J Shanghai Jiaotong Univ Sci

J Shanghai Jiaotong Univ Sci ›› 2021, Vol. 26 ›› Issue (3): 298-305.doi: 10.1007/s12204-021-2296-2

• Medical 3D Printing and Personalised Medicine • Previous Articles     Next Articles

 Automatic Segmentation Method for Cone-Beam Computed Tomography Image of the Bone Graft Region within Maxillary Sinus Based on the Atrous Spatial Pyramid Convolution Network

Automatic Segmentation Method for Cone-Beam Computed Tomography Image of the Bone Graft Region within Maxillary Sinus Based on the Atrous Spatial Pyramid Convolution Network

XU Jiangchang1 (许江长), HE Shamin2 (何莎敏), YU Dedong2 (于德栋), WU Yiqun2 (吴轶群), CHEN Xiaojun1,3 (陈晓军)   

  1. (1. Institute of Biomedical Manufacturing and Life Quality Engineering, State Key Laboratory of Mechanical System and
    Vibration, School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China; 2. Department of
    Second Dental Centre, Shanghai Ninth People’s Hospital, Shanghai Jiao Tong University School of Medicine; College of
    Stomatology, Shanghai Jiao Tong University; National Center for Stomatology; National Clinical Research Center for
    Oral Diseases; Shanghai Key Laboratory of Stomatology, Shanghai 200011, China; 3. Institute of Medical Robotics,
    Shanghai Jiao Tong University, Shanghai 200240, China)
  2. (1. Institute of Biomedical Manufacturing and Life Quality Engineering, State Key Laboratory of Mechanical System and
    Vibration, School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China; 2. Department of
    Second Dental Centre, Shanghai Ninth People’s Hospital, Shanghai Jiao Tong University School of Medicine; College of
    Stomatology, Shanghai Jiao Tong University; National Center for Stomatology; National Clinical Research Center for
    Oral Diseases; Shanghai Key Laboratory of Stomatology, Shanghai 200011, China; 3. Institute of Medical Robotics,
    Shanghai Jiao Tong University, Shanghai 200240, China)
  • Online:2021-06-28 Published:2021-06-02
  • Contact: CHEN Xiaojun(陈晓军) E-mail:xiaojunchen@sjtu.edu.cn
  • Supported by:
    the National Key Research and Development
    Program of China (No. 2017YFB1302900),
    the National Natural Science Foundation of China
    (Nos. 81971709, M-0019, and 82011530141), the Foundation
    of Science and Technology Commission of Shanghai
    Municipality (Nos. 19510712200, and 20490740700),
    and the Shanghai Jiao Tong University Foundation
    on Medical and Technological Joint Science
    Research (Nos. ZH2018ZDA15, YG2019ZDA06, and
    ZH2018QNA23), and the 2020 Key Research Project of
    Xiamen Municipal Government (No. 3502Z20201030)

PDF (PC)

138

Abstract: Sinus floor elevation with a lateral window approach requires bone graft (BG) to ensure sufficient  bone mass, and it is necessary to measure and analyse the BG region for follow-up of postoperative patients.  However, the BG region from cone-beam computed tomography (CBCT) images is connected to the margin of  the maxillary sinus, and its boundary is blurred. Common segmentation methods are usually performed manually  by experienced doctors, and are complicated by challenges such as low efficiency and low precision. In this study,  an auto-segmentation approach was applied to the BG region within the maxillary sinus based on an atrous  spatial pyramid convolution (ASPC) network. The ASPC module was adopted using residual connections to  compose multiple atrous convolutions, which could extract more features on multiple scales. Subsequently, a  segmentation network of the BG region with multiple ASPC modules was established, which effectively improved  the segmentation performance. Although the training data were insufficient, our networks still achieved good  auto-segmentation results, with a dice coefficient (Dice) of 87.13%, an Intersection over Union (Iou) of 78.01%,  and a sensitivity of 95.02%. Compared with other methods, our method achieved a better segmentation effect,  and effectively reduced the misjudgement of segmentation. Our method can thus be used to implement automatic  segmentation of the BG region and improve doctors’ work efficiency, which is of great importance for developing  preliminary studies on the measurement of postoperative BG within the maxillary sinus.   

Key words: atrous spatial pyramid convolution (ASPC), bone graft (BG) region, medical image segmentation,
residual connection

摘要： Sinus floor elevation with a lateral window approach requires bone graft (BG) to ensure sufficient  bone mass, and it is necessary to measure and analyse the BG region for follow-up of postoperative patients.  However, the BG region from cone-beam computed tomography (CBCT) images is connected to the margin of  the maxillary sinus, and its boundary is blurred. Common segmentation methods are usually performed manually  by experienced doctors, and are complicated by challenges such as low efficiency and low precision. In this study,  an auto-segmentation approach was applied to the BG region within the maxillary sinus based on an atrous  spatial pyramid convolution (ASPC) network. The ASPC module was adopted using residual connections to  compose multiple atrous convolutions, which could extract more features on multiple scales. Subsequently, a  segmentation network of the BG region with multiple ASPC modules was established, which effectively improved  the segmentation performance. Although the training data were insufficient, our networks still achieved good  auto-segmentation results, with a dice coefficient (Dice) of 87.13%, an Intersection over Union (Iou) of 78.01%,  and a sensitivity of 95.02%. Compared with other methods, our method achieved a better segmentation effect,  and effectively reduced the misjudgement of segmentation. Our method can thus be used to implement automatic  segmentation of the BG region and improve doctors’ work efficiency, which is of great importance for developing  preliminary studies on the measurement of postoperative BG within the maxillary sinus. 

关键词: atrous spatial pyramid convolution (ASPC), bone graft (BG) region, medical image segmentation,
residual connection

CLC Number: