
[11]	张立刚, 易杰, 姚礼, 黄桂林, 胡小华, 张霓霓, 代敏. 单囊型成釉细胞瘤行开窗减压术联合刮治术的临床研究[J]. 口腔医学研究, 2020, 36(4): 341-345.
[12]	段聪聪, 韩冰. 颌骨囊性病变相关手术治疗的发展现状[J]. 口腔医学研究, 2020, 36(10): 898-901.
[13]	曹雅, 王铁梅, 韩超, 孙国文, 杨功鑫, 吴大明, 林梓桐, 王从跃. 促结缔组织增生型成釉细胞瘤的临床及CT影像学分析[J]. 口腔医学研究, 2018, 34(7): 751-755.
[14]	冯英连,张磊,王铁梅,林梓桐,刘澍,黄晓峰. 3例颌骨角化型成釉细胞瘤:病例报道及影像学特征分析[J]. 口腔医学研究, 2018, 34(12): 1368-1372.
[15]	贾雪萍,唐恩溢. 彻底刮治术治疗下颌骨成釉细胞瘤的临床研究[J]. 口腔医学研究, 2017, 33(5): 567-570.