Automatic segmentation of human knee anatomy by a convolutional neural network applying a 3D MRI protocol

被引：18
|
作者：
Kulseng, Carl Petter Skaar [1 ]
Nainamalai, Varatharajan [2 ]
Grovik, Endre [3 ,4 ]
Geitung, Jonn-Terje [1 ,5 ,6 ]
Aroen, Asbjorn [7 ,8 ]
Gjesdal, Kjell-Inge [1 ,2 ,6 ]
机构：
[1] Sunnmore MR Klin, Langelandsvegen 15, N-6010 Alesund, Norway
[2] Norwegian Univ Sci & Technol, Larsgaardvegen 2, N-6025 Alesund, Norway
[3] Norwegian Univ Sci & Technol, Hogskoleringen 5, N-7491 Trondheim, Norway
[4] More & Romsdal Hosp Trust, Postboks 1600, N-6025 Alesund, Norway
[5] Univ Oslo, Fac Med, Klaus Torgards Vei 3, N-0372 Oslo, Norway
[6] Akershus Univ Hosp, Dept Radiol, Postboks 1000, N-1478 Lorenskog, Norway
[7] Akershus Univ Hosp, Inst Clin Med, Dept Orthoped Surg, Problemveien 7, N-0315 Oslo, Norway
[8] Norwegian Sch Sport Sci, Oslo Sports Trauma Res Ctr, Postboks 4014, N-0806 Oslo, Norway
来源：
BMC MUSCULOSKELETAL DISORDERS | 2023年 / 24卷 / 01期
关键词：
Magnetic Resonance Imaging; Musculoskeletal; Deep learning; Knee images segmentation; Visualization;
D O I：
10.1186/s12891-023-06153-y
中图分类号：
R826.8 [整形外科学]; R782.2 [口腔颌面部整形外科学]; R726.2 [小儿整形外科学]; R62 [整形外科学（修复外科学）];
学科分类号：
摘要：
BackgroundTo study deep learning segmentation of knee anatomy with 13 anatomical classes by using a magnetic resonance (MR) protocol of four three-dimensional (3D) pulse sequences, and evaluate possible clinical usefulness.MethodsThe sample selection involved 40 healthy right knee volumes from adult participants. Further, a recently injured single left knee with previous known ACL reconstruction was included as a test subject. The MR protocol consisted of the following 3D pulse sequences: T1 TSE, PD TSE, PD FS TSE, and Angio GE. The DenseVNet neural network was considered for these experiments. Five input combinations of sequences (i) T1, (ii) T1 and FS, (iii) PD and FS, (iv) T1, PD, and FS and (v) T1, PD, FS and Angio were trained using the deep learning algorithm. The Dice similarity coefficient (DSC), Jaccard index and Hausdorff were used to compare the performance of the networks.ResultsCombining all sequences collectively performed significantly better than other alternatives. The following DSCs (+/- standard deviation) were obtained for the test dataset: Bone medulla 0.997 (+/- 0.002), PCL 0.973 (+/- 0.015), ACL 0.964 (+/- 0.022), muscle 0.998 (+/- 0.001), cartilage 0.966 (+/- 0.018), bone cortex 0.980 (+/- 0.010), arteries 0.943 (+/- 0.038), collateral ligaments 0.919 (+/- 0.069), tendons 0.982 (+/- 0.005), meniscus 0.955 (+/- 0.032), adipose tissue 0.998 (+/- 0.001), veins 0.980 (+/- 0.010) and nerves 0.921 (+/- 0.071). The deep learning network correctly identified the anterior cruciate ligament (ACL) tear of the left knee, thus indicating a future aid to orthopaedics.ConclusionsThe convolutional neural network proves highly capable of correctly labeling all anatomical structures of the knee joint when applied to 3D MR sequences. We have demonstrated that this deep learning model is capable of automatized segmentation that may give 3D models and discover pathology. Both useful for a preoperative evaluation.
引用
收藏
页数：12
相关论文
共 50 条
  • [21] A New Convolutional Neural Network Architecture for Automatic Segmentation of Overlapping Human Chromosomes
    Sifan Song
    Tianming Bai
    Yanxin Zhao
    Wenbo Zhang
    Chunxiao Yang
    Jia Meng
    Fei Ma
    Jionglong Su
    Neural Processing Letters, 2022, 54 : 285 - 301
  • [22] Deep MRI brain extraction: A 3D convolutional neural network for skull stripping
    Kleesiek, Jens
    Urban, Gregor
    Hubert, Alexander
    Schwarz, Daniel
    Maier-Hein, Klaus
    Bendszus, Martin
    Biller, Armin
    NEUROIMAGE, 2016, 129 : 460 - 469
  • [23] Deformable MRI-Ultrasound Registration Using 3D Convolutional Neural Network
    Sun, Li
    Zhang, Songtao
    SIMULATION, IMAGE PROCESSING, AND ULTRASOUND SYSTEMS FOR ASSISTED DIAGNOSIS AND NAVIGATION, 2018, 11042 : 152 - 158
  • [24] Classification of MRI Migraine Medical Data Using 3D Convolutional Neural Network
    Ng, Hwei Geok
    Kerzel, Matthias
    Mehnert, Jan
    May, Arne
    Wermter, Stefan
    ARTIFICIAL NEURAL NETWORKS AND MACHINE LEARNING - ICANN 2018, PT III, 2018, 11141 : 300 - 309
  • [25] RP-Net: A 3D Convolutional Neural Network for Brain Segmentation From Magnetic Resonance Imaging
    Wang, Liansheng
    Xie, Cong
    Zeng, Nianyin
    IEEE ACCESS, 2019, 7 : 39670 - 39679
  • [26] A New Convolutional Neural Network Architecture for Automatic Segmentation of Overlapping Human Chromosomes
    Song, Sifan
    Bai, Tianming
    Zhao, Yanxin
    Zhang, Wenbo
    Yang, Chunxiao
    Meng, Jia
    Ma, Fei
    Su, Jionglong
    NEURAL PROCESSING LETTERS, 2022, 54 (01) : 285 - 301
  • [27] Hepatic artery segmentation with 3D convolutional neural networks
    Kock, Farina
    Chlebus, Grzegorz
    Thielke, Felix
    Schenk, Andrea
    Meine, Hans
    MEDICAL IMAGING 2022: COMPUTER-AIDED DIAGNOSIS, 2022, 12033
  • [28] 3D Hyper-dense Connected Convolutional Neural Network for Brain Tumor Segmentation
    Qamar, Saqib
    Jin, Hai
    Zheng, Ran
    Ahmad, Parvez
    2018 14TH INTERNATIONAL CONFERENCE ON SEMANTICS, KNOWLEDGE AND GRIDS (SKG), 2018, : 123 - 130
  • [29] 3D Thyroid Segmentation in CT Using Self-attention Convolutional Neural Network
    He, Xiuxiu
    Guo, Bang Jun
    Lei, Yang
    Liu, Yingzi
    Wang, Tonghe
    Curran, Walter J.
    Zhang, Long Jiang
    Liu, Tian
    Yang, Xiaofeng
    MEDICAL IMAGING 2020: COMPUTER-AIDED DIAGNOSIS, 2020, 11314
  • [30] 3D convolutional neural network for semantic scene segmentation based on unstructured point clouds
    Zhang R.
    Wang Y.
    Li G.
    Han Z.
    Li J.
    Li C.
    International Journal of Performability Engineering, 2018, 14 (07) : 1503 - 1512
12345